Research On Moving Object Detection And Tracking Algorithms Based On Video Image

Moving object detection and tracking in video image is one of the important research topics in the field of computer vision. It has important theoretical and practical significance.This paper analyzes the development of moving object detection and tracking technology at home and abroad, then we propose three new algorithms:According to the problems of large amount of computation, the slow moving. dynamic interference and the same gray, this paper presents an algorithm on the basis of an improved interframe differential algorithm and an improved Gaussian model. Firstly, according to a statistical histogram, an interesting region is extracted. Through a mean algorithm, an initial background model is established. The interesting region is divided into several blocks by a self-adaptive method. Secondly, according to an improved interframe difference algorithm, the interesting region is separated roughly. On the basis of these steps, we utilize an improved Gaussian model to separate the rough results precisely. At last, the results are processed by double-threshold background subtracting. Experimental results show this algorithm can detect moving vehicles rapidly and accurately.Aiming at the problems of efficiency, this paper presents moving objects detection based on improved Hough forest. This algorithm is divided into some steps. Firstly, the gradient and color were extracted and use these information to describe the objects during feature extraction. Secondly, we adopt the improved Hough forest to train the positive and negative samples and then we adopt voting way was improved by a weighted region based on Gaussian model. Experimental results show this algorithm can detect objects accurately at the same time.Object tracking under complex circumstances is a challenging task because of interference of background, obstacle occlusion, object deformation and so on. Given such conditions, robustly detecting through single-feature or fixed weights representation is difficult. For these problems, object tracking based on a fragment and a multi-feature adaptive fusion is proposed. Firstly, construct the object and the background's feature distribution coefficient is a measure to distinguish the target and background. Secondly, this paper imports the concept of fragments and distinguish the different types of occlusions. At last, this algorithm adopts different features fusion for different occlusion types. Experimental results show the algorithm can be adapted to different occlusion types, multi-feature adaptive fusion avoids the problems of lost or drift under the conditions of interference of background and object deformation, and this average accuracy rate is more than 93%.