Research On Moving Targets Detection Based On Image Sonar

Accompany the development of marine resources and underwater detection technology,sonar detection devices become an important part of underwater detection by virtue of its unique advantage. Using sonar devices to detect and recognize underwater targets is an important direction of recent ocean research, image sonar using acoustical imaging technology transforms underwater moving targets' information into intuitive sequence of images making it an important basis of underwater moving target detection field. Because of the complex underwater acoustic environments, underwater acoustic images collected by image sonar usually have the problem of low signal-to-noise ratio, targets information incomplete, so the moving object detection on sonar image plays a very important role and has broad application prospects.Firstly, acoustical imaging algorithms are analyzed and simulated. Aiming at solving the problems of sonar images, two image processing techniques have been researched, and performance validation of different image pre-processing methods are discussed through simulation. Image segmentation is one of the most key techniques of image target detection,study of two common image segmentation(such as: fixed thresholding method and Otsu thresholding method)has been conducted based on single image detection. We find these two algorithms have poor detection performance through experiments. Therefore, CFAR(Constant False Alarm Rate, CFAR) detection method is studied. After a detailed research of CFAR's theories, models and detection performance,apply CFAR algorithm to the sonar images which conventional sonar image segmentation algorithms fail to detect in, and successfully detected these targets.Moving target detection in sonar images is based on sonar image sequences, so a moving target detection research on multiple-frame images is carried out. Research of frame difference method and background estimation method has been conducted, and analyzes their detection performance by simulating. Among them, background modeling is the key of background estimation method. The Gaussian background modeling method and mixture Gaussian background modeling method are analyzed. Because underwater environment is complex, the confidence interval of background by mixture Gaussian background modeling is established, the culling of static goals and restraining of background noise is completed, and detect the targets end with CFAR detection algorithm. The processing result of simulation and pool experiments show that this underwater moving target detection algorithm based on sonar image has effective detection capability.