Research On Moving Object Detection And Tracking Algorithms

Moving object detection and tracking is one of the hot subjects in computer vision, which combines the knowledge of many fields, such as image processing, pattern recognition, artificial intelligence, automatic control, etal. It is an interdisciplinary, challenging and forefront subject. It has broadly applied in intelligent visual surveillance, intelligent transportation, human-computer interaction, video compression, medical image analysis, robot vision and navigation, etal. Therefore, this subject has important theoretical significance, practical value and broad prospects for development.This paper mainly studies the algorithms of moving object detection and tracking. In this paper, moving object detection and tracking algorithm are improved and achieved based on the results of previous people'study. Main tasks are as follows:Firstly, Moving object detection method based on Gaussian mixture model which is current commonly used is studied and on this basis, an improved method is proposed. The improved method divided the pixels of image into single model and multi-model point. Single model point is described with a single Gaussian model. Multi-model point is described with multiple Gaussian models. Thereby, the complexity of background modeling process and the process of moving object detection of the original Gaussian mixture model algorithm is reduced and the speed of moving object detection is increased, but the effect of moving object detection does not decrease.Secondly, A moving object detection method based on statistical histogram and single gaussian model is proposed. This method overcomes the disadvantage of Gaussian mixture model and the moving object detection method based on color deviation can not detect moving object completely when the color of moving object is similar to its corresponding background.Finally, Multiple vehicle objects's tracking of traffic scenario is achieved by Kalman filtering. A fast and efficient vehicle matching method in the adjacent frame based on the prediction model of Kalman is proposed. For the problem of multiple vehicles's touching in complex traffic system, a method of segmentating touching vehicles is proposed. When multiple vehicle objects touch each other, they are separated to individual vehicle objects through the segmentation algorithm, Tracking the vehicle objects which touch each other individually is achieved.