Research On The Moving Target Detecting And Tracking Algorithms For Off-position Detection

As the "Safe city" project and "Smart city" project was proposed by the Ministry ofPublic Security, many cities started the corresponding work to construct the system aboutaccident warning, safety production monitoring and control systems, which greatly promotedthe intelligent security industry, especially the field of intelligent video surveillance. Theresearch on object detecting and tracking algorithms for off-position detection proposed inthis paper is one part of the intelligent video surveillance, the intelligence algorithm is mainlyused to detect whether the personnel on duty was off position in the video scene, thealgorithms include the detection algorithm and the tracking algorithm for the personnel onduty.Since the personnel on duty usually belong to the foreground object in the video scene,detection of personnel on duty can be translated into foreground object detection in the videoscene. This paper compares comprehensively several foreground object detection algorithms,selects a kind of algorithm named gaussian mixture model to constructing video backgroundmodel, and then subtract background model from the original image to get the foregroundobject in the video. Finally, the foreground object should be filtered to separate the dutyperson from the difference images, which is be completed by image segmentation algorithmsin the paper.According to the actual application in this paper, it is impossible to detect duty personframe by frame instead of tracking duty person. In this paper, we propose an improved objecttracing algorithm which is suitable for off-position detection in intelligent security filed. Thisalgorithm takes advantage of mean shift and particle filter, pre-traces the object by themean-shift algorithm and then calculates the accurate position by particle algorithmadvoiding from the loss of tracing accuracy and the large calcultion, which shortenscomputing time on the premise of insuring tracing accuracy.Finally, this paper designs off-position detection system, introduces the flow chart ofeach algorithm module in the system, then selects a video about surveillance environment,and dectect whether the person in the video is off-position. Last part of the paper summarizesthe shortcomings and improvements of off-position detection system.