Analysis And Simulation Of Moving Video Target Trace Extraction Algorithm

Visual information, especially dynamic visual information, takes the most part of environment information that is perceived by human beings. The dynamic image of moving object offers more information than what single image can provide. Through analysis of many frames of image, we can gain information which we cannot get from single image. Image recognition and tracking of moving object is a process of detecting-recognition-tracking, which is based on dynamic image analysis and combine methods of image recognition and image tracking.The research in this dissertation is video sequence, i.e. image sequence. Object detection in videos involves verifying the presence of an object in image sequences and possibly locating it precisely. Object tracking is to monitor an object's spatial and temporal changes during a video sequence, including its presence, position, size, shape, etc. These two processes are closely related because tracking usually starts with detecting objects, while detecting an object repeatedly in subsequent image sequence is often necessary to verify tracking. Due to changes in illumination of the scene, background perturbations, shadows, vibration of the camera and occlusions between moving objects, precise detection and tracking of moving objects is still a challenge field of research.Based on the review and analysis of current methods of detecting and tracking moving objects, we lay emphasis on the research of moving objects detection and tracking in image sequence obtained by a stationary camera. Firstly, large amount of material was read to compare the difference among these methods. Thanks to this, the data structure and the algorithm were established.Based on the analysis of familiar methods of detecting moving objects, a method was present for detecting moving objects based on difference image. Detection of moving objects is done by four steps: difference frames, threshold self-adapting, morphological filtering, connected component labeling. The position coordinate of the target was calculated with the help of the centre determine method. Its moving orbit was shown. After we found that the computation was too complex in the OTSU method, we increase efficiency of it. At the same time, we researched on the predicting of the target when it is sheltered from our sight.This dissertation also makes simulation and programming on kinds of algorithms, and optimizes the last one, which make the programming compact and reliable. It also enhances the stability greatly. The experimental results indicate the algorithms operate quickly, accurately and simple to be realized through hardware. It achieves the prospective goal.