Moving Target Detection And Tracking Based On Image Sequence

Moving target detection and tracking is an important research direction in the field of vision, it combines image processing, pattern recognition, deep learning, artificial intelligence and other related knowledge, It is widely used in video surveillance, intelligent robot, industrial testing and other fields, and has great prospects for development. Moving target detection and tracking refers to the moving object in image sequences, and get back to target real-time location and tracking of moving targets. According to the relative movement between camera and whether the moving target can be divided into static under the background of moving object detection and tracking and dynamic under the background of moving object detection and tracking.This paper first introduces the basic knowledge of digital image processing, such as common color space model, mathematical morphology processing methods; Also analyzed the characteristics of the noise and some of the common sources of noise, and it gave the median filtering, gaussian filtering, wavelet denoising method.In terms of moving target detection, this paper introduces the background difference method, the interframe difference and optical flow method, three common moving target detection method, and gives them the applicable scenario and the main advantages and disadvantages. In view of the background difference method in the background model can't reflect the changes in the background of faults. This paper proposes a hybrid gaussian model and the background difference method combined with the background of moving object detection method, with multiple gaussian background model describes the change of the background, as the input image sequence dynamic update background model, and then with the current frame difference to detect the target.In terms of movement target tracking, first of all, this paper introduces several kinds of commonly used moving object tracking method, including the tracking method based on contour, tracking method based on the model, the tracking method based on feature points and tracking method based on region. MeanShift algorithm and Camshift algorithm presented in this paper, aiming at the shortcomings of the Camshift algorithm,.This paper puts forward a kind of Kalman filtering with Camshift algorithm tracking method of linear fitting, joined in the process of target tracking Kalman filter prediction of target location, other objects in the context of object is sheltered can accurately track to the target object.Based on the previous methods, this paper in Visual Studio 2010 integrated development platform using opencv vision library and c++ programming language developed based on the movement of the camera target tracking system.The system has a target detection, target tracking. Save the image, and other functions. The experimental results show that the system can better realize the static background tracking a moving target.