A RAPID METHOD FOR WATER TARGET EXTRACTION IN SAR IMAGE



Similar documents
VEHICLE DETECTION AND CLASSIFICATION FROM HIGH RESOLUTION SATELLITE IMAGES

Copyright 2005 IEEE. Reprinted from IEEE MTT-S International Microwave Symposium 2005

Using quantum computing to realize the Fourier Transform in computer vision applications

Face detection is a process of localizing and extracting the face region from the

Doppler Shift Estimation and Jamming Detection for Cellular Networks

Power functions: f(x) = x n, n is a natural number The graphs of some power functions are given below. n- even n- odd

A LabVIEW Based Experimental Platform for Ultrasonic Range Measurements

A NOVEL ALGORITHM WITH IM-LSI INDEX FOR INCREMENTAL MAINTENANCE OF MATERIALIZED VIEW

Automatic Traffic Estimation Using Image Processing

Ch9 Morphological Image Processing

A PHOTOGRAMMETRIC APPRAOCH FOR AUTOMATIC TRAFFIC ASSESSMENT USING CONVENTIONAL CCTV CAMERA

Synthetic Aperture Radar: Principles and Applications of AI in Automatic Target Recognition

Pattern recognition using multilayer neural-genetic algorithm

A performance analysis of EtherCAT and PROFINET IRT

Digital Systems Design Based on DSP algorithms in FPGA for Fault Identification in Rotary Machines

Maintenance Scheduling Optimization for 30kt Heavy Haul Combined Train in Daqin Railway

1. Overview of Nios II Embedded Development

Frequency Range Extension of Spectrum Analyzers with Harmonic Mixers

A Similarity Search Scheme over Encrypted Cloud Images based on Secure Transformation

Image Compression through DCT and Huffman Coding Technique

Analecta Vol. 8, No. 2 ISSN

1. Overview of Nios II Embedded Development

A MPCP-Based Centralized Rate Control Method for Mobile Stations in FiWi Access Networks

Extraction of Satellite Image using Particle Swarm Optimization

Noise estimation in remote sensing imagery using data masking

FIXED INCOME ATTRIBUTION

Combining an Alternating Sequential Filter (ASF) and Curvelet for Denoising Coronal MRI Images

Doppler Shift Estimation for TETRA Cellular Networks

Medical Image Segmentation of PACS System Image Post-processing *

Detection and Restoration of Vertical Non-linear Scratches in Digitized Film Sequences

COMPARISON OF OBJECT BASED AND PIXEL BASED CLASSIFICATION OF HIGH RESOLUTION SATELLITE IMAGES USING ARTIFICIAL NEURAL NETWORKS

Real-Time Reservoir Model Updating Using Ensemble Kalman Filter Xian-Huan Wen, SPE and Wen. H. Chen, SPE, ChevronTexaco Energy Technology Company

PARAMETRIC DISTRIBUTION FAMILIES USEFUL FOR MODELLING NON-LIFE INSURANCE PAYMENTS DATA. TAIL BEHAVIOUR. Sandra Teodorescu 1

Intensity transformations

Morphological segmentation of histology cell images

PIXEL-LEVEL IMAGE FUSION USING BROVEY TRANSFORME AND WAVELET TRANSFORM

Why focus on assessment now?

AM Receiver. Prelab. baseband

A Fast Algorithm for Multilevel Thresholding

Signature Region of Interest using Auto cropping

Retrospective Test for Loss Reserving Methods - Evidence from Auto Insurers

SIMPLIFIED CBA CONCEPT AND EXPRESS CHOICE METHOD FOR INTEGRATED NETWORK MANAGEMENT SYSTEM


Multisensor Data Fusion and Applications

MetaMorph Software Basic Analysis Guide The use of measurements and journals

Vision based Vehicle Tracking using a high angle camera

!!! Technical Notes : The One-click Installation & The AXIS Internet Dynamic DNS Service. Table of contents

8. ENERGY PERFORMANCE ASSESSMENT OF COMPRESSORS 8.1 Introduction

The Geometry of Perspective Projection

Euler Vector: A Combinatorial Signature for Gray-Tone Images

Manufacturing and Fractional Cell Formation Using the Modified Binary Digit Grouping Algorithm

THE EFFECT OF THE SPINDLE SYSTEM ON THE POSITION OF CIRCULAR SAW TEETH A STATIC APPROACH

COLOR-BASED PRINTED CIRCUIT BOARD SOLDER SEGMENTATION

Simultaneous Gamma Correction and Registration in the Frequency Domain

The Dynamic Background Generation Scheme Using an Image Frame

8. Hardware Acceleration and Coprocessing

The Design and Implementation of a Solar Tracking Generating Power System

An Energy-Based Vehicle Tracking System using Principal Component Analysis and Unsupervised ART Network

Quality assurance of solar thermal systems with the ISFH- Input/Output-Procedure

Appendix C White Paper to Support Bicycle Facility Design Toolkit on Lane Width Design Modifications DRAFT

8. Exception Handling

Linear and Non-linear Contrast Enhancement Image

A Simple Feature Extraction Technique of a Pattern By Hopfield Network

Image Forensic Analyses that Elude the Human Visual System

A Method of Caption Detection in News Video

Analysis of Antenna Mispointing Effect on SAR Image Quality

A Proposal for Estimating the Order Level for Slow Moving Spare Parts Subject to Obsolescence

Swept Sine Chirps for Measuring Impulse Response

Thresholding technique with adaptive window selection for uneven lighting image

Drainage Network Extraction from a SAREX 92 RADAR image

Mining the Situation: Spatiotemporal Traffic Prediction with Big Data

2. Getting Started with the Graphical User Interface

Parallel Data Selection Based on Neurodynamic Optimization in the Era of Big Data

Time-Optimal Online Trajectory Generator for Robotic Manipulators

A Dynamic Approach to Extract Texts and Captions from Videos

Wind Load Provisions of the Revised Building Code in Japan

Circle Object Recognition Based on Monocular Vision for Home Security Robot

Offshore Oilfield Development Planning under Uncertainty and Fiscal Considerations

Risk and Return: Estimating Cost of Capital

A logistic regression approach to estimating customer profit loss due to lapses in insurance

Options on Stock Indices, Currencies and Futures

Introduction. Chapter 1

Transcription:

A RAPI METHO FOR WATER TARGET EXTRACTION IN SAR IMAGE WANG ong a, QIN Ping a ept. o Surveying and Geo-inormatic Tongji University, Shanghai 200092, China. - wangdong75@gmail.com ept. o Inormation, East China Sea Marine Prediction Center, State Oceanic Administrative, Shanghai 200082, China. - qinping00@gmail.com Commission VII, WG VII/2 KEY WORS: Synthetic Aperture Radar Imagery, Water Target, Automatic Recognition, Grey Morphology, Nonlinear Filter ABSTRACT: For the inherent speckles in SAR imagery, the processing o oject extraction ecomes more diicult. Using the traditional target extraction arithmetic or optical imagery, we can not get ideal result on SAR imagery. Water oject in SAR imagery represents as some continuous region with low lightness and strong speckle noise. The traditional method to extract low lightness oject can e implement y segment the imagery to inary imagery using a speciy threshold. As the strong eect o the inherent speckle, how to gain an appropriate threshold is diicult. In this paper, the authors propose a sequential nonlinear iltering-ased method to extract the water oject in SAR imagery automatically. The experimental results using real SAR imagery show that our method can extract water ojects in SAR imagery rapidly and accurately. 1. INTROUCTION Synthetic Aperture Radar (SAR) can generate images o the ground in all weather conditions including rain and og at any time o day or night. It is a kind o important equipment or modern electric reconnaissance, widely used in civil and military. In recent year SAR imaging has een rapidly gaining prominence in applications such as remote sensing, surace surveillance, and automatic target recognition. In the navigation application o military, extracting the airport, water region, road, etc. in SAR imagery is important to produce the reerence imagery or matching with the real-time imagery. ierent rom some other remote sensing data, the SAR image do not oey the rule o projection ut the rule o time o light, and the SAR image data is most sensile on the geometric character o the surveyed oject, not only on the macroscopical level, ut also on the microscopical level. Both the natural character and the status o the oject are important parameters o SAR image. In the application o SAR image automatic target recognition, it can seldom get good eect y using the traditional oject extraction region segmentation operator which are requently used in optical images. A numer o algorithms or SAR image segmentation and target extraction have een proposed. In the case o low contrast and strong intererence, BI did the detection processing to extract some target with speciic shape in the remote sensing images on the method ased on the mathematical morphology theory (BI, 2005). In order to extract airport ojects in real aperture radar image, CHEN employed radon transorm technique to detect the runway and match the result with SAR imagery (CHEN, 2006). XIA proposed a method or oject extraction and classiication in SAR imagery using the threshold segmentation and character extraction techniques (XIA, 2005). In this paper, the authors proposed an algorithm ased on grey morphology and sequential nonlinear iltering to extract the water ojects in SAR imagery. 2. WATER OBJECTS EXTRACTION BASE ON GREY MORPHOLOGY AN SEQUENTIAL NONLINEAR FILTERING In the SAR imagery, water area represents as the region which has low lightness. Now, using threshold segmentation method to transorm the original image to inary image is mostly adopted to extract low lightness oject. Aected y the intense inherent noise, choosing an appropriate threshold automatically is an important and sophisticated theme. Using the method ased on hypothesis testing, ater interaction many time the threshold can e otained. In this paper, ased on the lightness o the water oject and the spatial distriution characteristic o it, the method ased on image segmentation threshold algorithm is not used. Employed the grey morphology and nonlinear iltering technique the author proposes a method to extract the water region in SAR image rapidly and eectively. From the asic gray-scale morphology operator, we can uild up a sequential nonlinear iltering model to extract water ojects in SAR imagery. Gray-scale dilation o y, denoted, is deined as ( )( max{ ( s x, t + ( x, ( s x),( t Gray-scale erosion o y, denoted Θ, is deined as (2.1) 237

The International Archives o the Photogrammetry, Remote Sensing and Spatial Inormation Sciences. Vol. XXXVII. Part B6. Beijing 2008 ( Θ)( min{ ( s + x, t + ( x, ( s + x),( t + (2.2) In Equations (2.1) and (2.2), and are the domains o and, respectively. For digital image processing, unction is input image, unction is structuring element, itsel a suimage unction. At each position o the structuring element the values o dilation and erosion at that point are the maximum value o ( + ) and minimum value o ( - ) in the interval spanned y. Gray-scale dilation and erosion are duals with respect to unction complementation and relection. The general eect o perorming dilation on a gray-scale image is twoold: (1) i all the values o the structuring element are positive, the output image tends to e righter than the input (2) dark details either are deduced or eliminated, depending on how their values and shapes relate to the structuring element. And or erosion operator, the eect is also dual: (1) i all the elements o structuring element are positive, the output image tends to e darker than the input image (2) the eect o right details in the input image that are smaller in area than the structuring element is reduced, with the degree o reduction eing determined y the gray-level values surrounding the right detail y the shape and amplitude values o the structuring element itsel. As the water ojects in SAR image represents as some consecutive regions consist o pixels with low lightnes we can comine the gray-level dilation and erosion to perorm graylevel closing operator. First perorm gray dilation to eliminate inconsecutive dark pixels and the water ojects were reduced as well and then perorm gray erosion to restore the water ojects without reintroducing the details removed y dilation (Easanuruk, 2005 Shih, 1998). As the inherent strong speckle noise in SAR imagery, we also introduced the order-statistics ilters to improve the method (Pita 1996). Order-statistics ilters are nonlinear spatial ilters whose response is ased on ordering (ranking) the pixels contained in the image area encompassed y the ilter, and then replacing the value o the center pixel with the value determined y the ranking result. Median ilter, as the most used order-statistics ilter, replaces the value o a pixel y the median o the gray levels in the neighourhood o that pixel (the original value o the pixel is included in the computation o the median). For certain types o random noise and particularly the pulse noise, they provide excellent noise-reduction capailitie with consideraly less lurring than linear smoothing ilters. The speckle in SAR imagery is multiplicative noise, and can e reduced y median ilter. The max ilter and min ilter are also order-statistics ilters. The max ilter using the 100th percentile results and is useul or inding the rightest points in an image. The min ilter is the 0th percentile and is useul to ind the darkest points in an image. They are also applicale or water oject extraction in SAR imagery. In this paper, we comine the gray-level morphology and nonlinear order-statistics ilter to make the oject extraction processing more eectively. We deine a speciic structuring element, and transorm the gray-level morphology operator to nonlinear ilter processing. Ater thi we can employ the sequential nonlinear ilter to extract water ojects in SAR imagery. Based on equations (2.1) and (2.2), deine as the N x N neighorhood, and unction as zero unction, that is its values in domain always are zero. So the equations (2.1) and (2.2) can transorm to equations (2.3) and (2.4) respectively. ( )( max{ ( s x, t ( Θ)( min{ ( s + x, t + ( s x), ( t ( s + x), ( t + (2.3) (2.4) So we can transorm the gray-level dilation processing to max ilter operator, and gray-level erosion to min ilter. In this paper, we comine the nonlinear ilters to uild up a sequential nonlinear ilter model. First we use max ilter to eliminate inconsecutive dark pixels in SAR image, and the water oject with consecutive distriution also e reduced as well. Second, we employed the median ilter to reduce the speckle noise. Finally we use the min ilter to restore the water ojects which were reduced in the irst step. In this paper, the size o neighourhood is 3x3. We can also perorm the max ilter several times in serie and perorm min ilter the same times as well. Ater these step we can get water ojects in SAR imagery and remove almost all the non-water ojects. Employed this sequential nonlinear ilter algorithm, we can extract low lightness pixels region eectively, and most o these region are water ojects. But it also contains some nonwater oject such as airport, road etc. So we need the advanced processing to select the water ojects rom the ilter results automatically. 3. WATER REGION MOEL AN WATER OBJECT AUTOMATIC SELECTION Ater the processing o sequential nonlinear ilter, we can get the low lightness targets in SAR imagery. It is necessary to mark the targets regions and select the water ojects rom the 238

The International Archives o the Photogrammetry, Remote Sensing and Spatial Inormation Sciences. Vol. XXXVII. Part B6. Beijing 2008 ilter results automatically. As the lightness o pixels in water region is very low, irst we can set the pixels which lightness over some speciic level (in this paper we select the middle level 128) to ackground lightness 255 and then employ the region grow algorithm to mark all the regions in the result image. For automatic selecting the water target we need calculate some characteristic o the marked regions. In this paper, we calculate the area value, average gray level and the histogram o every marked region. And ased on the characteristic o SAR image and knowledge o water oject, we uild up the water ojects model with the ollowing rules: (1) Area rule: regard the region which area less than some speciic value as non-water region. (2) Lightness rule: regard the region which average lightness or the lightness o the peak o histogram over some speciic value as non-water region. (3) Histogram distriution rule: as the water region consist o most low lightness pixels and some noise pixel its histogram should has a single peak in the let side. Thereore, we uild the histogram distriution rule regarding the region can not accord with all the ollowing conditions as non-water region: (a) gray level o the histogram peak less than the average gray o region ()the numer o pixels with gray level o the histogram peak more than 10 percent o total pixels in the region (c) the numer o pixels with gray level ranging rom the histogram peak to the peak level plus 5 more than 60 percent o total pixels in the region (d) the numer o pixels darker than the histogram peak level less than 1 percent o total pixels in the region. Based on these rules and the characteristic o region we can select the water ojects automatically. The Figure 1 is the lowchart o the approach proposed in this paper. Figure 1. Flowchart o the proposed approach 4. EXPERIMENTAL RESULTS We do experiments using the Wujiang SAR image which has water region, airport region and mountains. The result proved that the method proposed in this paper is easile and valid, and the water oject in the SAR image can e extracted rapidly and eiciently. Figure 2 is the original SAR imagery region (512 x 512). Figure 3 is the result image ater the sequential nonlinear iltering. Figure 4 is the result ater water oject selection automatically. Figure 5 is the water ojects extract y algorithm proposed in this paper. 239

The International Archives o the Photogrammetry, Remote Sensing and Spatial Inormation Sciences. Vol. XXXVII. Part B6. Beijing 2008 Figure 2. Original SAR imagery Figure 4. Result imagery ater water oject selection Figure 3. Result imagery ater sequential nonlinear iltering Figure 5. Water ojects in SAR imagery 240

The International Archives o the Photogrammetry, Remote Sensing and Spatial Inormation Sciences. Vol. XXXVII. Part B6. Beijing 2008 5. CONCLUSIONS Based on the characteristic o water target, we proposed a method employing sequential nonlinear ilter technique to extract water ojects in SAR imagery, and make experiments using real SAR imagery. The results show that this approach can extract water ojects in SAR imagery eectively and rapidly. For urther research, we can calculate more characteristic parameters o region, and apply this algorithm to other low lightness ojects extraction in SAR imagery. ACKNOWLEGMENT This work was supported y the Space Foundation o China under grant No.0747-0540SITC2099-4. REFERENCES BI Jin-yuan SONG Liang-tu, 2005. An Algorithm o Weak Target etection rom Remote Sensing ImageBased on Mathematic Morphology. Remote Sensing Inormation, pp.3-6. CHEN Yingying SHAO Yongshe, 2006. A Radon Transorm- Based Algorithm or the etection and Matching o Airport Ojects. Journal o Tongji University(Natural Science). Easanuruk, S. Mitatha, S. Intajag, S. Chitwong, S., 2005. Speckle Reduction Using Fuzzy Morphological Anisotropic iusion. In: 2005 Fith International Conerence on Inormation, Communications and Signal Processing. 12(06-09), pp.748-751. Pita I., 1996. State o the art o morphological and nonlinear digital image processing. In: proceedings o IEE Colloquium on Morphological and Nonlinear Image Processing Techniques. 10, pp.1-4. Shih, F.Y. Mitchell, O.R., 1998. Automated ast recognition and location o aritrarily shaped ojects y image morphology. In: Proceedings o Computer Society Conerence on CVPR 88. 6 (5-9), pp.774-779. XIA Xin LUO ai-sheng LUO Feng LIN Hong-jin, 2005. An Eicient Method or Typical Target Feature Extraction and Recognition in SAR Images. Journal o Sichuan University (Natural Science Edition). 42(5), pp.969-973. 241

The International Archives o the Photogrammetry, Remote Sensing and Spatial Inormation Sciences. Vol. XXXVII. Part B6. Beijing 2008 242