Everybody needs somebody: Modeling social and grouping behavior on a linear programming multiple people tracker

Size: px
Start display at page:

Download "Everybody needs somebody: Modeling social and grouping behavior on a linear programming multiple people tracker"

Transcription

1 Everybody needs somebody: Modelng socal and groupng behavor on a lnear programmng multple people tracker Laura Leal-Taxe, Gerard Pons-Moll and Bodo Rosenhahn Insttute for Informaton Processng (TNT) Lebnz Unversty Hannover, Germany leal@tnt.un-hannover.de Abstract Multple people trackng conssts n detectng the subjects at each frame and matchng these detectons to obtan full trajectores. In sem-crowded envronments, pedestrans often occlude each other, makng trackng a challengng task. Most trackng methods make the assumpton that each pedestran s moton s ndependent, thereby gnorng the complex and mportant nteracton between subjects. In ths paper, we present an approach whch ncludes the nteracton between pedestrans n two ways: frst, consderng socal and groupng behavor, and second, usng a global optmzaton scheme to solve the data assocaton problem. Results on three challengng publcly avalable datasets show our method outperforms state-of-the-art trackng systems. (a) (b) (c) Fgure : Includng socal and groupng behavor to the network flow graph. (a) Constant velocty assumpton. (b) Avodance forces. (c) Group attracton forces. groupng behavor to obtan a robust tracker able to work n crowded scenaros... Related work. Introducton The optmzaton strategy deals wth the data assocaton problem, whch s usually solved on a frame-by-frame bass or one track at a tme. Several methods can be used such as Markov Chan Monte Carlo (MCMC) [5] or nference n Bayesan networks [2]. In [3] an effcent approxmatve Dynamc Programmng (DP) scheme s presented, n whch trajectores are estmated one after the other, whch obvously does not guarantee a global optmum for all trajectores. Recent works show that global optmzaton can be more relable n crowded scenes as t solves the matchng problem jontly for all tracks. The multple object trackng problem s defned as a lnear constraned optmzaton flow problem and Lnear Programmng (LP) s commonly used to fnd the global optmum. The dea was frst used for people trackng n [2], although ths method needs to know a pror the number of targets to track, whch lmts ts applcaton n real trackng stuatons. Other works formulate the trackng problem as a maxmum flow [4] or a mnmum cost problem [24, 28], both effcently solved usng LP and wth a far superor performance when compared wth DP [3]. Most trackng systems work wth the assumpton that the moton model for each target s ndependent. Ths smplfy- Multple people trackng s a key problem for many computer vson tasks, such as survellance, anmaton or actvty recognton. In crowded envronments occlusons and false detectons are common, and although there have been substantal advances n the last years, trackng s stll a challengng task. Trackng s often dvded n two steps: detecton and data assocaton. Researchers have presented mprovements on the object detector [5, 26] as well as on the optmzaton technques [4, 6] and even specfc algorthms have been developed for trackng n crowded scenes []. Though each object can be tracked separately, recent works have proven that trackng objects jontly and takng nto consderaton ther nteracton can gve much better results n complex scenes. Current research s manly focused on two aspects to explot the nteracton between pedestrans: the use of a global optmzaton strategy [4, 28] and a socal moton model [22, 27]. The focus of ths paper s to marry the concepts of global optmzaton and socal and Ths work was partally funded by the German Research Foundaton, DFG projects RO 2497/7- and RO 2524/2-.

2 ng assumpton s especally problematc n crowded scenes. In order to avod collsons and reach the chosen destnaton at the same tme, a pedestran follows a seres of socal rules or socal forces. These have been defned n what s called the Socal Force Model (SFM) [], whch has been used for abnormal crowd behavor detecton [9], crowd smulaton [2] and has only recently been appled to multple people trackng: n [25], an energy mnmzaton approach s used to predct the future poston of each pedestran consderng all the terms of the socal force model. In [22] and [7], the socal forces are ncluded n the moton model of the Kalman or Extended Kalman flter. In [] a method s presented to detect small groups of people n a crowd, but t s only recently that groupng behavor has been ncluded n a trackng framework [7, 23, 27]. In [23] groups are ncluded n a graphcal model whch contans cycles and, therefore, Dual-Decomposton s needed to fnd the soluton, whch obvously s computatonally much more expensve than usng Lnear Programmng. Moreover, the results presented n [23] are only for short tme wndows. On the other hand, the formulatons of [7,27] are predctve by nature and therefore too local and unable to deal wth trajectory changes (e.g. when people meet and stop to talk). Socal behavor models have only been ntroduced wthn a predctve framework, whch are suboptmal due to the recursve nature of flterng. Therefore, n contrast to prevous works, we propose to nclude socal and groupng models nto a global optmzaton framework whch allows us to better estmate the true maxmum a-posteror probablty of the trajectores..2. Contrbutons We present a novel approach for multple people trackng whch takes nto account the nteracton between pedestrans n two ways: frst, usng global optmzaton for data assocaton and second, ncludng socal as well as groupng behavor. The key nsght s that people plan ther trajectores n advance n order to avod collsons, therefore, a graph model whch takes nto account future and past frames s the perfect framework to nclude socal and groupng behavor. We formulate multple object trackng as a mnmum-cost network flow problem, and present a new graph model whch yelds to better results than exstng global optmzaton approaches. The socal force model (SFM) and groupng behavor (GR) are ncluded n an effcent way wthout alterng the lnearty of the problem. Results on several challengng publc datasets show the mprovement of the trackng results n crowded envronments. 2. Multple people trackng Trackng s commonly dvded n two steps: object detecton and data assocaton. Frst, the objects are detected n each frame of the sequence and second, the detectons are matched to form complete trajectores. In ths secton we defne the data assocaton problem and descrbe how to convert t to a mnmum-cost network flow problem, whch can be effcently solved usng Lnear Programmng. The dea s to buld a graph n whch the nodes represent the pedestran detectons. These nodes are fully connected to past and future observatons by edges, whch determne the relaton between two observatons wth a cost. Thereby, the matchng problem s equvalent to a mnmum-cost network flow problem: fndng the optmal set of trajectores s equvalent to sendng flow through the graph so as to mnmze the cost. 2.. Problem statement Let O = {o } be a set of object detectons wth o = (p, t ), where p = (x, y, z) s the 3D poston and t s the tme stamp. A trajectory s defned as a lst of ordered object detectons T k = {o k, o k2,, o kn }, and the goal of multple object trackng s to fnd the set of trajectores T = {T k } that best explans the detectons. Ths s equvalent to maxmzng the a-posteror probablty of T gven the set of detectons O. Assumng detectons are condtonally ndependent, the objectve functon s expressed as: T = argmax T P (o T )P (T ) () P (o T ) s the lkelhood of the detecton. In order to reduce the space of T, we make the assumpton that the trajectores cannot overlap (.e., a detecton cannot belong to two trajectores), but unlke [28], we do not defne the moton of each subject to be ndependent, therefore, we deal wth a much larger search space. We extend ths space by ncludng the followng dependences for each trajectory T k : Constant velocty assumpton: the observaton o T k depends on past observatons [o, o 2 ] Groupng behavor: If T k belongs to a group, the set of members of the group T k,gr has an nfluence on T k Avodance term: T k s affected by the set of trajectores T k,sfm whch are close to T k at some pont n tme and do not belong to the same group as T k The frst and thrd dependences are grouped nto the SFM term. The sets T k,sfm and T k,gr are dsjont,.e., a pedestran can have an attractve effect or a repulsve effect on another pedestran, but not both. Therefore, we decompose P (T ) as: P (T ) = P (T k T k,sfm T k,gr ) (2) T k T = P (T k,sfm T k )P (T k,gr T k )P (T k ) T k T

3 Frame t- where the trajectores are represented by a Markov chan: Q P (T ) = Pn (ok )... P (ok ok ) (3) Frame t+ Frame t Tk T Pk,SFM (ok,sfm ok, ok )Pk,GR (ok,gr ok, ok )... Pout (okn ) where Pk,SFM evaluates how well the socal rules are kept f ok s matched to ok, and Pk,GR descrbes how well the structure of the group s kept. Tk T log P (TGR Tk ) + e5 b5 We lnearze the objectve functon by defnng a set of flow flags f,j = {, } whch ndcate f an edge (, j) s n the path of a trajectory or not. In a mnmum cost network flow problem, the objectve s to fnd the values of the varables that mnmze the total cost of the flows over the network. Defnng the costs as negatve log-lkelhoods, and combnng Equatons (), (2) and (3), the followng objectve functon s obtaned: T = argmn log P (Tk ) log P (TSFM Tk ) e2 b Trackng wth Lnear Programmng T s (s, e ) log P (o T ) b e (e e3 b3, b j) e6 b6 (b, e ) b4 (b, t) e4 t Fgure 2: Example of a graph wth the specal source s and snk t nodes, 6 detectons whch are represented by two nodes each: the begnnng b and the end e. = argmn T + Cn, fn, + Below we detal the three types of edges present n the graphcal model and the cost for each type: C,out f,out (C,j + CSFM,,j + CGR,,j )f,j + C f,j subject to the followng constrants: Edge capactes: we assume that each detecton can only correspond to one trajectory, therefore, the edge capactes have an upper bound of uj and: fn, + f f,out + f (4) Flow conservaton at the nodes: P P fn, + f = j f,j j fj, = f,out + f (5) To map ths formulaton nto a cost-flow network, we defne G = (N, E) to be a drected network wth a cost C,j and a capacty uj assocated wth every edge (, j) E. An example of such a network s shown n Fgure 2; t contans two specal nodes, the source s and the snk t; all flow that goes through the graph starts at the s node and ends at the t node. Thereby, each flow represents a trajectory Tk and the path that each flow follows ndcates whch observatons belong to each of the trajectores. Each observaton o s represented wth two nodes, the begnnng node b N and the end node e N (see Fgure 2). A detecton edge connects b and e. Lnk edges. The edges (e, bj ) connect the end nodes e wth the begnnng nodes bj n followng frames, wth cost C,j and flow f,j, whch s f o and oj belong to Tk and f Fmax, and otherwse. f s the frame number dfference between nodes j and and Fmax s the maxmum allowed frame gap. The costs of the lnk edges represent the spatal relaton between dfferent subjects. Assumng that a subject cannot move a lot from one frame to the next, we defne the costs to be a decreasng functon of the dstance between detectons s successve frames. The tme gap between observatons s also taken nto account n order to be able to work at any frame rate, therefore velocty measures are used nstead of dstances. The veloctes are mapped to probabltes wth a Gauss error functon as shown n Equaton (6), assumng the pedestrans cannot exceed a maxmum velocty Vmax. The choce of parameter Vmax s detaled n Secton 4. E(Vt, Vmax ) = erf Vt + Vmax 2 Vmax 4 (6) The advantage of usng Equaton (6) over a lnear functon s that the probablty of lower veloctes decreases more slowly, whle the probablty for hgher veloctes decreases more rapdly. Ths s consstent wth the probablty dstrbuton of speed learned from tranng data.

4 Therefore, the cost of a lnk edge s defned as: C,j = log (P (o j o )) + C( f) ( ) pj p = log E ) t, V max + C( f) ( ) where C( f) = log B f j s the cost dependng on the frame dfference between detectons. Detecton edges. The edges (b, e ) connect the begnnng node b and end node e, wth cost C and flow f, whch s f o belongs to T k, and otherwse. ( C = log ( P det (o )) + log BB mn p BB p ) If all the costs of the edges are postve, the soluton to the mnmum-cost problem s the trval null flow. Consequently, we represent each observaton wth two nodes and a detecton edge wth negatve cost. The hgher the lkelhood of a detecton P det (o ) the more negatve the cost of the detecton edge, hence, confdent detectons are lkely to be n the path of the flow n order to mnmze the total cost. If a map of the scene s avalable, we can also nclude ths nformaton n the detecton cost. If a detecton s far away from a possble entry/ext pont, we add an extra negatve cost to the detecton edge, n order to favor that observaton to be matched. The added cost depends on the dstance to the closest entry/ext pont p BB, and s only computed for dstances hgher than BB mn =.5m. Entrance and ext edges. The edges (s, e ) connect the source s wth all the end nodes e, wth cost C n, and flow f n,, whch s f T k starts at o and otherwse. Smlarly, (b, t) connects the end node b wth snk t, wth cost C,out and flow f,out, whch s f T k ends at o, and otherwse. By connectng the s node wth the end nodes (or t to begn nodes), we make sure that when a track starts (or ends) t does not beneft from the negatve cost of the detecton edge. Therefore, we defne C n = C out = and the flow constrants of Eqs. (4) and (5). In [28], the authors propose to create the opposte edges (s, b ) and (e, t). The advantage of our formulaton s that t does not depend on P n and P out, whch are data dependent terms that need to be calculated durng optmzaton. 3. Modelng socal behavor If a pedestran does not encounter any obstacles, the natural path to follow s a straght lne. But what happens when the space gets more and more crowded and the pedestran can no longer follow the straght path? Socal nteracton between pedestrans s especally mportant when the envronment s crowded. In ths secton we consder how to nclude the socal behavor, whch we dvde nto the Socal ) Force Model (SFM) and the Group behavor (GR), nto our mnmum-cost network flow problem. 3.. Socal Force Model The socal force model states that the moton of a pedestran can be descrbed as f they were subject to socal forces. There are three man terms that need to be consdered: the desre of a pedestran to mantan a certan speed, the desre to keep a comfortable dstance from other pedestrans and the desre to reach a destnaton. Snce we cannot know a pror the destnaton of the pedestran n a real trackng system, we focus on the frst two terms. Constant velocty assumpton. The pedestran tres to keep a certan speed and drecton, therefore we assume that n t + t we have the same speed as n t and predct the pedestran s poston n t + t accordngly. Avodance term. The pedestran also tres to avod collsons and keep a comfortable dstance from other pedestrans. We model ths term as a repulson feld wth an exponental dstance-decay functon wth value α learned from tranng data. a t+ t = g m g exp ( pt+ t p t+ t m α t The only pedestrans that have ths repulson effect on subject are the ones whch do not belong to the same group as and p t+ t p t+ t m m. The dfferent avodance terms are combned lnearly. Now the predcton of the pedestran s next poston s also nfluenced by the avodance term from all pedestrans: p t+ t = p t + (vt + at+ t t) t (7) The dstance between predcton and real measurements s used to compute the cost: ( ) p C SFM,,j = log E t+ t p t+ t j t, V max In Fgure 3 we plot the probablty dstrbuton computed usng dfferent terms. Note, ths s just for vsualzaton purposes, snce we do not compute the probablty for each pont on the scene, but only for the postons where the detector has fred. There are 4 pedestrans n the scene, the purple one and 3 green ones walkng n a group. As shown n 3(b), f we only use the predcted postons (yellow heads) gven the prevous speeds, there s a collson between the purple pedestran and the green marked wth a collde. The avodance term shfts the probablty mode to a more plausble poston. )

5 (a) (b) (c) (d) (e) Fgure 3: Three green pedestrans walk n a group, the predcted postons n the next frame are marked by yellow heads. The : 72 Y: 29 purple pedestran s lnearly predcted poston (yellow head) clearly nterferes wth the trajectory of the group. Representaton of the probablty (blue s red s ) dstrbuton for the purple s next poston usng: 3(a) only dstances, 3(b) only SFM (constant velocty assumpton and avodance term), 3(c) only GR (consderng the purple pedestran belongs to the group), (d) dstances+sfm and 3(e) dstances+sfm+gr Group Model The socal behavor [] also ncludes an attracton force whch occurs when a pedestran s attracted to a frend, shop, etc. We model the attracton between members of a group. Before modelng group behavor we determne whch tracks form each group and at whch frame the group begns and ends (to deal wth splttng and formaton of groups). The dea s that f two pedestrans are close to each other over a reasonable perod of tme, they are lkely to belong to the same group. From the tranng sequence n [22], we learn the dstance and speed probablty dstrbutons of the members of a group P g vs. ndvdual pedestrans P. If m and n are two trajectores whch appear on the scene at t = [, N], we compute the flag G m,n that ndcates f m and n belong to the same group. N, P g (m, n) > N P (m, n) G m,n = t= t=, otherwse For every observaton o, we defne a group label g whch ndcates to whch group the observaton belongs to, f any. If several pedestrans form a group, they tend to keep a smlar speed, therefore, f belongs to a group, we can use the mean speed of all the other members of the group to predct the next poston for usng Equaton (7). The dstance between ths predcted poston and the real measurements s used n (6) to obtan the probablty for the groupng term. An example s shown n Fgure 3(c), where we can see that the maxmum probablty provded by the group term keeps the group confguraton. In Fgure 3(d) we show the combned probablty of the dstance and SFM nformaton, whch narrows the space of probable postons. Fnally, 3(e) represents the combned probablty of DIST, SFM and GR. 4. Implementaton detals To compute the SFM and groupng costs, we need to have nformaton about the veloctes of the pedestrans, whch can only be obtaned f we already have the trajectores. We solve ths chcken-and-egg problem teratvely; on the frst teraton, the trajectores are estmated only wth the nformaton defned n Secton 2.2. The mnmum cost soluton s found usng the Smplex algorthm [8], wth the mplementaton gven n [8]. To reduce the computatonal cost, we prune the graph usng the physcal constrants represented by the edge costs. If any of the costs C j, C SFM,,j or C GR,,j s nfnte the edge (, j) s erased from the graphcal model. For long sequences, we dvde the vdeo nto several batches and optmze for each batch. For temporal consstency, the batches have an overlap of F max = frames. Wth our non-optmzed code, the runtme for a sequence of 8 frames (4 seconds), 4837 detectons, batches of frames and 6 teratons s 3 seconds on a 3GHz machne. All parameters defned n the prevous sectons are learned from tranng data; n our case we use one sequence of the publcly avalable dataset [22]. The parameter to penalze for the frame dfference s B j =.3, the avodance term α =.5. Our approach works well for a wde range of V max and F max. Values between 5 and 25 were tested for both parameters, and the dfference between worst and best trackng accuracy obtaned was %. For all experments shown n the followng sectons, we use V max = 7 and F max =. 5. Expermental results In ths secton we show the trackng results of our method on three publcly avalable datasets and compare wth exstng state-of-the-art trackng approaches usng the CLEAR metrcs [3], DA (detecton accuracy), TA (trackng accuracy), DP (detecton precson) and TP (trackng precson). 5.. Evaluaton wth mssng data, nose and outlers We evaluate the mpact of every component of the proposed approach wth one of the sequences of the dataset [22], whch contans mages from a crowded publc place,

6 Trackng accuracy Mssng data (%) Trackng accuracy Outlers (%) Trackng accuracy Nose level Fgure 4: Experments are repeated 5 tmes and average result, maxmum and mnmum are plotted. Blue star = results wth DIST, Green damond = results wth SFM, Red square = results wth SFM+GR. From left to rght: Experment wth smulated mssng data, wth outlers, and wth random nose. (a) (b) (c) Fgure 5: Top row: Trackng results wth only DIST. Bottom row: Trackng results wth SFM+GR. Green = correct trajectores, Blue = observaton mssng from the set, Red = wrong match. 5(a) Wrong match wth DIST, corrected wth SFM. 5(b) Mssng detectons cause the matches to shft due the global optmzaton; correct result wth SFM. 5(c) Mssed detecton for subject 3 on two consecutve frames. Wth SFM, subject 2 n the frst frame (yellow arrow) s matched to subject 3 n the last frame (yellow arrow), creatng an dentty swtch; correct result wth groupng nformaton. wth several groups as well as walkng and standng pedestrans. Usng the ground truth (GT) pedestran postons as the baselne for our experments, we perform three types of tests, mssng data, outlers and nose, and compare the results obtaned wth: DIST: proposed network model wth veloctes SFM: addng the Socal Force Model (Secton 3.) SFM+GR: addng SFM and groupng behavor (Secton 3.2) Mssng data. Ths experment shows the robustness of our approach gven mssed detectons. Ths s evaluated by randomly erasng a certan percentage of detectons from the GT set. The percentages evaluated are [, 4, 8, 2, 6, 2] from the total number of detectons over the whole sequence. As we can see n Fgure 4, both SFM and SFM+GR ncrease the trackng accuracy when compared to DIST. Outlers. Wth an ntal set of detectons of GT wth 2% mssng data, tests are performed wth [,, 2, 3, 4, 5] percentage of outlers added n random postons over the ground plane. In Fgure 4, the results show that the SFM s especally mportant when the tracker s dealng wth outlers. Wth 5% of outlers, the dentty swtches wth SFM+GR are reduced 7% w.r.t the DIST results. Nose. Ths test s used to determne the performance of our approach gven nosy detectons, whch are very common manly due to small errors n the 2D-3D mappng. From the GT set wth 2% mssng data, random nose s added to every detecton. The varances of the nose tested are [,.2,.4,.6,.8,.] of the sze of the scene observed. As expected, group nformaton s the most robust to nose; f the poston of pedestran A s not correctly estmated, other pedestrans n the group wll contrbute to the estmaton of the true trajectory of A. These results corroborate that havng good behavoral

7 models becomes more mportant as the observatons deterorate. In Fgure 5 we plot the trackng results of a sequence wth 2% smulated mssng data. Only usng dstance nformaton can see dentty swtches as shown n Fgure 5(a). In Fgure 5(b) we can see how mssng data affects the matchng results. The matches are shfted, ths chan reacton s due to the global optmzaton. In both cases, the use of SFM allows the tracker to nterpolate the necessary detectons and fnd the correct trajectores. Fnally, n Fgure 5(c) we plot the wrong result whch occurs because track 3 has two consecutve mssng detectons. Even wth SFM, track 2 s swtched for 3, snce the swtch does not create extreme changes n velocty. In ths case, the groupng nformaton s key to obtanng good trackng results. More results are shown n Fgure 7, frst row. n the greedy phase of predctve approaches, where people fght for detectons. The red false detecton n the frst frame takes the detecton n the second frame that should belong to the green trajectory (whch ends n the frst frame). In the thrd frame, the red trajectory overtakes the yellow trajectory and a new blue trajectory starts where the green should have been. None of the resultng trajectores volate the SFM and GR condtons. On the other hand, our global optmzaton framework takes full advantage of the SFM and GR nformaton and correctly recovers all the trajectores. More results of the proposed algorthm can be seen n Fgure 7, last row Trackng results 5.2. Frame: Frame: Frame: Frame: 4 Frame: We evaluate the proposed algorthm on two publcly avalable datasets: a crowded town center [2] and the wellknown PETS29 dataset [9]. We compare results wth () [2] usng the results provded by the authors; (2) [28], a trackng algorthm based on network flows, for whch we use our own mplementaton of the algorthm; (3) [22], whch ncludes socal behavor, usng the code provded by the authors; (4) [27], whch ncludes socal and groupng behavor, usng our own mplementaton. For a far comparson, we do not use appearance nformaton for any method. Frame: 46 Frame: 3 Frame: Fgure 6: Predctve approaches [22, 27] (frst row) vs. Proposed method (second row) Town Center dataset We perform trackng experments on a vdeo of a crowded town center [2]. To show the mportance of socal behavor and the robustness of our algorthm at low frame rates, we track at 2.5fps (takng one every tenth frame). HOG Detectons Benfold et al. [2] Zhang et al. [28] Pellegrn et al. [22] Yamaguch et al. [27] Proposed DA TA DP TP IDsw Table : Town Center sequence. Note, the precson reported n [2] s about 9% hgher than the nput detectons precson; ths s because the authors use the moton estmaton obtaned wth a KLT feature tracker to mprove the exact poston of the detectons, whle we use the raw detectons. Stll, our algorthm reports 64% less ID swtches. As shown n Table, our algorthm outperforms [22], whch ncludes socal behavor, and [27], whch ncludes also groupng nformaton, by almost 4% n accuracy and wth 5% less ID swtches. In Fgure 6 we can see an example where [22, 27] fal. The errors are created Results on the PETS29 dataset In addton, we perform monocular trackng on the PETS29 sequence L, Vew and obtan the detectons usng the Mxture of Gaussans (MOG) background subtracton method. We obtan a trackng accuracy of 67% compared to 64.5% for Pellegrn et al. [22]. Ths dataset s very challengng from a socal behavor pont of vew, because the subjects often change drecton and groups form and splt frequently. Snce our approach s based on a probablstc framework, t s better suted for unexpected behavor changes (lke destnaton changes), where other predctve approaches fal [22, 27]. 6. Conclusons In ths paper we argued for ntegratng pedestran behavoral models n a lnear programmng framework. Our algorthm fnds the MAP estmate of the trajectores total posteror ncludng socal and groupng models usng a mnmum-cost network flow wth an mproved novel graph structure that outperforms exstng approaches. People nteracton s persstent rather than transent, hence the proposed probablstc formulaton fully explots the power of 53 5

8 Frame: 35 Frame: Frame: 8445 Frame: 445 Frame: Frame: Frame: 25 Frame: 9 Frame: 4 Fgure 7: Frst row: Results on the BIWI dataset (Secton 5.). The scene s heavly crowded, socal and groupng behavor are key to obtanng good trackng results. Last row: Results on the Town Center dataset (Secton 5.2.). behavoral models as opposed to standard predctve and recursve approaches such as Kalman flterng. Experments on three publc datasets reveal the mportance of usng socal nteracton models for trackng n dffcult condtons such as n crowded scenes wth the presence of mssed detectons, false alarms and nose. Results show that our approach s superor to state-of-the-art multple people trackers. As future work, we plan on extendng our approach to even more crowded scenaros where ndvduals cannot be detected and therefore features mght be used as n [6]. References [] S. Al and M. Shah. Floor felds for trackng n hgh densty crowded scenes. ECCV, 28. [2] B. Benfold and I. Red. Stable mult-target trackng n real-tme survellance vdeo. CVPR, 2. 7 [3] J. Berclaz, F. Fleuret, and P. Fua. Robust people trackng wth global trajectory optmzaton. CVPR, 26. [4] J. Berclaz, F. Fleuret, E. Tu retken, and P. Fua. Multple object trackng usng k-shortest paths optmzaton. TPAMI, 2. [5] M. Bretensten, F. Rechln, B. Lebe, E. Koller-Meer, and L. van Gool. Robust trackng-by-detecton usng a detector confdence partcle flter. ICCV, 29. [6] G. Brostow and R. Cpolla. Unsupervsed detecton of ndependent moton n crowds. CVPR, [7] W. Cho and S. Savarese. Multple target trackng n world coordnate wth sngle, mnmally calbrated camera. ECCV, 2. 2 [8] G. Dantzg. Lnear programmng and extensons. Prnceton Unversty Press, Prncenton, NJ, [9] J. Ferryman. Pets 29 dataset: Performance and evaluaton of trackng and survellance [] W. GE, R. Collns, and B. Ruback. Automatcally detectng the small group structure of a crowd. WACV, [] D. Helbng and P. Molna r. Socal force model for pedestran dynamcs. Physcal Revew E, 5:4282, , 5 [2] H. Jang, S. Fels, and J. Lttle. A lnear programmng approach for multple object trackng. CVPR, 27. [3] R. Kastur, D. Goldgof, P. Soundararajan, V. Manohar, J. Garofolo, M. Boonstra, V. Korzhova, and J. Zhang. Framework for performance evaluaton for face, text and vehcle detecton and trackng n vdeo: data, metrcs, and protocol. TPAMI, 3(2), [4] R. Kaucc, A. Perera, G. Brooksby, J. Kaufhold, and A. Hoogs. A unfed framework for trackng through occlusons and across sensor gaps. CVPR, 25. [5] Z. Khan, T. Balch, and F. Dellaert. Mcmc-based partcle flterng for trackng a varable number of nteractng targets. TPAMI, 25. [6] B. Lebe, K. Schndler, N. Cornels, and L. van Gool. Coupled detecton and trackng from statc cameras and movng vehcles. TPAMI, 3(), 28. [7] M. Luber, J. Stork, G. Tpald, and K. Arras. People trackng wth human moton predctons from socal forces. ICRA, 2. 2 [8] A. Makhorn. Gnu lnear programmng kt (glpk) [9] R. Mehran, A. Oyama, and M. Shah. Abnormal crowd behavor detecton usng socal force model. CVPR, [2] P. Nllus, J. Sullvan, and S. Carlsson. Mult-target trackng - lnkng denttes usng bayesan network nference. CVPR, 26. [2] N. Pelechano, J. Allbeck, and N. Badler. Controllng ndvdual agents n hgh-densty crowd smulaton. Eurographcs/ACM SIGGRAPH Symposum on Computer Anmaton, [22] S. Pellegrn, A. Ess, K. Schndler, and L. van Gool. You ll never walk alone: modelng socal behavor for mult-target trackng. ICCV, 29., 2, 5, 7 [23] S. Pellegrn, A. Ess, and L. van Gool. Improvng data assocaton by jont modelng of pedestran trajectores and groupngs. ECCV, 2. 2 [24] H. Prsavash, D. Ramanan, and C. Fowlkes. Globally-optmal greedy algorthms for trackng a varable number of objects. CVPR, 2. [25] P. Scovanner and M. Tappen. Learnng pedestran dynamcs from the real world. ICCV, [26] B. Wu and R. Nevata. Detecton and trackng of multple, partally occluded humans by bayesan combnaton of edgelet part detectors. IJCV, 75(2), 27. [27] K. Yamaguch, A. Berg, L. Ortz, and T. Berg. Who are you wth and where are you gong? CVPR, 2., 2, 7 [28] L. Zhang, Y. L, and R. Nevata. Global data assocaton for multobject trackng usng network flows. CVPR, 28., 2, 4, 7

Who are you with and Where are you going?

Who are you with and Where are you going? Who are you wth and Where are you gong? Kota Yamaguch Alexander C. Berg Lus E. Ortz Tamara L. Berg Stony Brook Unversty Stony Brook Unversty, NY 11794, USA {kyamagu, aberg, leortz, tlberg}@cs.stonybrook.edu

More information

The Development of Web Log Mining Based on Improve-K-Means Clustering Analysis

The Development of Web Log Mining Based on Improve-K-Means Clustering Analysis The Development of Web Log Mnng Based on Improve-K-Means Clusterng Analyss TngZhong Wang * College of Informaton Technology, Luoyang Normal Unversty, Luoyang, 471022, Chna wangtngzhong2@sna.cn Abstract.

More information

Forecasting the Direction and Strength of Stock Market Movement

Forecasting the Direction and Strength of Stock Market Movement Forecastng the Drecton and Strength of Stock Market Movement Jngwe Chen Mng Chen Nan Ye cjngwe@stanford.edu mchen5@stanford.edu nanye@stanford.edu Abstract - Stock market s one of the most complcated systems

More information

Detecting Global Motion Patterns in Complex Videos

Detecting Global Motion Patterns in Complex Videos Detectng Global Moton Patterns n Complex Vdeos Mn Hu, Saad Al, Mubarak Shah Computer Vson Lab, Unversty of Central Florda {mhu,sal,shah}@eecs.ucf.edu Abstract Learnng domnant moton patterns or actvtes

More information

benefit is 2, paid if the policyholder dies within the year, and probability of death within the year is ).

benefit is 2, paid if the policyholder dies within the year, and probability of death within the year is ). REVIEW OF RISK MANAGEMENT CONCEPTS LOSS DISTRIBUTIONS AND INSURANCE Loss and nsurance: When someone s subject to the rsk of ncurrng a fnancal loss, the loss s generally modeled usng a random varable or

More information

What is Candidate Sampling

What is Candidate Sampling What s Canddate Samplng Say we have a multclass or mult label problem where each tranng example ( x, T ) conssts of a context x a small (mult)set of target classes T out of a large unverse L of possble

More information

Recurrence. 1 Definitions and main statements

Recurrence. 1 Definitions and main statements Recurrence 1 Defntons and man statements Let X n, n = 0, 1, 2,... be a MC wth the state space S = (1, 2,...), transton probabltes p j = P {X n+1 = j X n = }, and the transton matrx P = (p j ),j S def.

More information

Credit Limit Optimization (CLO) for Credit Cards

Credit Limit Optimization (CLO) for Credit Cards Credt Lmt Optmzaton (CLO) for Credt Cards Vay S. Desa CSCC IX, Ednburgh September 8, 2005 Copyrght 2003, SAS Insttute Inc. All rghts reserved. SAS Propretary Agenda Background Tradtonal approaches to credt

More information

Module 2 LOSSLESS IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Module 2 LOSSLESS IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur Module LOSSLESS IMAGE COMPRESSION SYSTEMS Lesson 3 Lossless Compresson: Huffman Codng Instructonal Objectves At the end of ths lesson, the students should be able to:. Defne and measure source entropy..

More information

CONSISTENT VEHICLES TRACKING BY USING A COOPERATIVE DISTRIBUTED VIDEO SURVEILLANCESYSTEM

CONSISTENT VEHICLES TRACKING BY USING A COOPERATIVE DISTRIBUTED VIDEO SURVEILLANCESYSTEM Internatonal Research Journal of Appled and Basc Scences 2013 Avalable onlne at www.rjabs.com ISSN 2251-838X / Vol, 4 (12):3658-3663 Scence Explorer Publcatons CONSISTENT VEHICLES TRACKING BY USING A COOPERATIVE

More information

An Alternative Way to Measure Private Equity Performance

An Alternative Way to Measure Private Equity Performance An Alternatve Way to Measure Prvate Equty Performance Peter Todd Parlux Investment Technology LLC Summary Internal Rate of Return (IRR) s probably the most common way to measure the performance of prvate

More information

Multi-View Tracking of Multiple Targets with Dynamic Cameras

Multi-View Tracking of Multiple Targets with Dynamic Cameras Mult-Vew Tracng of Multple Targets wth Dynamc Cameras Tll Kroeger 1, Ralf Dragon 1, Luc Van Gool 1,2 1 Computer Vson Laboratory, ETH Zurch 2 ESAT - PSI / IBBT, K.U. Leuven {roegert, dragonr, vangool}@vson.ee.ethz.ch

More information

A Multi-mode Image Tracking System Based on Distributed Fusion

A Multi-mode Image Tracking System Based on Distributed Fusion A Mult-mode Image Tracng System Based on Dstrbuted Fuson Ln zheng Chongzhao Han Dongguang Zuo Hongsen Yan School of Electroncs & nformaton engneerng, X an Jaotong Unversty X an, Shaanx, Chna Lnzheng@malst.xjtu.edu.cn

More information

Realistic Image Synthesis

Realistic Image Synthesis Realstc Image Synthess - Combned Samplng and Path Tracng - Phlpp Slusallek Karol Myszkowsk Vncent Pegoraro Overvew: Today Combned Samplng (Multple Importance Samplng) Renderng and Measurng Equaton Random

More information

INVESTIGATION OF VEHICULAR USERS FAIRNESS IN CDMA-HDR NETWORKS

INVESTIGATION OF VEHICULAR USERS FAIRNESS IN CDMA-HDR NETWORKS 21 22 September 2007, BULGARIA 119 Proceedngs of the Internatonal Conference on Informaton Technologes (InfoTech-2007) 21 st 22 nd September 2007, Bulgara vol. 2 INVESTIGATION OF VEHICULAR USERS FAIRNESS

More information

Distributed Multi-Target Tracking In A Self-Configuring Camera Network

Distributed Multi-Target Tracking In A Self-Configuring Camera Network Dstrbuted Mult-Target Trackng In A Self-Confgurng Camera Network Crstan Soto, B Song, Amt K. Roy-Chowdhury Department of Electrcal Engneerng Unversty of Calforna, Rversde {cwlder,bsong,amtrc}@ee.ucr.edu

More information

Vehicle Detection and Tracking in Video from Moving Airborne Platform

Vehicle Detection and Tracking in Video from Moving Airborne Platform Journal of Computatonal Informaton Systems 10: 12 (2014) 4965 4972 Avalable at http://www.jofcs.com Vehcle Detecton and Trackng n Vdeo from Movng Arborne Platform Lye ZHANG 1,2,, Hua WANG 3, L LI 2 1 School

More information

DEFINING %COMPLETE IN MICROSOFT PROJECT

DEFINING %COMPLETE IN MICROSOFT PROJECT CelersSystems DEFINING %COMPLETE IN MICROSOFT PROJECT PREPARED BY James E Aksel, PMP, PMI-SP, MVP For Addtonal Informaton about Earned Value Management Systems and reportng, please contact: CelersSystems,

More information

The Greedy Method. Introduction. 0/1 Knapsack Problem

The Greedy Method. Introduction. 0/1 Knapsack Problem The Greedy Method Introducton We have completed data structures. We now are gong to look at algorthm desgn methods. Often we are lookng at optmzaton problems whose performance s exponental. For an optmzaton

More information

Vision Mouse. Saurabh Sarkar a* University of Cincinnati, Cincinnati, USA ABSTRACT 1. INTRODUCTION

Vision Mouse. Saurabh Sarkar a* University of Cincinnati, Cincinnati, USA ABSTRACT 1. INTRODUCTION Vson Mouse Saurabh Sarkar a* a Unversty of Cncnnat, Cncnnat, USA ABSTRACT The report dscusses a vson based approach towards trackng of eyes and fngers. The report descrbes the process of locatng the possble

More information

Causal, Explanatory Forecasting. Analysis. Regression Analysis. Simple Linear Regression. Which is Independent? Forecasting

Causal, Explanatory Forecasting. Analysis. Regression Analysis. Simple Linear Regression. Which is Independent? Forecasting Causal, Explanatory Forecastng Assumes cause-and-effect relatonshp between system nputs and ts output Forecastng wth Regresson Analyss Rchard S. Barr Inputs System Cause + Effect Relatonshp The job of

More information

Damage detection in composite laminates using coin-tap method

Damage detection in composite laminates using coin-tap method Damage detecton n composte lamnates usng con-tap method S.J. Km Korea Aerospace Research Insttute, 45 Eoeun-Dong, Youseong-Gu, 35-333 Daejeon, Republc of Korea yaeln@kar.re.kr 45 The con-tap test has the

More information

Hallucinating Multiple Occluded CCTV Face Images of Different Resolutions

Hallucinating Multiple Occluded CCTV Face Images of Different Resolutions In Proc. IEEE Internatonal Conference on Advanced Vdeo and Sgnal based Survellance (AVSS 05), September 2005 Hallucnatng Multple Occluded CCTV Face Images of Dfferent Resolutons Ku Ja Shaogang Gong Computer

More information

A hybrid global optimization algorithm based on parallel chaos optimization and outlook algorithm

A hybrid global optimization algorithm based on parallel chaos optimization and outlook algorithm Avalable onlne www.ocpr.com Journal of Chemcal and Pharmaceutcal Research, 2014, 6(7):1884-1889 Research Artcle ISSN : 0975-7384 CODEN(USA) : JCPRC5 A hybrd global optmzaton algorthm based on parallel

More information

Face Verification Problem. Face Recognition Problem. Application: Access Control. Biometric Authentication. Face Verification (1:1 matching)

Face Verification Problem. Face Recognition Problem. Application: Access Control. Biometric Authentication. Face Verification (1:1 matching) Face Recognton Problem Face Verfcaton Problem Face Verfcaton (1:1 matchng) Querymage face query Face Recognton (1:N matchng) database Applcaton: Access Control www.vsage.com www.vsoncs.com Bometrc Authentcaton

More information

A DYNAMIC CRASHING METHOD FOR PROJECT MANAGEMENT USING SIMULATION-BASED OPTIMIZATION. Michael E. Kuhl Radhamés A. Tolentino-Peña

A DYNAMIC CRASHING METHOD FOR PROJECT MANAGEMENT USING SIMULATION-BASED OPTIMIZATION. Michael E. Kuhl Radhamés A. Tolentino-Peña Proceedngs of the 2008 Wnter Smulaton Conference S. J. Mason, R. R. Hll, L. Mönch, O. Rose, T. Jefferson, J. W. Fowler eds. A DYNAMIC CRASHING METHOD FOR PROJECT MANAGEMENT USING SIMULATION-BASED OPTIMIZATION

More information

Feature selection for intrusion detection. Slobodan Petrović NISlab, Gjøvik University College

Feature selection for intrusion detection. Slobodan Petrović NISlab, Gjøvik University College Feature selecton for ntruson detecton Slobodan Petrovć NISlab, Gjøvk Unversty College Contents The feature selecton problem Intruson detecton Traffc features relevant for IDS The CFS measure The mrmr measure

More information

Project Networks With Mixed-Time Constraints

Project Networks With Mixed-Time Constraints Project Networs Wth Mxed-Tme Constrants L Caccetta and B Wattananon Western Australan Centre of Excellence n Industral Optmsaton (WACEIO) Curtn Unversty of Technology GPO Box U1987 Perth Western Australa

More information

Logistic Regression. Lecture 4: More classifiers and classes. Logistic regression. Adaboost. Optimization. Multiple class classification

Logistic Regression. Lecture 4: More classifiers and classes. Logistic regression. Adaboost. Optimization. Multiple class classification Lecture 4: More classfers and classes C4B Machne Learnng Hlary 20 A. Zsserman Logstc regresson Loss functons revsted Adaboost Loss functons revsted Optmzaton Multple class classfcaton Logstc Regresson

More information

Inter-Ing 2007. INTERDISCIPLINARITY IN ENGINEERING SCIENTIFIC INTERNATIONAL CONFERENCE, TG. MUREŞ ROMÂNIA, 15-16 November 2007.

Inter-Ing 2007. INTERDISCIPLINARITY IN ENGINEERING SCIENTIFIC INTERNATIONAL CONFERENCE, TG. MUREŞ ROMÂNIA, 15-16 November 2007. Inter-Ing 2007 INTERDISCIPLINARITY IN ENGINEERING SCIENTIFIC INTERNATIONAL CONFERENCE, TG. MUREŞ ROMÂNIA, 15-16 November 2007. UNCERTAINTY REGION SIMULATION FOR A SERIAL ROBOT STRUCTURE MARIUS SEBASTIAN

More information

Learning from Multiple Outlooks

Learning from Multiple Outlooks Learnng from Multple Outlooks Maayan Harel Department of Electrcal Engneerng, Technon, Hafa, Israel She Mannor Department of Electrcal Engneerng, Technon, Hafa, Israel maayanga@tx.technon.ac.l she@ee.technon.ac.l

More information

Traffic State Estimation in the Traffic Management Center of Berlin

Traffic State Estimation in the Traffic Management Center of Berlin Traffc State Estmaton n the Traffc Management Center of Berln Authors: Peter Vortsch, PTV AG, Stumpfstrasse, D-763 Karlsruhe, Germany phone ++49/72/965/35, emal peter.vortsch@ptv.de Peter Möhl, PTV AG,

More information

Extending Probabilistic Dynamic Epistemic Logic

Extending Probabilistic Dynamic Epistemic Logic Extendng Probablstc Dynamc Epstemc Logc Joshua Sack May 29, 2008 Probablty Space Defnton A probablty space s a tuple (S, A, µ), where 1 S s a set called the sample space. 2 A P(S) s a σ-algebra: a set

More information

An Interest-Oriented Network Evolution Mechanism for Online Communities

An Interest-Oriented Network Evolution Mechanism for Online Communities An Interest-Orented Network Evoluton Mechansm for Onlne Communtes Cahong Sun and Xaopng Yang School of Informaton, Renmn Unversty of Chna, Bejng 100872, P.R. Chna {chsun,yang}@ruc.edu.cn Abstract. Onlne

More information

Institute of Informatics, Faculty of Business and Management, Brno University of Technology,Czech Republic

Institute of Informatics, Faculty of Business and Management, Brno University of Technology,Czech Republic Lagrange Multplers as Quanttatve Indcators n Economcs Ivan Mezník Insttute of Informatcs, Faculty of Busness and Management, Brno Unversty of TechnologCzech Republc Abstract The quanttatve role of Lagrange

More information

RequIn, a tool for fast web traffic inference

RequIn, a tool for fast web traffic inference RequIn, a tool for fast web traffc nference Olver aul, Jean Etenne Kba GET/INT, LOR Department 9 rue Charles Fourer 90 Evry, France Olver.aul@nt-evry.fr, Jean-Etenne.Kba@nt-evry.fr Abstract As networked

More information

GRAVITY DATA VALIDATION AND OUTLIER DETECTION USING L 1 -NORM

GRAVITY DATA VALIDATION AND OUTLIER DETECTION USING L 1 -NORM GRAVITY DATA VALIDATION AND OUTLIER DETECTION USING L 1 -NORM BARRIOT Jean-Perre, SARRAILH Mchel BGI/CNES 18.av.E.Beln 31401 TOULOUSE Cedex 4 (France) Emal: jean-perre.barrot@cnes.fr 1/Introducton The

More information

CS 2750 Machine Learning. Lecture 3. Density estimation. CS 2750 Machine Learning. Announcements

CS 2750 Machine Learning. Lecture 3. Density estimation. CS 2750 Machine Learning. Announcements Lecture 3 Densty estmaton Mlos Hauskrecht mlos@cs.ptt.edu 5329 Sennott Square Next lecture: Matlab tutoral Announcements Rules for attendng the class: Regstered for credt Regstered for audt (only f there

More information

POLYSA: A Polynomial Algorithm for Non-binary Constraint Satisfaction Problems with and

POLYSA: A Polynomial Algorithm for Non-binary Constraint Satisfaction Problems with and POLYSA: A Polynomal Algorthm for Non-bnary Constrant Satsfacton Problems wth and Mguel A. Saldo, Federco Barber Dpto. Sstemas Informátcos y Computacón Unversdad Poltécnca de Valenca, Camno de Vera s/n

More information

Can Auto Liability Insurance Purchases Signal Risk Attitude?

Can Auto Liability Insurance Purchases Signal Risk Attitude? Internatonal Journal of Busness and Economcs, 2011, Vol. 10, No. 2, 159-164 Can Auto Lablty Insurance Purchases Sgnal Rsk Atttude? Chu-Shu L Department of Internatonal Busness, Asa Unversty, Tawan Sheng-Chang

More information

J. Parallel Distrib. Comput.

J. Parallel Distrib. Comput. J. Parallel Dstrb. Comput. 71 (2011) 62 76 Contents lsts avalable at ScenceDrect J. Parallel Dstrb. Comput. journal homepage: www.elsever.com/locate/jpdc Optmzng server placement n dstrbuted systems n

More information

RESEARCH ON DUAL-SHAKER SINE VIBRATION CONTROL. Yaoqi FENG 1, Hanping QIU 1. China Academy of Space Technology (CAST) yaoqi.feng@yahoo.

RESEARCH ON DUAL-SHAKER SINE VIBRATION CONTROL. Yaoqi FENG 1, Hanping QIU 1. China Academy of Space Technology (CAST) yaoqi.feng@yahoo. ICSV4 Carns Australa 9- July, 007 RESEARCH ON DUAL-SHAKER SINE VIBRATION CONTROL Yaoq FENG, Hanpng QIU Dynamc Test Laboratory, BISEE Chna Academy of Space Technology (CAST) yaoq.feng@yahoo.com Abstract

More information

An interactive system for structure-based ASCII art creation

An interactive system for structure-based ASCII art creation An nteractve system for structure-based ASCII art creaton Katsunor Myake Henry Johan Tomoyuk Nshta The Unversty of Tokyo Nanyang Technologcal Unversty Abstract Non-Photorealstc Renderng (NPR), whose am

More information

How To Understand The Results Of The German Meris Cloud And Water Vapour Product

How To Understand The Results Of The German Meris Cloud And Water Vapour Product Ttel: Project: Doc. No.: MERIS level 3 cloud and water vapour products MAPP MAPP-ATBD-ClWVL3 Issue: 1 Revson: 0 Date: 9.12.1998 Functon Name Organsaton Sgnature Date Author: Bennartz FUB Preusker FUB Schüller

More information

Logistic Regression. Steve Kroon

Logistic Regression. Steve Kroon Logstc Regresson Steve Kroon Course notes sectons: 24.3-24.4 Dsclamer: these notes do not explctly ndcate whether values are vectors or scalars, but expects the reader to dscern ths from the context. Scenaro

More information

An Enhanced Super-Resolution System with Improved Image Registration, Automatic Image Selection, and Image Enhancement

An Enhanced Super-Resolution System with Improved Image Registration, Automatic Image Selection, and Image Enhancement An Enhanced Super-Resoluton System wth Improved Image Regstraton, Automatc Image Selecton, and Image Enhancement Yu-Chuan Kuo ( ), Chen-Yu Chen ( ), and Chou-Shann Fuh ( ) Department of Computer Scence

More information

IMPACT ANALYSIS OF A CELLULAR PHONE

IMPACT ANALYSIS OF A CELLULAR PHONE 4 th ASA & μeta Internatonal Conference IMPACT AALYSIS OF A CELLULAR PHOE We Lu, 2 Hongy L Bejng FEAonlne Engneerng Co.,Ltd. Bejng, Chna ABSTRACT Drop test smulaton plays an mportant role n nvestgatng

More information

NON-LINEAR MULTIMODAL OBJECT TRACKING BASED ON 2D LIDAR DATA

NON-LINEAR MULTIMODAL OBJECT TRACKING BASED ON 2D LIDAR DATA Metrol. Meas. Syst. Vol. XVI (009), No 3, pp. 359-369 METROLOGY AND MEASUREMENT SYSTEMS Index 330930, ISSN 0860-89 www.metrology.pg.gda.pl NON-LINEAR MULTIMODAL OBJECT TRACKING BASED ON D LIDAR DATA Mchael

More information

ANALYZING THE RELATIONSHIPS BETWEEN QUALITY, TIME, AND COST IN PROJECT MANAGEMENT DECISION MAKING

ANALYZING THE RELATIONSHIPS BETWEEN QUALITY, TIME, AND COST IN PROJECT MANAGEMENT DECISION MAKING ANALYZING THE RELATIONSHIPS BETWEEN QUALITY, TIME, AND COST IN PROJECT MANAGEMENT DECISION MAKING Matthew J. Lberatore, Department of Management and Operatons, Vllanova Unversty, Vllanova, PA 19085, 610-519-4390,

More information

Forecasting the Demand of Emergency Supplies: Based on the CBR Theory and BP Neural Network

Forecasting the Demand of Emergency Supplies: Based on the CBR Theory and BP Neural Network 700 Proceedngs of the 8th Internatonal Conference on Innovaton & Management Forecastng the Demand of Emergency Supples: Based on the CBR Theory and BP Neural Network Fu Deqang, Lu Yun, L Changbng School

More information

CHOLESTEROL REFERENCE METHOD LABORATORY NETWORK. Sample Stability Protocol

CHOLESTEROL REFERENCE METHOD LABORATORY NETWORK. Sample Stability Protocol CHOLESTEROL REFERENCE METHOD LABORATORY NETWORK Sample Stablty Protocol Background The Cholesterol Reference Method Laboratory Network (CRMLN) developed certfcaton protocols for total cholesterol, HDL

More information

L10: Linear discriminants analysis

L10: Linear discriminants analysis L0: Lnear dscrmnants analyss Lnear dscrmnant analyss, two classes Lnear dscrmnant analyss, C classes LDA vs. PCA Lmtatons of LDA Varants of LDA Other dmensonalty reducton methods CSCE 666 Pattern Analyss

More information

Robust Design of Public Storage Warehouses. Yeming (Yale) Gong EMLYON Business School

Robust Design of Public Storage Warehouses. Yeming (Yale) Gong EMLYON Business School Robust Desgn of Publc Storage Warehouses Yemng (Yale) Gong EMLYON Busness School Rene de Koster Rotterdam school of management, Erasmus Unversty Abstract We apply robust optmzaton and revenue management

More information

1. Measuring association using correlation and regression

1. Measuring association using correlation and regression How to measure assocaton I: Correlaton. 1. Measurng assocaton usng correlaton and regresson We often would lke to know how one varable, such as a mother's weght, s related to another varable, such as a

More information

Human Tracking by Fast Mean Shift Mode Seeking

Human Tracking by Fast Mean Shift Mode Seeking JOURAL OF MULTIMEDIA, VOL. 1, O. 1, APRIL 2006 1 Human Trackng by Fast Mean Shft Mode Seekng [10 font sze blank 1] [10 font sze blank 2] C. Belezna Advanced Computer Vson GmbH - ACV, Venna, Austra Emal:

More information

On the Optimal Control of a Cascade of Hydro-Electric Power Stations

On the Optimal Control of a Cascade of Hydro-Electric Power Stations On the Optmal Control of a Cascade of Hydro-Electrc Power Statons M.C.M. Guedes a, A.F. Rbero a, G.V. Smrnov b and S. Vlela c a Department of Mathematcs, School of Scences, Unversty of Porto, Portugal;

More information

Disagreement-Based Multi-System Tracking

Disagreement-Based Multi-System Tracking Dsagreement-Based Mult-System Trackng Quannan L 1, Xnggang Wang 2, We Wang 3, Yuan Jang 3, Zh-Hua Zhou 3, Zhuowen Tu 1 1 Lab of Neuro Imagng, Unversty of Calforna, Los Angeles 2 Huazhong Unversty of Scence

More information

Support Vector Machines

Support Vector Machines Support Vector Machnes Max Wellng Department of Computer Scence Unversty of Toronto 10 Kng s College Road Toronto, M5S 3G5 Canada wellng@cs.toronto.edu Abstract Ths s a note to explan support vector machnes.

More information

An Introduction to 3G Monte-Carlo simulations within ProMan

An Introduction to 3G Monte-Carlo simulations within ProMan An Introducton to 3G Monte-Carlo smulatons wthn ProMan responsble edtor: Hermann Buddendck AWE Communcatons GmbH Otto-Llenthal-Str. 36 D-71034 Böblngen Phone: +49 70 31 71 49 7-16 Fax: +49 70 31 71 49

More information

Stochastic Protocol Modeling for Anomaly Based Network Intrusion Detection

Stochastic Protocol Modeling for Anomaly Based Network Intrusion Detection Stochastc Protocol Modelng for Anomaly Based Network Intruson Detecton Juan M. Estevez-Tapador, Pedro Garca-Teodoro, and Jesus E. Daz-Verdejo Department of Electroncs and Computer Technology Unversty of

More information

The Use of Analytics for Claim Fraud Detection Roosevelt C. Mosley, Jr., FCAS, MAAA Nick Kucera Pinnacle Actuarial Resources Inc.

The Use of Analytics for Claim Fraud Detection Roosevelt C. Mosley, Jr., FCAS, MAAA Nick Kucera Pinnacle Actuarial Resources Inc. Paper 1837-2014 The Use of Analytcs for Clam Fraud Detecton Roosevelt C. Mosley, Jr., FCAS, MAAA Nck Kucera Pnnacle Actuaral Resources Inc., Bloomngton, IL ABSTRACT As t has been wdely reported n the nsurance

More information

Risk-based Fatigue Estimate of Deep Water Risers -- Course Project for EM388F: Fracture Mechanics, Spring 2008

Risk-based Fatigue Estimate of Deep Water Risers -- Course Project for EM388F: Fracture Mechanics, Spring 2008 Rsk-based Fatgue Estmate of Deep Water Rsers -- Course Project for EM388F: Fracture Mechancs, Sprng 2008 Chen Sh Department of Cvl, Archtectural, and Envronmental Engneerng The Unversty of Texas at Austn

More information

Staff Paper. Farm Savings Accounts: Examining Income Variability, Eligibility, and Benefits. Brent Gloy, Eddy LaDue, and Charles Cuykendall

Staff Paper. Farm Savings Accounts: Examining Income Variability, Eligibility, and Benefits. Brent Gloy, Eddy LaDue, and Charles Cuykendall SP 2005-02 August 2005 Staff Paper Department of Appled Economcs and Management Cornell Unversty, Ithaca, New York 14853-7801 USA Farm Savngs Accounts: Examnng Income Varablty, Elgblty, and Benefts Brent

More information

Single and multiple stage classifiers implementing logistic discrimination

Single and multiple stage classifiers implementing logistic discrimination Sngle and multple stage classfers mplementng logstc dscrmnaton Hélo Radke Bttencourt 1 Dens Alter de Olvera Moraes 2 Vctor Haertel 2 1 Pontfíca Unversdade Católca do Ro Grande do Sul - PUCRS Av. Ipranga,

More information

Enabling P2P One-view Multi-party Video Conferencing

Enabling P2P One-view Multi-party Video Conferencing Enablng P2P One-vew Mult-party Vdeo Conferencng Yongxang Zhao, Yong Lu, Changja Chen, and JanYn Zhang Abstract Mult-Party Vdeo Conferencng (MPVC) facltates realtme group nteracton between users. Whle P2P

More information

Examensarbete. Rotating Workforce Scheduling. Caroline Granfeldt

Examensarbete. Rotating Workforce Scheduling. Caroline Granfeldt Examensarbete Rotatng Workforce Schedulng Carolne Granfeldt LTH - MAT - EX - - 2015 / 08 - - SE Rotatng Workforce Schedulng Optmerngslära, Lnköpngs Unverstet Carolne Granfeldt LTH - MAT - EX - - 2015

More information

denote the location of a node, and suppose node X . This transmission causes a successful reception by node X for any other node

denote the location of a node, and suppose node X . This transmission causes a successful reception by node X for any other node Fnal Report of EE359 Class Proect Throughput and Delay n Wreless Ad Hoc Networs Changhua He changhua@stanford.edu Abstract: Networ throughput and pacet delay are the two most mportant parameters to evaluate

More information

) of the Cell class is created containing information about events associated with the cell. Events are added to the Cell instance

) of the Cell class is created containing information about events associated with the cell. Events are added to the Cell instance Calbraton Method Instances of the Cell class (one nstance for each FMS cell) contan ADC raw data and methods assocated wth each partcular FMS cell. The calbraton method ncludes event selecton (Class Cell

More information

"Research Note" APPLICATION OF CHARGE SIMULATION METHOD TO ELECTRIC FIELD CALCULATION IN THE POWER CABLES *

Research Note APPLICATION OF CHARGE SIMULATION METHOD TO ELECTRIC FIELD CALCULATION IN THE POWER CABLES * Iranan Journal of Scence & Technology, Transacton B, Engneerng, ol. 30, No. B6, 789-794 rnted n The Islamc Republc of Iran, 006 Shraz Unversty "Research Note" ALICATION OF CHARGE SIMULATION METHOD TO ELECTRIC

More information

PAS: A Packet Accounting System to Limit the Effects of DoS & DDoS. Debish Fesehaye & Klara Naherstedt University of Illinois-Urbana Champaign

PAS: A Packet Accounting System to Limit the Effects of DoS & DDoS. Debish Fesehaye & Klara Naherstedt University of Illinois-Urbana Champaign PAS: A Packet Accountng System to Lmt the Effects of DoS & DDoS Debsh Fesehaye & Klara Naherstedt Unversty of Illnos-Urbana Champagn DoS and DDoS DDoS attacks are ncreasng threats to our dgtal world. Exstng

More information

Multiframe Many Many Point Correspondence for Vehicle Tracking in High Density Wide Area Aerial Videos

Multiframe Many Many Point Correspondence for Vehicle Tracking in High Density Wide Area Aerial Videos Int J Comput Vs DOI 10.1007/s11263-013-0624-1 Multframe Many Many Pont Correspondence for Vehcle Trackng n Hgh Densty Wde Area Aeral Vdeos Imran Saleem Mubarak Shah Receved: 21 December 2011 / Accepted:

More information

A Multi-Camera System on PC-Cluster for Real-time 3-D Tracking

A Multi-Camera System on PC-Cluster for Real-time 3-D Tracking The 23 rd Conference of the Mechancal Engneerng Network of Thaland November 4 7, 2009, Chang Ma A Mult-Camera System on PC-Cluster for Real-tme 3-D Trackng Vboon Sangveraphunsr*, Krtsana Uttamang, and

More information

Enterprise Master Patient Index

Enterprise Master Patient Index Enterprse Master Patent Index Healthcare data are captured n many dfferent settngs such as hosptals, clncs, labs, and physcan offces. Accordng to a report by the CDC, patents n the Unted States made an

More information

An MILP model for planning of batch plants operating in a campaign-mode

An MILP model for planning of batch plants operating in a campaign-mode An MILP model for plannng of batch plants operatng n a campagn-mode Yanna Fumero Insttuto de Desarrollo y Dseño CONICET UTN yfumero@santafe-concet.gov.ar Gabrela Corsano Insttuto de Desarrollo y Dseño

More information

Number of Levels Cumulative Annual operating Income per year construction costs costs ($) ($) ($) 1 600,000 35,000 100,000 2 2,200,000 60,000 350,000

Number of Levels Cumulative Annual operating Income per year construction costs costs ($) ($) ($) 1 600,000 35,000 100,000 2 2,200,000 60,000 350,000 Problem Set 5 Solutons 1 MIT s consderng buldng a new car park near Kendall Square. o unversty funds are avalable (overhead rates are under pressure and the new faclty would have to pay for tself from

More information

Luby s Alg. for Maximal Independent Sets using Pairwise Independence

Luby s Alg. for Maximal Independent Sets using Pairwise Independence Lecture Notes for Randomzed Algorthms Luby s Alg. for Maxmal Independent Sets usng Parwse Independence Last Updated by Erc Vgoda on February, 006 8. Maxmal Independent Sets For a graph G = (V, E), an ndependent

More information

Simulation and optimization of supply chains: alternative or complementary approaches?

Simulation and optimization of supply chains: alternative or complementary approaches? Smulaton and optmzaton of supply chans: alternatve or complementary approaches? Chrstan Almeder Margaretha Preusser Rchard F. Hartl Orgnally publshed n: OR Spectrum (2009) 31:95 119 DOI 10.1007/s00291-007-0118-z

More information

SPEE Recommended Evaluation Practice #6 Definition of Decline Curve Parameters Background:

SPEE Recommended Evaluation Practice #6 Definition of Decline Curve Parameters Background: SPEE Recommended Evaluaton Practce #6 efnton of eclne Curve Parameters Background: The producton hstores of ol and gas wells can be analyzed to estmate reserves and future ol and gas producton rates and

More information

8 Algorithm for Binary Searching in Trees

8 Algorithm for Binary Searching in Trees 8 Algorthm for Bnary Searchng n Trees In ths secton we present our algorthm for bnary searchng n trees. A crucal observaton employed by the algorthm s that ths problem can be effcently solved when the

More information

Statistical Methods to Develop Rating Models

Statistical Methods to Develop Rating Models Statstcal Methods to Develop Ratng Models [Evelyn Hayden and Danel Porath, Österrechsche Natonalbank and Unversty of Appled Scences at Manz] Source: The Basel II Rsk Parameters Estmaton, Valdaton, and

More information

Descriptive Models. Cluster Analysis. Example. General Applications of Clustering. Examples of Clustering Applications

Descriptive Models. Cluster Analysis. Example. General Applications of Clustering. Examples of Clustering Applications CMSC828G Prncples of Data Mnng Lecture #9 Today s Readng: HMS, chapter 9 Today s Lecture: Descrptve Modelng Clusterng Algorthms Descrptve Models model presents the man features of the data, a global summary

More information

Implementation of Deutsch's Algorithm Using Mathcad

Implementation of Deutsch's Algorithm Using Mathcad Implementaton of Deutsch's Algorthm Usng Mathcad Frank Roux The followng s a Mathcad mplementaton of Davd Deutsch's quantum computer prototype as presented on pages - n "Machnes, Logc and Quantum Physcs"

More information

A statistical approach to determine Microbiologically Influenced Corrosion (MIC) Rates of underground gas pipelines.

A statistical approach to determine Microbiologically Influenced Corrosion (MIC) Rates of underground gas pipelines. A statstcal approach to determne Mcrobologcally Influenced Corroson (MIC) Rates of underground gas ppelnes. by Lech A. Grzelak A thess submtted to the Delft Unversty of Technology n conformty wth the requrements

More information

2. SYSTEM MODEL. the SLA (unlike the only other related mechanism [15] we can compare it is never able to meet the SLA).

2. SYSTEM MODEL. the SLA (unlike the only other related mechanism [15] we can compare it is never able to meet the SLA). Managng Server Energy and Operatonal Costs n Hostng Centers Yyu Chen Dept. of IE Penn State Unversty Unversty Park, PA 16802 yzc107@psu.edu Anand Svasubramanam Dept. of CSE Penn State Unversty Unversty

More information

1. Fundamentals of probability theory 2. Emergence of communication traffic 3. Stochastic & Markovian Processes (SP & MP)

1. Fundamentals of probability theory 2. Emergence of communication traffic 3. Stochastic & Markovian Processes (SP & MP) 6.3 / -- Communcaton Networks II (Görg) SS20 -- www.comnets.un-bremen.de Communcaton Networks II Contents. Fundamentals of probablty theory 2. Emergence of communcaton traffc 3. Stochastc & Markovan Processes

More information

AN APPOINTMENT ORDER OUTPATIENT SCHEDULING SYSTEM THAT IMPROVES OUTPATIENT EXPERIENCE

AN APPOINTMENT ORDER OUTPATIENT SCHEDULING SYSTEM THAT IMPROVES OUTPATIENT EXPERIENCE AN APPOINTMENT ORDER OUTPATIENT SCHEDULING SYSTEM THAT IMPROVES OUTPATIENT EXPERIENCE Yu-L Huang Industral Engneerng Department New Mexco State Unversty Las Cruces, New Mexco 88003, U.S.A. Abstract Patent

More information

1 Example 1: Axis-aligned rectangles

1 Example 1: Axis-aligned rectangles COS 511: Theoretcal Machne Learnng Lecturer: Rob Schapre Lecture # 6 Scrbe: Aaron Schld February 21, 2013 Last class, we dscussed an analogue for Occam s Razor for nfnte hypothess spaces that, n conjuncton

More information

How To Calculate The Accountng Perod Of Nequalty

How To Calculate The Accountng Perod Of Nequalty Inequalty and The Accountng Perod Quentn Wodon and Shlomo Ytzha World Ban and Hebrew Unversty September Abstract Income nequalty typcally declnes wth the length of tme taen nto account for measurement.

More information

PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 12

PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 12 14 The Ch-squared dstrbuton PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 1 If a normal varable X, havng mean µ and varance σ, s standardsed, the new varable Z has a mean 0 and varance 1. When ths standardsed

More information

To Fill or not to Fill: The Gas Station Problem

To Fill or not to Fill: The Gas Station Problem To Fll or not to Fll: The Gas Staton Problem Samr Khuller Azarakhsh Malekan Julán Mestre Abstract In ths paper we study several routng problems that generalze shortest paths and the Travelng Salesman Problem.

More information

How Sets of Coherent Probabilities May Serve as Models for Degrees of Incoherence

How Sets of Coherent Probabilities May Serve as Models for Degrees of Incoherence 1 st Internatonal Symposum on Imprecse Probabltes and Ther Applcatons, Ghent, Belgum, 29 June 2 July 1999 How Sets of Coherent Probabltes May Serve as Models for Degrees of Incoherence Mar J. Schervsh

More information

Vehicle Tracking Using Particle Filter for Parking Management System

Vehicle Tracking Using Particle Filter for Parking Management System 2014 4th Internatonal Conference on Artfcal Intellgence wth Applcatons n Engneerng and Technology Vehcle Trackng Usng Partcle Flter for Parkng Management System Kenneth Tze Kn Teo, Renee Ka Yn Chn, N.S.V.

More information

Tracking with Non-Linear Dynamic Models

Tracking with Non-Linear Dynamic Models CHAPTER 2 Trackng wth Non-Lnear Dynamc Models In a lnear dynamc model wth lnear measurements, there s always only one peak n the posteror; very small non-lneartes n dynamc models can lead to a substantal

More information

Multi-Robot Tracking of a Moving Object Using Directional Sensors

Multi-Robot Tracking of a Moving Object Using Directional Sensors Mult-Robot Trackng of a Movng Object Usng Drectonal Sensors Xaomng Hu, Karl H. Johansson, Manuel Mazo Jr., Alberto Speranzon Dept. of Sgnals, Sensors & Systems Royal Insttute of Technology, SE- 44 Stockholm,

More information

A DATA MINING APPLICATION IN A STUDENT DATABASE

A DATA MINING APPLICATION IN A STUDENT DATABASE JOURNAL OF AERONAUTICS AND SPACE TECHNOLOGIES JULY 005 VOLUME NUMBER (53-57) A DATA MINING APPLICATION IN A STUDENT DATABASE Şenol Zafer ERDOĞAN Maltepe Ünversty Faculty of Engneerng Büyükbakkalköy-Istanbul

More information

A Decision Support System for Crowd Control

A Decision Support System for Crowd Control A Decson Support System for Crowd Control Johan Schubert, Lug Ferrara, Pontus Hörlng, Johan Walter Department of Decson Support Systems Dvson of Command and Control Systems Swedsh Defence Research Agency

More information

THE METHOD OF LEAST SQUARES THE METHOD OF LEAST SQUARES

THE METHOD OF LEAST SQUARES THE METHOD OF LEAST SQUARES The goal: to measure (determne) an unknown quantty x (the value of a RV X) Realsaton: n results: y 1, y 2,..., y j,..., y n, (the measured values of Y 1, Y 2,..., Y j,..., Y n ) every result s encumbered

More information

Data Broadcast on a Multi-System Heterogeneous Overlayed Wireless Network *

Data Broadcast on a Multi-System Heterogeneous Overlayed Wireless Network * JOURNAL OF INFORMATION SCIENCE AND ENGINEERING 24, 819-840 (2008) Data Broadcast on a Mult-System Heterogeneous Overlayed Wreless Network * Department of Computer Scence Natonal Chao Tung Unversty Hsnchu,

More information

7.5. Present Value of an Annuity. Investigate

7.5. Present Value of an Annuity. Investigate 7.5 Present Value of an Annuty Owen and Anna are approachng retrement and are puttng ther fnances n order. They have worked hard and nvested ther earnngs so that they now have a large amount of money on

More information

THE DISTRIBUTION OF LOAN PORTFOLIO VALUE * Oldrich Alfons Vasicek

THE DISTRIBUTION OF LOAN PORTFOLIO VALUE * Oldrich Alfons Vasicek HE DISRIBUION OF LOAN PORFOLIO VALUE * Oldrch Alfons Vascek he amount of captal necessary to support a portfolo of debt securtes depends on the probablty dstrbuton of the portfolo loss. Consder a portfolo

More information