A Study Of Multi-pedestrian Motion Detecting And Tracking Under Compressed Domain

Video encoded moving target detection is actually directed to the compression of moving target detection based domain. The current study focused on compressed domain split the two most widely used video coding standard: MPEG-2 and H.264. Most of the pedestrian detection and tracking algorithm technology called pixel domain, using the original pixel data from fundamental to ensure the accuracy of the trajectory of the object. However, they often lead to high computational complexity which is fatal drawback in large-scale monitoring system in real-time operation. In addition, most of the network transmission of video content is not original pixel data but sorting through compression coding such as H.264 bit stream, in order to improve communication efficiency.In contrast to this, in the compression algorithm information by encoding the domain may significantly reduce the computational complexity, such as motion vectors and DCT coefficients. In the compressed domain video object segmentation, compared with the pixel domain segmentation method is more suitable for the needs of practical applications. Given that most practical applications of video sequences have been compressed into a format directly in the compressed domain video object segmentation, may be waived completely decode compressed video; and, from the compressed video extracted by the entropy coding only motion vector ( motion vector) and DCT coefficients can be directly used as a movement and texture features object segmentation needs, which in the macroblock (macroblock) or block (block) level so that the data extracted feature quantity than the pixel domain processing required a lot less, thereby significantly reducing the amount of calculation segmentation algorithm. Therefore, from the compressed domain video object segmentation with fast characteristics suitable for applications with real-time requirements, solve pixel domain segmentation method is difficult to meet the requirements of real-time split, but want to achieve pixel-level segmentation accuracy, usually also the object The boundary blocks completely decoded edge thinning processing in the pixel region.Based on the latest compression standard H.264 encoding platform, the compressed domain moving target detection and tracking algorithm study. Choose two ways in which the study: To analyze H.264 compression standard can be used to find a new moving target detection parameters to detect by the quantization parameter QP moving target; another method by combining H.264 compression domain pixel domain to improve the detection accuracy, this method on the basis of simple yet improved detection accuracy. QP coefficient detected by using a moving target, and then the local decompression to improve detection accuracy by LBP pixel domain algorithm. The first method in the conventional compressed domain moving object detection method proposed a new information available,the second method ensures interference light, shadows and other cases on the basis of real-time algorithm to solve a certain extent, the accuracy ratio Traditional compressed domain detection greatly increased.