Moving Target Tracking Using Active Contour Model In Ship Navigation Video Surveillance

Moving target detection and tracking technique has long been regarded as an important research in computer vision and image processing field. In recent years, along with the rapid development of multimedia technology and advanced computer performances, the applications of the various kinds of video images are more and more extensive. The moving target detection and tracking techniques have also increasingly attracted a significant amount of attention and are extensively used. For example, in artificial intelligent system, the moving target is needed to be tracked and recognized and in the intelligent transportation system (ITS), it is necessary to monitor moving vehicles, etc.This paper considers ships detection and tracking method for channel safety under intelligent video surveillance system. The commonly used moving target tracking method can be divided into two categories: (1) the tracking method based on the characteristics of the target image region, (2) the tracking method based on the target contour features. Active contour model-based tracking method belongs to category (2), and so this tracking method is based on the target contour features. The advantage of using active contour model-based tracking is that it can reduce computational complexity.Active contour model, also know as snake model, is energy-minimizing deformable contour with continuous and closed path. The shape and the position of the curve are controlled by the internal and external energy of the image contour. Internal spline forces serve to impose a piecewise smoothness constraint, and the external constraint forces are responsible for putting the contour near the desired image features. After this model was first proposed by Kass in 1987, many researches had been made to improve this model. Moreover, it is applied in many computer vision research fields, for examples, static image boundary detection, image segmentation and moving target tracking in image sequence, etc.Since the initial positions of the active contours are mostly selected manually, its application in moving target tracking has many limitations. This paper researches the method for tracking moving ships based on video image sequence. Tracking of moving ships based on video image sequence needs overall consideration such as the effect of wave, the reflection of water surface etc. This paper, by combining the active contour model and the motion information of the target, put forwards a method in order to automatically choose the initial position of the contour based on binary image of the target, and enhances the practicability of the active contour model in moving target tracking, and then this active contour-based tracking method is used to track single and multiple moving targets.This dissertation is mainly organized as following chapters:First, in the introduction, the research background for moving target detection and tracking, the related research works and the existing problems for active contour model based moving target detection and tracking methods are presented. In chapter two, the introduction to active contour model, contour representation, energy function definition and energy minimization process using greedy algorithm, etc. are presented. Chapter three develops an algorithm for moving target detection and tracking in video image sequence. Since the technique research in this paper is moving target tracking using a static camera, moving target is detected by using background elimination method. First, set up the Gaussian mixture background model for each pixel point of image frames without moving objects, and then proceed the detection of moving target by using this background model. The speed of the moving ship can be estimated by using the displacement quantity of the detected target among the several frames of that target entering into the surveillance area. And then develops an algorithm to select the initial contour automatically based on the binary image of the detected target. After getting the contour of target, the points of the active contour are saved in image array in the form of Freeman chain code. After getting the initial contour of the target and the estimated velocity of the moving target, it can be proceed to tracking. In chapter four, the experimental results for tracking single and multiple moving targets by a single camera surveillance system are presented. Finally, some conclusions are formulated together with the summary of the contributions presented in the dissertation. Additionally, some remarks concerning future research are also given.