Man-made Targets Detection In Large Scene PolSAR Image

Man-Made targets detection plays an important role in applications of remote sensing and it is the basis of disaster rescue and military reconnaissance.The detection performance and processing speed will affect the practical application directly.Polarimetric SAR is an active Radar system,the all-day and all-weather characteristic make it has good performance in the applications of target detection which both accuracy and processing speed are taken into consideration as two important indexes.The ability of PolSAR imaging has been greatly improved and there are more and more PolSAR images with large scene and big data.This provides the data base for PolSAR image targets detection applications.However,most traditional research of target detection pay more attention on the index of accuracy,so that the processing time of traditional algorithms will be increase with the increasing data.In most target detection applications,processing time is an important index and this is a challenge of traditional target detection algorithms.Complete the detection task with a high speed in the requirement of accuracy is the prime problem to solve.In this paper,man-made targets detection in large scene PolSAR image is considered to be the main research task.First,the analysis of PolSAR image representation and man-made target characteristics are taken into consideration and then,the characteristics of man-made targets in PolSAR image are clarified.The basic target detection theory is combined to construct the foundation of follow research.Mathematical tools are used to improve the processing efficiency of traditional algorithms and the fast-Wishart based detection algorithm is designed.Then,the image uniformity description factor,IUDF,is designed in combination with image analysis.IUDF has the ability of target areas extraction and the processing of IUDF extraction is much faster than traditional texture feature.IUDF is used to design three different target detection algorithm with different image information usage and the indexes of accuracy and processing time have a wide range which satisfy most requirements of practical applications.Experiments of man-made targets detection in land environment and ocean ship detection are designed to test the performance of d man-made targets detection algorithms in large scene PolSAR images and the accuracy,important targets detection and processing time are taken into consideration as three indexes.Experimental results show that,the four detection algorithms designed in this paper have the ability of man-made targets detection in large scene PolSAR image.The performance of fast-Wishart based detection algorithm is related to the selection of supervisory information and there is great potential of performance improving.Three IUDF based detection algorithms complete the detection task in a high processing speed with a higher accuracy than traditional algorithms.Different image information usage make the three algorithms a wide range of the indexes of accuracy and processing speed and this will satisfy most requirements of practical applications.