Video Moving Target Detection Algorithm In The Complex Context Of Research

Video moving object detection technique is one of the most important fields in digital video processing and application, and especially it has been widely used in military field such as weapon control and guiding system. With the rapid development of the electronic technology, the working speed of the digital signal processing device is becoming faster and faster, and which enables the object tracing in real time. At present, the object tracking in real time becomes a hot topic in all over the world.In this thesis, the author makes the design and implementation of the system and algorithms under the DM642 hardware development kit and CCS software integration development environment. This thesis focuses on (1) image sequence pre-processing algorithm which can improve the image quality; and (2) moving target detection under complex background caused by camera moving.It is necessary to pre-process images since it may cause image blurring or dilution when a moving camera takes images. After summarizing the advantages and disadvantages of some algorithms, an improved histogram equalization algorithm is put forward. Compared with the traditional algorithms, the new one can not only enhance the contrast of the image but also can make the details much more clearly, especially useful to low contrast images. Moreover, a large number of tests showed that the new algorithm has wide application prospects.When the detection system is tracking the moving objects under the servo control, the camera is also moving, which results in the background movement. In order to segment the moving objects from moving video background, it is necessary to compensate the moving background. Therefore, the thesis introduces the key technique of the global motion compensation in details and puts forward an object detection algorithm based on the global motion estimation. In this algorithm, the affine parameter model is taken to estimate the moving parameters of the background. As for the choosing of the feature points, the algorithm that model cooperated with the Mao operator is advanced. Experimental results indicated that the studied algorithm can accurately estimate the background motion caused by the camera moving, consequently, detect the motion objective accurately.