一种基于小波变换的SAR图像舰船尾迹检测算法
|
摘要
合成孔径雷达(SAR)是利用雷达与目标的相对运动把尺寸较小的真实天线孔径用数据处理的方法合成一较大的等效天线孔径的雷达。合成孔径雷达工作时按一定的重复频率发、收脉冲,真实天线依次占一虚构线阵天线单元位置。把这些单元天线接收信号的振幅与相对发射信号的相位叠加起来,便合成一个等效合成孔径天线的接收信号。合成孔径雷达的特点是分辨率高,能全天候工作,能有效地识别伪装和穿透掩盖物。主要用于航空测量、航空遥感、卫星海洋观测、航天侦察、图像匹配制导等。
舰船检测与监视是世界各海岸地带国家的传统任务,在民用及公安部门有广泛的应用,例如:舰船的寻找和救助、捕渔船监视、非法移民、保卫领土、反毒品等等。由于合成孔径雷达(SAR)具有全天时和全天候的观测能力,能够宏观、长期、动态、实时地对陆地和海洋进行观测,近年来,利用SAR图像进行舰船检测的研究和技术开发在海洋遥感领域得到高度重视,成为SAR最重要的海洋应用之一。
本论文首先介绍了合成孔径雷达图像上海面舰船尾迹的特征以及广泛用于SAR图像上线性目标检测的Radon变换算法,同时也介绍了小波变换的特点和应用。由于Radon变换出色的抗噪性能,目前大部分SAR图像舰船尾迹检测方法都是基于Radon变换的线性检测算法,国外的研究人员如Murphy对SAR图像进行Radon变换,在变换域进行增强处理,再用滤波逆投影算法变换回图像域;Califano等人将Radon变换限制在窗口内;Copeland等人则定义了沿线段积分的Radon变换;国内的研究人员有的利用形态学图像处理技术剔除非尾迹峰值点的干扰,降低虚警率;有的则采用邻近像素求和方法来实现恒虚警率(CFAR)检测。
其次,论文具体讨论了应用Radon算法检测合成孔径雷达图像舰船尾迹过程中,Radon算法的主要缺陷以及常规Radon变换域中尾迹峰值检测的不确定性问题,针对这些问题,论文提出了一种基于小波变换的尾迹检测算法。该算法首先对原始SAR图像进行预处理,初步消除强目标点和孤立噪声的影响;然后进行Radon变换,在变换域中通过自适应域值的方法提取所有可能的局部峰值点,对这些峰值的一维“截面”进行连续小波变换峰值匹配生成相关决策参数,根据提取到的参数,通过设定科学的判决函数在特征空间进行决策,得到真实的尾迹线性特征;最后在原始SAR图像利用方向梯度算法确定已检测到的尾迹起点和走向。
最后,通过真实ERS-1 SAR图像上的测试,证明了该方法能有效消除Radon变换的多个缺陷以及虚假线性特征带来的不利影响,降低检测的虚警率,准确有效地捕捉到SAR图像中的舰船尾迹。
Synthetic Aperture Radar (SAR) is a form of radar in which, the large antenna aperture radar is equivalently replaced with a number of real small antenna aperture radars by data processing, taking advantage of the relative motion between the sensors and the objectives. The SAR sensor sends and receives the pulses in a certain frequency and the real antenna by turns occupies the units of a virtual linear array antenna. The received signal of an equivalent synthetic aperture antenna is formed by combining the amplitude of the received signals of the antenna units and the phases relative to the send signals. Featured by high resolution, SAR can effectively identify the real target and penetrate the cover regardless of the weather condition. SAR is widely applied in aviation, aerospace remote sensing, satellite observation of the marine, aerospace reconnaissance, image matching guidance and so on.
Ship detection and surveillance is the traditional task of the world’s costal countries, which has a wide range of application in civilian and public security departments such as search and rescue ships, surveillance of fishing vessels, illegal immigration, defend territory and anti-drug. As the Synthetic Aperture Radar (SAR) has the ability of macro, long-term, dynamic and real-time on land and sea observations regardless of the time and weather, in recent years, great importance has been attached to ship detection research and technology development using SAR images in ocean remote sensing, which has become one of the most important applications of SAR ocean image.
Firstly, the dissertation introduces the feature of the ship wakes in SAR ocean images and the Radon transform which has been widely used in linear target detection in SAR images. In addition, the feature and application of the wavelet transform is introduced as well. At present, due to the exceptional anti-noise characteristic of the Radon transform, most ship wake detection algorithms in SAR image are based on Radon transform. A great deal of research has been dedicated to developing Radon transform-based algorithms for linear targets. Murphy proves the viability of the technique for both simulated and real SAR images. Skingley and Rey address post-processing problems, including the detection of peaks and troughs in the Hough domain and the removal of false alarms. Califano develops the multiple window parameter transform based on Radon transform algorithm. All these algorithms improve certain deficiencies of the Radon transform and reduce the false alarm rate. Some domestic researchers use mathematical morphologic image processing technology to reduce the effect of the fake wake peaks in the Radon domain, thus reducing the false alarm rate. Some use the technique of“Adjacent pixels summing(APS)”to develop the CFAR detection algorithm.
Secondly, in this dissertation aimed at the problem of the drawbacks of Radon transform and the uncertainty in detecting the ship wake peaks in the conventional Radon transform domain, a detection algorithm based on wavelet transform is developed. Firstly the algorithm preprocesses the original SAR image to eliminate the strong and isolated noise. Secondly it extracts all possible local peaks, performs continuous wavelet transform to match all these one-dimensional local peak sections and make decision in the feature space using the decision vectors formed from the extracted parameters. Finally, a maximum directional-derivative method is used to locate the starting points of the detected ship wakes.
Finally, the algorithm is tested on real ERS-1 SAR images and the results demonstrate its effectiveness in reducing the adverse effect caused by the drawbacks of Radon transform and fake linear features, which reduces the false alarm rate and captures the real ship wakes in SAR image accurately
舰船检测与监视是世界各海岸地带国家的传统任务,在民用及公安部门有广泛的应用,例如:舰船的寻找和救助、捕渔船监视、非法移民、保卫领土、反毒品等等。由于合成孔径雷达(SAR)具有全天时和全天候的观测能力,能够宏观、长期、动态、实时地对陆地和海洋进行观测,近年来,利用SAR图像进行舰船检测的研究和技术开发在海洋遥感领域得到高度重视,成为SAR最重要的海洋应用之一。
本论文首先介绍了合成孔径雷达图像上海面舰船尾迹的特征以及广泛用于SAR图像上线性目标检测的Radon变换算法,同时也介绍了小波变换的特点和应用。由于Radon变换出色的抗噪性能,目前大部分SAR图像舰船尾迹检测方法都是基于Radon变换的线性检测算法,国外的研究人员如Murphy对SAR图像进行Radon变换,在变换域进行增强处理,再用滤波逆投影算法变换回图像域;Califano等人将Radon变换限制在窗口内;Copeland等人则定义了沿线段积分的Radon变换;国内的研究人员有的利用形态学图像处理技术剔除非尾迹峰值点的干扰,降低虚警率;有的则采用邻近像素求和方法来实现恒虚警率(CFAR)检测。
其次,论文具体讨论了应用Radon算法检测合成孔径雷达图像舰船尾迹过程中,Radon算法的主要缺陷以及常规Radon变换域中尾迹峰值检测的不确定性问题,针对这些问题,论文提出了一种基于小波变换的尾迹检测算法。该算法首先对原始SAR图像进行预处理,初步消除强目标点和孤立噪声的影响;然后进行Radon变换,在变换域中通过自适应域值的方法提取所有可能的局部峰值点,对这些峰值的一维“截面”进行连续小波变换峰值匹配生成相关决策参数,根据提取到的参数,通过设定科学的判决函数在特征空间进行决策,得到真实的尾迹线性特征;最后在原始SAR图像利用方向梯度算法确定已检测到的尾迹起点和走向。
最后,通过真实ERS-1 SAR图像上的测试,证明了该方法能有效消除Radon变换的多个缺陷以及虚假线性特征带来的不利影响,降低检测的虚警率,准确有效地捕捉到SAR图像中的舰船尾迹。
Synthetic Aperture Radar (SAR) is a form of radar in which, the large antenna aperture radar is equivalently replaced with a number of real small antenna aperture radars by data processing, taking advantage of the relative motion between the sensors and the objectives. The SAR sensor sends and receives the pulses in a certain frequency and the real antenna by turns occupies the units of a virtual linear array antenna. The received signal of an equivalent synthetic aperture antenna is formed by combining the amplitude of the received signals of the antenna units and the phases relative to the send signals. Featured by high resolution, SAR can effectively identify the real target and penetrate the cover regardless of the weather condition. SAR is widely applied in aviation, aerospace remote sensing, satellite observation of the marine, aerospace reconnaissance, image matching guidance and so on.
Ship detection and surveillance is the traditional task of the world’s costal countries, which has a wide range of application in civilian and public security departments such as search and rescue ships, surveillance of fishing vessels, illegal immigration, defend territory and anti-drug. As the Synthetic Aperture Radar (SAR) has the ability of macro, long-term, dynamic and real-time on land and sea observations regardless of the time and weather, in recent years, great importance has been attached to ship detection research and technology development using SAR images in ocean remote sensing, which has become one of the most important applications of SAR ocean image.
Firstly, the dissertation introduces the feature of the ship wakes in SAR ocean images and the Radon transform which has been widely used in linear target detection in SAR images. In addition, the feature and application of the wavelet transform is introduced as well. At present, due to the exceptional anti-noise characteristic of the Radon transform, most ship wake detection algorithms in SAR image are based on Radon transform. A great deal of research has been dedicated to developing Radon transform-based algorithms for linear targets. Murphy proves the viability of the technique for both simulated and real SAR images. Skingley and Rey address post-processing problems, including the detection of peaks and troughs in the Hough domain and the removal of false alarms. Califano develops the multiple window parameter transform based on Radon transform algorithm. All these algorithms improve certain deficiencies of the Radon transform and reduce the false alarm rate. Some domestic researchers use mathematical morphologic image processing technology to reduce the effect of the fake wake peaks in the Radon domain, thus reducing the false alarm rate. Some use the technique of“Adjacent pixels summing(APS)”to develop the CFAR detection algorithm.
Secondly, in this dissertation aimed at the problem of the drawbacks of Radon transform and the uncertainty in detecting the ship wake peaks in the conventional Radon transform domain, a detection algorithm based on wavelet transform is developed. Firstly the algorithm preprocesses the original SAR image to eliminate the strong and isolated noise. Secondly it extracts all possible local peaks, performs continuous wavelet transform to match all these one-dimensional local peak sections and make decision in the feature space using the decision vectors formed from the extracted parameters. Finally, a maximum directional-derivative method is used to locate the starting points of the detected ship wakes.
Finally, the algorithm is tested on real ERS-1 SAR images and the results demonstrate its effectiveness in reducing the adverse effect caused by the drawbacks of Radon transform and fake linear features, which reduces the false alarm rate and captures the real ship wakes in SAR image accurately
引文
[1]向敬成,张明友.雷达系统.电子工业出版社,2001,188~189
[2]崔麦会,周建军,陈超.合成孔径雷达的物理原理及其在军事上的应用.国防技术基础,2004,(4),4~5
[3]曹忠权,谢平,金花.星载合成孔径雷达遥感技术的地学应用.地球物理学进展,2004,19(2):291~295
[4]匡纲要,高贵,蒋咏梅.合成孔径雷达目标检测理论、算法及应用.国防科技大学出版社,2007,8~9
[5]周红建,李相迎,彭雄宏.从卫星SAR海洋图像中检测船目标.国防科技大学学报,1999,21(1),67~84
[6] J. K. Tunaley, E. H. Buller, K. H. Wu, M. T. Rey. The simulation of the SAR image of a ship wake[J],IEEE Trans. on GRS,1991,29(1): 149~156.
[7]王爱明,朱敏慧.海洋舰船尾迹合成孔径雷达成像仿真研究.北京:中国科学院电子学研究所,2002
[8]张宇,张永刚,黄韦艮.一种利用SAR图像检测船舶尾迹的方法.国土资源遥感,2003,(1),56~58
[9]邹焕新,匡纲要,郁文贤.基于特征空间决策的SAR图像舰船尾迹检测算法.系统工程与电子技术,2004,26(6),728~730
[10] Lyden J D ,et al . Analysis of synthetic aperture radar imagery of surface ship wakes [A]. Proceedings of IGARSS′86 [C]. Zurich: IGARSS, 8-11 Sept, 1986. 801~805.
[11] L. M. Murphy, Linear feature detection and enhancement in noisy images via the Radon transform [J]. Pattern Recognition Lett., Sept. 1986, vol. 4, no.4, pp. 279~284
[12] Eldhuset K. Automatic ship and ship wake detection in spaceborne SAR images from coastal regions [A]. Proceedings of IGARSS′88 [C] . Edinburgh, Scotland :IGARSS ,13 - 16 Sept ,1988. 1529~1533.
[13] Hendry A, et al . Automated linear feature detection and its application to curve location in synthetic aperture radar imagery [A]. Proceedings of IGARSS′88 [ C] . Edinburgh , Scotland : IGARSS , 13~16 Sept , 1988. 1521~1524.
[14] M. T. Rey, J. K. Tunaley, J. T. Folinsbee, P. A. Jahans, J. A. Dixon, and M. R. Vant, Application of Radon transform techniques to wake detection in Seasat-A SAR images [J]. IEEE Trans. Geosci. Remote Sensing, July 1990, vol. 28, pp. 553~560
[15] J. Skingley and A. J. Rye,“The Hough Transform Applied to SAR images for Thin Line Detection,”Pattern Recognition Letters, Jun. 1987, vol. 6, pp. 61~67
[16] A. Califano and R. Bolle, The multiple window parameter transform [J]. IEEE Trans. Pattern Anal. Mach. Intell., Dec. 1992, vol. 14, pp. 1157~1170
[17] A. C. Copeland, Gopalan Ravichandran and M. M. Trivedi, Localized Radon transform-based detection of ship wakes in SAR images [J]. IEEE Trans. Geosci. Remote Sensing, Jan. 1995, vol. 33, pp. 35~45
[18] Lin I-I, et al. Computer-based algorithm for ship detection from ERS SAR imagery [EB/OL]. http:/ / www. ers- symosium.org/papers/participants/data/lini/, 1997-03-18.
[19] Fitch J P, Lehman S K, Dowla F U , et al . Ship Wake-Detection Procedure Using Conjugate Gradient Trained Artificial Neural Network [J]. IEEE Transactions on Geoscience and Remote Sensing, 1991, 29 (5):718~726.
[20]王世庆,金亚秋. SAR图像船行尾迹检测的Radon变换和形态学图像处理技术[J].遥感学报,2001,5(4),289~294
[21]周红建,陈越,王正志,彭雄宏.应用Radon变换方法检测窄V形船舶航迹.中国图象图形学报,2000,5(11),902~905
[22]汤子跃,朱敏慧等.一种SAR图像舰船尾迹的CFAR检测方法[J ] .电子学报,2002,30(9),1336~1339.
[23]郑治真,沈萍,杨选辉,万玉莉.小波变换及其MATLAB工具的应用.地震出版社. 2001,1~3
[24]俞一彪,孙兵.数字信号处理理论与应用.东南大学出版社,2005,189~190
[25]钟诚.小波变换及其应用研究.中国科技信息. 2008,(2),70~71
[26]王立,常青,张科. Radon变换在低信噪比图像中的线段检测.红外与激光工程,2003,32(2),163~166
[27]胡昌华,张军波,夏军,等编著.基于MATLAB的系统分析与设计—小波分析.西安电子科技大学出版社,2000.
[2]崔麦会,周建军,陈超.合成孔径雷达的物理原理及其在军事上的应用.国防技术基础,2004,(4),4~5
[3]曹忠权,谢平,金花.星载合成孔径雷达遥感技术的地学应用.地球物理学进展,2004,19(2):291~295
[4]匡纲要,高贵,蒋咏梅.合成孔径雷达目标检测理论、算法及应用.国防科技大学出版社,2007,8~9
[5]周红建,李相迎,彭雄宏.从卫星SAR海洋图像中检测船目标.国防科技大学学报,1999,21(1),67~84
[6] J. K. Tunaley, E. H. Buller, K. H. Wu, M. T. Rey. The simulation of the SAR image of a ship wake[J],IEEE Trans. on GRS,1991,29(1): 149~156.
[7]王爱明,朱敏慧.海洋舰船尾迹合成孔径雷达成像仿真研究.北京:中国科学院电子学研究所,2002
[8]张宇,张永刚,黄韦艮.一种利用SAR图像检测船舶尾迹的方法.国土资源遥感,2003,(1),56~58
[9]邹焕新,匡纲要,郁文贤.基于特征空间决策的SAR图像舰船尾迹检测算法.系统工程与电子技术,2004,26(6),728~730
[10] Lyden J D ,et al . Analysis of synthetic aperture radar imagery of surface ship wakes [A]. Proceedings of IGARSS′86 [C]. Zurich: IGARSS, 8-11 Sept, 1986. 801~805.
[11] L. M. Murphy, Linear feature detection and enhancement in noisy images via the Radon transform [J]. Pattern Recognition Lett., Sept. 1986, vol. 4, no.4, pp. 279~284
[12] Eldhuset K. Automatic ship and ship wake detection in spaceborne SAR images from coastal regions [A]. Proceedings of IGARSS′88 [C] . Edinburgh, Scotland :IGARSS ,13 - 16 Sept ,1988. 1529~1533.
[13] Hendry A, et al . Automated linear feature detection and its application to curve location in synthetic aperture radar imagery [A]. Proceedings of IGARSS′88 [ C] . Edinburgh , Scotland : IGARSS , 13~16 Sept , 1988. 1521~1524.
[14] M. T. Rey, J. K. Tunaley, J. T. Folinsbee, P. A. Jahans, J. A. Dixon, and M. R. Vant, Application of Radon transform techniques to wake detection in Seasat-A SAR images [J]. IEEE Trans. Geosci. Remote Sensing, July 1990, vol. 28, pp. 553~560
[15] J. Skingley and A. J. Rye,“The Hough Transform Applied to SAR images for Thin Line Detection,”Pattern Recognition Letters, Jun. 1987, vol. 6, pp. 61~67
[16] A. Califano and R. Bolle, The multiple window parameter transform [J]. IEEE Trans. Pattern Anal. Mach. Intell., Dec. 1992, vol. 14, pp. 1157~1170
[17] A. C. Copeland, Gopalan Ravichandran and M. M. Trivedi, Localized Radon transform-based detection of ship wakes in SAR images [J]. IEEE Trans. Geosci. Remote Sensing, Jan. 1995, vol. 33, pp. 35~45
[18] Lin I-I, et al. Computer-based algorithm for ship detection from ERS SAR imagery [EB/OL]. http:/ / www. ers- symosium.org/papers/participants/data/lini/, 1997-03-18.
[19] Fitch J P, Lehman S K, Dowla F U , et al . Ship Wake-Detection Procedure Using Conjugate Gradient Trained Artificial Neural Network [J]. IEEE Transactions on Geoscience and Remote Sensing, 1991, 29 (5):718~726.
[20]王世庆,金亚秋. SAR图像船行尾迹检测的Radon变换和形态学图像处理技术[J].遥感学报,2001,5(4),289~294
[21]周红建,陈越,王正志,彭雄宏.应用Radon变换方法检测窄V形船舶航迹.中国图象图形学报,2000,5(11),902~905
[22]汤子跃,朱敏慧等.一种SAR图像舰船尾迹的CFAR检测方法[J ] .电子学报,2002,30(9),1336~1339.
[23]郑治真,沈萍,杨选辉,万玉莉.小波变换及其MATLAB工具的应用.地震出版社. 2001,1~3
[24]俞一彪,孙兵.数字信号处理理论与应用.东南大学出版社,2005,189~190
[25]钟诚.小波变换及其应用研究.中国科技信息. 2008,(2),70~71
[26]王立,常青,张科. Radon变换在低信噪比图像中的线段检测.红外与激光工程,2003,32(2),163~166
[27]胡昌华,张军波,夏军,等编著.基于MATLAB的系统分析与设计—小波分析.西安电子科技大学出版社,2000.