Calibration of ranging sensors for mobile mapping applications

Transcription

1 1/24 Calibration of ranging sensors for mobile mapping applications N. Seube(*), T. Touze(*), A. Picard(*), M. Rondeau(**) (*)Ocean Sensing and Mapping Lab. ENSTA Bretagne Brest FRANCE and (**)CIDCO Rimouski CANADA, Qc HYDRO 12, Rotterdam

2 Framework of the research 2/24 Cooperation ENSTA (France) - CIDCO (Canada) on : Coastal erosion monitoring surveys Infrastructure surveys Error analysis of hydrographic survey systems Motivation of this work : Vessel mounted LiDAR for port infrastructure inspection at very low scale Mobile mapping from land vehicules for coastal monitoring, dam lake survey Enhancement of existing boresight calibration used in hydrography

3 Calibration of ranging devices for mobile mapping 3/24 Calibration problems : Synchronization (IMU-POSIT-RANGING) : time-tagging Boresight angles (IMU-Ranging), misalignment, lever-arms, range Environment dependent parameters (acoustic, radiometric): not addressed in this talk Classical approaches in boresight calibration : Visual point cloud matching (patch test) DEM matching (patch test) Natural surface matching (e.g; point matching) Surface matching

4 Basics of Calibration 4/24 IMU-Ranging Latency calibration : Survey plateform dynamics produce time-varying attitude errors Latency recovery from a data set is almost impossible Latency acts on geolocated POINTS Geometric calibration : Boresight errors are micro-rotations Lever arms acts as translations Boresight recovery can be done by DIRECTION fitting, lever-arm by POINT fitting

5 Drawback of existing methods 5/24 Latency estimation No clear methodology Accuracy not supported in acquisition softwares (resolution= 1ms...) Latency recovery from data set is not reliable Boresight Patch test : force the decoupling of a 3-D problem to a sequence of 3 1-D problems, Poor accuracy (0,1 deg) On site calibration data are corrupted by positioning devices

6 Our objective 6/24 Latency recovery with 0,1ms resolution 3D boresight/lever arms estimation, with target accuracy of less than 0,01 deg Proposed methodology : 1 IMU-Ranging latency estimation 2 3-D Boresight angle adjustment 3 Boresight + Lever-arm adjustment by control point fitting

7 Sources of (IMU-Ranging device) time-tagging errors 7/24 IMU IMUs internal computation integrates navigation dynamics All IMUs have a specific latency, not always documented Acquisition System Geolocation computation may induce latency Comm boards quality and acquisition buffers may induce latency

8 Objective of Latency calibration 8/24 Airborne LiDAR: 0,1ms accuracy (source: Skaloud, Litchi, 2006, ISPRS) Vessel mounted LiDAR: 0,5 ms (for centimeter accuracy for marine infrastructure surveys) Land vehicle monted LiDAR : 0,2 ms (high motion dynamics) IMU-Ranging device TOTAL latency definition: IMU time delay between physical measurement and data output Transmission delay to the acquisition device Stacking in acquisition buffers Acquisition software data assembling (orientation, ranging, position) used for points geolocation

9 A positioning free latency calibration method n = (N, E, D): navigation frame, bs: mobile ranging device body frame, bi: IMU frame. M :target point, M same point viewed with rotational motion x n = RbI n RbI bs x S x n = RbI n (t dt)rbi bs x S where R n bi and RbI bs : DCM from bi to n and from bs to bi. 9/24

10 Latency calibration (cont d) 10/24 x n = R n bi RbI n (t dt)x n Assuming a constant angular velocity: Rn bi (t dt) = Rn bi d [ dt R bi n ].dt = (Id + dtω bi n/bi )RbI n thus x n = RbI n (Id + dtωbi n/bi )RbI n x n = x n + dtrbi n ΩbI n/bi RbI n x n = x n + dtω n n/bi x n

11 Latency calibration (cont d) 11/24 Target point shift due to the latency dt: = x x. n = dt Ω n n/bi x n = dt ωbi/n n x n = dt RbI n ωbi bi/n x n Finally, the latency estimation equation is : dt = n ω bi bi/n x n (1)

12 Framework and Motivation Experimental set-up 12/24

13 Experimental set-up 13/24 Target choice: precision sphere Equipments : Leica HDS6200, IxSea Octans4 Acquisition software: Qinsy HDS6200-PC : Ethernet, PPS Octans4-PC: serial link, no PPS Alternate scans of the sphere: up to (-18, 18) deg/sec IMU+LiDAR motion controlled by a 3D rotating table (IXBlue TRI30)

14 Experimental results 14/24 Example of two spheres scans :

15 Experimental results 15/24 ω D (deg/sec) Latency estimate (ms) Sphere center STD (10 5 m) Latency STD (ms) Table: Latency estimates for various angular speeds. Latency standard deviation decreases with the angular speed value. In these results, the OCTANS4 IMU time-tags in taking into account its known latency of 2.35ms. So the total latency estimate is 4.21ms. Conclusion : Accuracy = 0,01ms, Precision = 0,09ms Applicable to any ranging device. Depends only on acquisition system and IMU configuration Able to detect the presence of serial links buffers

16 On site boresight calibration for terrestrial mobile mapping Point Geolocation equation: x n = X n + RbI n (RbI bs r bs + RB bil bb) x n : souding posit in navigation frame X n : IMU posit in navigation frame (provided by hybridized GPS/IMU) RbI n : IMU attitude (direction cosine matrix) r bs : LiDAR ranging output in LiDAR frame l bb : Lever-arm Posit-LiDAR in Body Frame 16/24

17 Principle of the method 17/24 v bs : fit of static LiDAR returns on a given planar surface Geolocation equation: v n = R n bi RbI bs v bs Criterion: All vectors v n should lie on the same plan, defined by its normal n. Observation equation : n v n = 0 for all static scans Different observation sites and different attitude with respect to the plan, in order to identify boresight errors.

18 Numerical simulation results 18/24 From: n v n = 0 we obtain observations equations depending on the boresight parameters. The system can be solved by non linear optimization tools. Simulation results (take into account LiDAR and posit errors): Distance: 50m of a vertical wall Set of 12 static scans at various roll, pitch, heading Boresight precision: 0,02 deg Boresight accuracy: 0,005 deg

19 Framework and Motivation Mobile mapping system Mobile mapping system mounted on an Argo vehicle during the Erasmus IP 2012 camp. 19/24

20 Framework and Motivation Boresight angle residuals STD (Leica HDS6200 / IXBLUE LANDINS / Magellan Proflex 6000) system 20/24

21 Practical results (Vassiviere ERASMUS camp 2012) 21/24 28 observation sites of a wall, at various pitch, roll and heading angles roll angle STD = deg pitch angle STD = deg heading angle STD = deg

22 Survey quality 22/24 A posteriori error : STD = 5cm max for a survey achieved at 5km/h on rocky beaches (high plateform dynamics).

23 Tie point check 23/24 Basket ball, scanned from 7 different sites at a distance of 15m. Sphere center position STD = 1,5cm (less than RTK posit uncertainty).

24 24/24 Positionning errors propagates through calibration methods and alterate its accuracy Latency calibration method compatible with ALS requirements Positioning free IMU-LiDAR boresight calibration procedure Current work: Field validation of boresight calibration Adaptation of boresight, lever-arm and boresight calibration methods to MBES calibration