The Research Of Object Matching In Intelligent Surveillance System

With the rapid development of computer hardware and Artificial Intelligence technology, many fields focus on the technology of the intelligent computer. As one of its branches, Intelligent Surveillance System has been widely used in the fields such like production, security, police. As the one of core technologies in Intelligence Surveillance System, object matching technology plays a key role in Intelligent Surveillance System. The practical application, which is diverse and complex, more and more puts forward higher request to the matching algorithm. The algorithm needs to be reliable to image scale,rotation, change in illumination and cost little time. This paper focuses on the most popular object matching algorithm, analyses its disadvantage and improves it to the Intelligent Surveillance System.Firstly, we study the core technologies of the Intelligent Surveillance System including object detection, object tracking, object matching. And make a research in object matching, focus on SIFT algorithm, which has the most widely use among all object matching algorithm. And combine it with the classic Harris Corner Detector algorithm, for secondary filtering of the interest points, as a result, we get more accurate Harris-SIFT feature, and increase the rate of object matching.Secondly, we point out the fail of SIFT algorithm when the object has a flip transformation, figure out reasons for this phenomenon and propose an improve algorithm based on the difference of the modulus values, which is invariant to flip. The improved SIFT algorithm(F-SIFT) has a high quality in object flip scene.