基于船载雷达图像的海上目标检测技术研究
|
摘要
导航雷达是舰船航行中必不可少的目标探测工具,雷达目标的检测和自动跟踪在军用和民用方面都具有重要价值。导航雷达在实际应用中受各种杂波、噪声和同频干扰等影响,雷达回波信号和雷达图像受到严重影响,目标检测率亟待提高,针对多谱探测预警课题的实际问题,本文在以下几个方面进行了理论研究和实验验证。
首先,本文对导航雷达图像预处理和目标检测技术的历史及现状进行了分析和总结,根据调研所了解到的情况对多谱预警中雷达图像目标检测的课题作了详细的规划,先对雷达图像进行预处理研究,然后进行目标检测。
雷达图像预处理部分：先对雷达图像特点作了分析,然后重点分析中值滤波和小波图像去噪,用中值滤波、自适应中值滤波、小波阈值降噪、小波结合中值或维纳滤波等方法对实采雷达图像做了仿真和分析,从PSNR值来看,小波图像去噪比中值滤波在PSNR值上有很大提高,中值滤波中十字形(或菱形)窗口比矩形和叉形窗口效果好,小波与中值、维纳结合的PSNR比小波单独去噪没有明显优势；在视觉效果上,维纳滤波与小波结合去噪的效果要优于小波单独去噪,全面衡量得出sym4小波降噪最优
雷达目标检测部分：分析了阈值分割法和边缘检测法,阈值分割法中Otsu和Ridler计算的阂值偏大,经典的KSW和最小误差法取得的阈值又偏小,针对雷达图像提出的改进的KSW法对怎样选取合适的权重系数没有理论可供参考,本文在已经提出的针对雷达图像改进的KSW方法的基础上,提出一种选择合适的权重系数的方法,最后用形态学滤波进行后处理；目标边缘检测中用四种不同的检测算子进行处理,其中Canny算子检测结果最好。实验结果表明改进的阈值计算法可以很好地分割目标,阈值分割法比边缘检测法检测结果好。
最后进行目标判别,因为在雷达图像中可能会有杂波的强反射回波被误检为目标,也可能会有目标回波太弱而被漏检,本文采取如下判别方法：对连续3帧雷达图像进行改进的阈值分割法的目标检测,将3帧检测结果中的任意两帧进行与运算剔除虚警,然后进行或运算,最后结合先验知识进行判别。实验结果证明该方法可以准确地判断出目标。
Navigation radar target detection is essential to ship, radar target detection and automatic tracking are of great value in terms of military and civilian. Navigation radar plays an important role during ship navigation and collision avoidance. Navigation radar is effected by the clutter, noise, and the same frequency interference in practical applications, the radar echo signal and radar images have been seriously affected, the target detection rate needs to be improved, the practical problems of multi-spectral detection of early warning topics, in the following several aspects of the theoretical study and experimental verification.
Firstly, the history and current situation of the marine radar and radar target detection are analyzed and summarized.Then it gives a detailed study plan for the topic multi-spectral target detection,radar image pre-processing research and target detection.
Part of the radar image pre-processing:first the characteristics of radar images are analyzed and the radar image pre-processing methods are summed up from the references.Then it focused on the median filtering and wavelet denoising methods.In median filtering it discussed different template shapes and window size of the effect for image preprocessing. It analyzed the adaptive median filter. In wavelet threshold denoising, wavelet basis, decomposition scale and the choice of threshold function are discussed, and then use the median filtering,adaptive median filtering, wavelet threshold denoising, wavelet combined with the median and wiener filtering methods to process real radar images. From the PSNR point of view, the wavelet image denoising PSNR has greatly improved than the median filter and cross window median filter does better than the rectangular and diamond-shaped window.Wavelet with median or wiener has no advantage in PSNR compared to wave denoising alone.From visual effects, wiener filtering with wavelet is better than the wavelet denoising, through a comprehensive measure the optimal is sym4wavelet.
Part of the radar target detection:analyzed threshold segmentation and edge detection.In threshold segmentation method the Otsu and Ridler calculated threshold is too large, the threshold obtained by KSW is too small, this paper presents an improved threshold calculation method, and finally use morphological filtering for processing; Edge detection used four different detection operator for processing, and test results were analyzed and summarized. Simulation results show that the threshold segmentation method of calculation based on the improved threshold detection results is better than the canny target edge detection method.
Finally, target discrimination.Because the strong echo of the clutter in the radar images may be false alarms as the goal, there may be a target echo is too weak to be missed, this article take the following discriminant method:use improved threshold segmentation method to do target detection for three continuous radar images, any two of the three test results do "and" operator, and then proceed to do "or" operator, and finally combined with a priori knowledge of discrimination. The experimental results show that the method can accurately determine the target.
首先,本文对导航雷达图像预处理和目标检测技术的历史及现状进行了分析和总结,根据调研所了解到的情况对多谱预警中雷达图像目标检测的课题作了详细的规划,先对雷达图像进行预处理研究,然后进行目标检测。
雷达图像预处理部分：先对雷达图像特点作了分析,然后重点分析中值滤波和小波图像去噪,用中值滤波、自适应中值滤波、小波阈值降噪、小波结合中值或维纳滤波等方法对实采雷达图像做了仿真和分析,从PSNR值来看,小波图像去噪比中值滤波在PSNR值上有很大提高,中值滤波中十字形(或菱形)窗口比矩形和叉形窗口效果好,小波与中值、维纳结合的PSNR比小波单独去噪没有明显优势；在视觉效果上,维纳滤波与小波结合去噪的效果要优于小波单独去噪,全面衡量得出sym4小波降噪最优
雷达目标检测部分：分析了阈值分割法和边缘检测法,阈值分割法中Otsu和Ridler计算的阂值偏大,经典的KSW和最小误差法取得的阈值又偏小,针对雷达图像提出的改进的KSW法对怎样选取合适的权重系数没有理论可供参考,本文在已经提出的针对雷达图像改进的KSW方法的基础上,提出一种选择合适的权重系数的方法,最后用形态学滤波进行后处理；目标边缘检测中用四种不同的检测算子进行处理,其中Canny算子检测结果最好。实验结果表明改进的阈值计算法可以很好地分割目标,阈值分割法比边缘检测法检测结果好。
最后进行目标判别,因为在雷达图像中可能会有杂波的强反射回波被误检为目标,也可能会有目标回波太弱而被漏检,本文采取如下判别方法：对连续3帧雷达图像进行改进的阈值分割法的目标检测,将3帧检测结果中的任意两帧进行与运算剔除虚警,然后进行或运算,最后结合先验知识进行判别。实验结果证明该方法可以准确地判断出目标。
Navigation radar target detection is essential to ship, radar target detection and automatic tracking are of great value in terms of military and civilian. Navigation radar plays an important role during ship navigation and collision avoidance. Navigation radar is effected by the clutter, noise, and the same frequency interference in practical applications, the radar echo signal and radar images have been seriously affected, the target detection rate needs to be improved, the practical problems of multi-spectral detection of early warning topics, in the following several aspects of the theoretical study and experimental verification.
Firstly, the history and current situation of the marine radar and radar target detection are analyzed and summarized.Then it gives a detailed study plan for the topic multi-spectral target detection,radar image pre-processing research and target detection.
Part of the radar image pre-processing:first the characteristics of radar images are analyzed and the radar image pre-processing methods are summed up from the references.Then it focused on the median filtering and wavelet denoising methods.In median filtering it discussed different template shapes and window size of the effect for image preprocessing. It analyzed the adaptive median filter. In wavelet threshold denoising, wavelet basis, decomposition scale and the choice of threshold function are discussed, and then use the median filtering,adaptive median filtering, wavelet threshold denoising, wavelet combined with the median and wiener filtering methods to process real radar images. From the PSNR point of view, the wavelet image denoising PSNR has greatly improved than the median filter and cross window median filter does better than the rectangular and diamond-shaped window.Wavelet with median or wiener has no advantage in PSNR compared to wave denoising alone.From visual effects, wiener filtering with wavelet is better than the wavelet denoising, through a comprehensive measure the optimal is sym4wavelet.
Part of the radar target detection:analyzed threshold segmentation and edge detection.In threshold segmentation method the Otsu and Ridler calculated threshold is too large, the threshold obtained by KSW is too small, this paper presents an improved threshold calculation method, and finally use morphological filtering for processing; Edge detection used four different detection operator for processing, and test results were analyzed and summarized. Simulation results show that the threshold segmentation method of calculation based on the improved threshold detection results is better than the canny target edge detection method.
Finally, target discrimination.Because the strong echo of the clutter in the radar images may be false alarms as the goal, there may be a target echo is too weak to be missed, this article take the following discriminant method:use improved threshold segmentation method to do target detection for three continuous radar images, any two of the three test results do "and" operator, and then proceed to do "or" operator, and finally combined with a priori knowledge of discrimination. The experimental results show that the method can accurately determine the target.
引文
[1]Memll I Skolnik雷达手册(第三版)[M].南京电子技术研究所.北京：电子工业出版社,2010：909-937.
[2]丁献文,万荣胜,王建,等.航海雷达图像噪声抑制方法研究[J].海洋技术,2011,30(3)：13-16.
[3]沈继红,李英,戴运桃,等.X波段雷达图像同频干扰的抑制方法研究[J].仪器仪表学报,2011,32(5)：1089-1094.
[4]高成志.基于X波段导航雷达海面回波图像的去噪和插值修复研究[D].天津：天津大学,2010.
[5]李鸿杰.XX雷达杂波抑制方法研究[D].南京：南京航空航天大学,2010.
[6]宋占杰,孙皓,王鑫,等.基于导航雷达海面回波图像去噪研究[J].海洋技术,2010,29(])：82-86.
[7]杨娜.基于小波变换的雷达图像处理[D].大连：大连海事大学,2009.
[8]郝燕玲,唐艳红,卢志忠.X波段航海雷达图像噪声检测与滤除方法研究[J].国土资源遥感,2008,2(76)：14-17.
[9]黄琼丹.基于雷达视频的目标检测和录取方法研究[D].西安：西安电子科技大学,2006.
[10]田守东.基于雷达图像的目标检测技术研究[D].哈尔滨：哈尔滨工程大学,2010.
[11]白素萍,张子鹤.基于数字图像处理的雷达图像动目标点迹提取和跟踪[J].舰船电子工程,1999,109：14-19.
[12]焦凤萍.合成孔径雷达图像的预处理方法研究[D].安徽：安徽大学,2007.
[13]Briggs J N. Target detection by marine radar[J]. Aerospace and Electronic Systems Magazine,2005,20 (6):39-40.
[14]Kabakchiev C, Garvanov I, Behar V. CFAR detection and parameter estimation of moving marine targets using forward scatter radar[J]. IEEE Conferences, Radar Symposium (IRS), Proceedings International,2011:85-90.
[15]Panagopoulos S, Soraghan J J. Small-target detection in sea clutter[J]. Geoscience and Remote Sensing, IEEE Transactions on,2004,42 (7):1355-1361.
[16]刘斌,杨劲松,范开国,等.基于船载雷达图像的海上船只检测方法[J].海洋学研究,2009,27(4)：33-37.
[17]种劲松.合成孔径雷达图像舰船目标检测算法与应用研究[D].北京：中国科学院研究生院,2002.
[18]种劲松,朱慧敏.高分辨率合成孔径雷达图像舰船目标检测方法[J].测试技术学报2003,17(1)：15-18.
[19]陈珊.合成孔径雷达图像上舰船目标的检测[D].上海：上海交通大学,2008.
[20]叶海军.基于纹理特征和数学形态学的SAR图像目标检测方法[J].中国电子科学研究院学报,2009,4(4)：436-440.
[21]陈晓楠,索继东,柳晓鸣.基于改进小波阈值函数的船用雷达回波去噪[J].沈阳工业大学学报,2010,32(2)：196-199.
[22]方莉,张萍.经典图像去噪算法研究综述[J].工业控制计算机,2010,23(11)：73-74.
[23]雷浩鹏.数字图像去噪算法研究及应用[D].长沙：长沙理工大学,2010.
[24]张黎,王立克,杨峰,等.小波阈值图像去噪研究与应用[J].图像处理,2006,10(3)：293-295.
[25]Donoho D L. De-noising by soft-thresholding [J]. IEEE Trans, on IT,1995,41(3):612-627.
[26]Donoho D L. Adapting to unknown smoothness via wavelet shrinkage [J]. Journal of American Statistical Association,1995,12(90):1200-1224.
[27]Wan Jun, Zhang Xiaohui, Rao Jionghui. Research and Application of Denoising Method Based on Wavelet Threshold[C]. Information Engineering and Computer Science (1CIECS),2nd International Conference on,2010:1-4.
[28]Qingwu Li, Chunyuan He. Application of Wavelet Threshold to Image De-noising[C]. Innovative Computing, Information and Control, First International Conference on,2006 (2):693-696.
[29]冯文强.非局部算法在图像去噪中的应用[D].安徽：中国科学技术大学,2011.
[30]韩思奇,王蕾.图像分割的阈值法综述[J].系统工程与电子技术,2002,24(6)：91-94.
[31]Otsu N. A Threshold Selection Method from Gray-Level Histograms[J]. IEEE Trans. Systems Man and Cybernetics,1979,9(1):62-66.
[32]Sapna Varshney S, Rajpal N, Purwar R. Comparative study of image segmentation techniques and object matching using segmentation[C]. Methods and Models in Computer Science, Proceeding of International Conference on,2009:1-6.
[33]Zhen Wang, Meng Yang. A fast clustering algorithm in image segmentation[C]. Computer Engineering and Technology,2nd International Conference on,2010 (6):592-594.
[34]吴一全,朱兆达.图像处理中闽值选取方法30年的进展(一、二)[J].数据采集与处理,1993,8(3)：193-201.
[35]Kittler J, Illingworth J. Minimum error thresholding[J]. Pattern Recognition,1986,19 (1):41-47.
[36]Veltkamp R C. Shape matching: similarity measures and algorithms[C]. Shape Modeling and Applications, International Conference on,2001:188-197.
[37]刘兴超.单帧红外图像弱小目标检测算法研究[D].武汉：华中科技大学,2005.
[38]李琦,傅俊诚,李自勤,等.激光雷达含噪图像边缘检测算法比较[J].2003,32(3)：239-243.
[39]连洁,韩传久,潘路.基于Canny算法的红外小目标边缘检测方法[J].微计算机信息,2007,23(6-3)：308-310.
[40]John Canny. A Computational Approach for Edge Detection[J]. IEEE Trans. Pattern Anal. Machine Intell,1986,8 (6):679-698.
[41]Vincent L. Morphological Grayscale Reconstruction in Image Analysis:Application and Efficient Algorithms[J]. IEEE Trans. On Image Processing,1993,2(2):176-201.
[42]DongPing Ming, Qun Wang, Jiancheng Luo. Evaluation of High Spatial Resolution Remote Sensing Image Segmentation Algorithms[C]. Image and Signal Processing,,2nd International Congress on,2009:1-5.
[43]Kapur J N, Sahoo P K. A New Method for Gray-level picture Thresholding Using the Entropy of the Histogram[J]. Computer Vision Graphics and Image Processing,1985, 29(2):273-285.
[44]Zhao Chunjiang, Deng Yong. A Modified Sobel Edge Detection Using Dempster-Shafer Theory[C]. Image and Signal Processing,2nd International Congress on,2009:1-4.
[45]Torre Vincent, Poggio, Tomaso A. On Edge Detection[J]. Pattern Analysis and Machine Intelligence, IEEE Transactions on,1986,8 (2):147-163.
[46]Qing Liu, Cheng-yu Lai. Edge detection based on mathematical morphology theory[C]. Image Analysis and Signal Processing, International Conference on,2011: 151-154.
[47]Rafael C Gonzalez数字图像处理(MATLAB版)[M].阮秋琦.北京：电子工业出版社,2005.
[48]Hanselman D, Littlefield B R. Mastering MATLAB 6:A Comprehensive Tutorial and Reference[M]. Prentice Hall, Upper Saddle River,2001.
[2]丁献文,万荣胜,王建,等.航海雷达图像噪声抑制方法研究[J].海洋技术,2011,30(3)：13-16.
[3]沈继红,李英,戴运桃,等.X波段雷达图像同频干扰的抑制方法研究[J].仪器仪表学报,2011,32(5)：1089-1094.
[4]高成志.基于X波段导航雷达海面回波图像的去噪和插值修复研究[D].天津：天津大学,2010.
[5]李鸿杰.XX雷达杂波抑制方法研究[D].南京：南京航空航天大学,2010.
[6]宋占杰,孙皓,王鑫,等.基于导航雷达海面回波图像去噪研究[J].海洋技术,2010,29(])：82-86.
[7]杨娜.基于小波变换的雷达图像处理[D].大连：大连海事大学,2009.
[8]郝燕玲,唐艳红,卢志忠.X波段航海雷达图像噪声检测与滤除方法研究[J].国土资源遥感,2008,2(76)：14-17.
[9]黄琼丹.基于雷达视频的目标检测和录取方法研究[D].西安：西安电子科技大学,2006.
[10]田守东.基于雷达图像的目标检测技术研究[D].哈尔滨：哈尔滨工程大学,2010.
[11]白素萍,张子鹤.基于数字图像处理的雷达图像动目标点迹提取和跟踪[J].舰船电子工程,1999,109：14-19.
[12]焦凤萍.合成孔径雷达图像的预处理方法研究[D].安徽：安徽大学,2007.
[13]Briggs J N. Target detection by marine radar[J]. Aerospace and Electronic Systems Magazine,2005,20 (6):39-40.
[14]Kabakchiev C, Garvanov I, Behar V. CFAR detection and parameter estimation of moving marine targets using forward scatter radar[J]. IEEE Conferences, Radar Symposium (IRS), Proceedings International,2011:85-90.
[15]Panagopoulos S, Soraghan J J. Small-target detection in sea clutter[J]. Geoscience and Remote Sensing, IEEE Transactions on,2004,42 (7):1355-1361.
[16]刘斌,杨劲松,范开国,等.基于船载雷达图像的海上船只检测方法[J].海洋学研究,2009,27(4)：33-37.
[17]种劲松.合成孔径雷达图像舰船目标检测算法与应用研究[D].北京：中国科学院研究生院,2002.
[18]种劲松,朱慧敏.高分辨率合成孔径雷达图像舰船目标检测方法[J].测试技术学报2003,17(1)：15-18.
[19]陈珊.合成孔径雷达图像上舰船目标的检测[D].上海：上海交通大学,2008.
[20]叶海军.基于纹理特征和数学形态学的SAR图像目标检测方法[J].中国电子科学研究院学报,2009,4(4)：436-440.
[21]陈晓楠,索继东,柳晓鸣.基于改进小波阈值函数的船用雷达回波去噪[J].沈阳工业大学学报,2010,32(2)：196-199.
[22]方莉,张萍.经典图像去噪算法研究综述[J].工业控制计算机,2010,23(11)：73-74.
[23]雷浩鹏.数字图像去噪算法研究及应用[D].长沙：长沙理工大学,2010.
[24]张黎,王立克,杨峰,等.小波阈值图像去噪研究与应用[J].图像处理,2006,10(3)：293-295.
[25]Donoho D L. De-noising by soft-thresholding [J]. IEEE Trans, on IT,1995,41(3):612-627.
[26]Donoho D L. Adapting to unknown smoothness via wavelet shrinkage [J]. Journal of American Statistical Association,1995,12(90):1200-1224.
[27]Wan Jun, Zhang Xiaohui, Rao Jionghui. Research and Application of Denoising Method Based on Wavelet Threshold[C]. Information Engineering and Computer Science (1CIECS),2nd International Conference on,2010:1-4.
[28]Qingwu Li, Chunyuan He. Application of Wavelet Threshold to Image De-noising[C]. Innovative Computing, Information and Control, First International Conference on,2006 (2):693-696.
[29]冯文强.非局部算法在图像去噪中的应用[D].安徽：中国科学技术大学,2011.
[30]韩思奇,王蕾.图像分割的阈值法综述[J].系统工程与电子技术,2002,24(6)：91-94.
[31]Otsu N. A Threshold Selection Method from Gray-Level Histograms[J]. IEEE Trans. Systems Man and Cybernetics,1979,9(1):62-66.
[32]Sapna Varshney S, Rajpal N, Purwar R. Comparative study of image segmentation techniques and object matching using segmentation[C]. Methods and Models in Computer Science, Proceeding of International Conference on,2009:1-6.
[33]Zhen Wang, Meng Yang. A fast clustering algorithm in image segmentation[C]. Computer Engineering and Technology,2nd International Conference on,2010 (6):592-594.
[34]吴一全,朱兆达.图像处理中闽值选取方法30年的进展(一、二)[J].数据采集与处理,1993,8(3)：193-201.
[35]Kittler J, Illingworth J. Minimum error thresholding[J]. Pattern Recognition,1986,19 (1):41-47.
[36]Veltkamp R C. Shape matching: similarity measures and algorithms[C]. Shape Modeling and Applications, International Conference on,2001:188-197.
[37]刘兴超.单帧红外图像弱小目标检测算法研究[D].武汉：华中科技大学,2005.
[38]李琦,傅俊诚,李自勤,等.激光雷达含噪图像边缘检测算法比较[J].2003,32(3)：239-243.
[39]连洁,韩传久,潘路.基于Canny算法的红外小目标边缘检测方法[J].微计算机信息,2007,23(6-3)：308-310.
[40]John Canny. A Computational Approach for Edge Detection[J]. IEEE Trans. Pattern Anal. Machine Intell,1986,8 (6):679-698.
[41]Vincent L. Morphological Grayscale Reconstruction in Image Analysis:Application and Efficient Algorithms[J]. IEEE Trans. On Image Processing,1993,2(2):176-201.
[42]DongPing Ming, Qun Wang, Jiancheng Luo. Evaluation of High Spatial Resolution Remote Sensing Image Segmentation Algorithms[C]. Image and Signal Processing,,2nd International Congress on,2009:1-5.
[43]Kapur J N, Sahoo P K. A New Method for Gray-level picture Thresholding Using the Entropy of the Histogram[J]. Computer Vision Graphics and Image Processing,1985, 29(2):273-285.
[44]Zhao Chunjiang, Deng Yong. A Modified Sobel Edge Detection Using Dempster-Shafer Theory[C]. Image and Signal Processing,2nd International Congress on,2009:1-4.
[45]Torre Vincent, Poggio, Tomaso A. On Edge Detection[J]. Pattern Analysis and Machine Intelligence, IEEE Transactions on,1986,8 (2):147-163.
[46]Qing Liu, Cheng-yu Lai. Edge detection based on mathematical morphology theory[C]. Image Analysis and Signal Processing, International Conference on,2011: 151-154.
[47]Rafael C Gonzalez数字图像处理(MATLAB版)[M].阮秋琦.北京：电子工业出版社,2005.
[48]Hanselman D, Littlefield B R. Mastering MATLAB 6:A Comprehensive Tutorial and Reference[M]. Prentice Hall, Upper Saddle River,2001.