Ship Identification Tracking Technology Based On A Sequence Of Images Of Water Sports

Video image target detection and tracking is one of the core technologies in computer vision, which is widely used in many intelligent control industries. The detection, identification and tracking technologies of moving ships on water have important uses in terms of maritime security, management and scheduling. Therefore realizing detection, identification and tracking technologies of moving ships have great significance in water surface vessels’information statistics and preventing ship collision.This paper mainly researches maritime ship identification and tracking technologies. Firstly, it surveys the current development status of maritime moving target detection and tracking, and the existing image processing technologies are briefed; Then, the gray model based on interval distribution is employed to detect targets, and the HU invariant moments are adopted as image feature for extraction, and a SVM classifier as the recognizing approach; Finally, the meanshift and particle filter tracking algorithm are selected for moving ship tracking, and the effectiveness of these proposed methods is verified in the end. The main research contents include the following:1. After a study of existing visual interval models, a method based on the gray space model is used to detect ship targets in image sequences. Significant interesting objects in the moving target are extracted rapidly through the threshold segmentation and background disturbance removal, morphological denoising and other image processing operations, which lays a good foundation for the ship targets recognition.2. An in-depth study is carried out on pattern recognition of target shape features on the basis of target detection, Then use the invariant moment to describe the characteristics of target in the case of the template matching being not ideal. The study also takes the normalized characteristics as template information database and trained target template features. Finally, based on the SVM,the identify characteristics are classified and then the target type of ships is output. With the above procedures, precise identification of ships is realized.3. In-depth study of MeanShift method and particle filter method, which are applied to the tracking system. Since the traditional Kalman filter can’t effectively track the moving targets in complex scenarios, this paper uses the MeanShift method and particle filtering method for realization of ship target tracking under cross-blocked, and analyzes the advantage of two methods in the tracking error and the processing time. As a result, the proposed method achieves a relative efficacy, and meanwhile improves the system’s real-time and accuracy.