Research On Algorithm Of Object Tracing And Detecting For Video Surveillance System

The main research of this thesis was based on object detecting for indoor video surveillance system. The application of intelligent video surveillance system has been becoming comprehensive and getting much attention in the worldwide. As technology progresses, intelligent video surveillance system develops in the direction of more intelligentized. Targets tracking and detecting is considered as the most important problem in the process, it is the basic of the latter intelligentized processing.This thesis researches on the targets tracking and detecting technology based on the indoor relative fixed background, then put forward a more effective method and prove its availability and veracity by experiment.First, introduce the developing status and background, and analyze the correlative research methods.Then compare the kinds of related technique to object detecting and distilling with each other, by reading various research papers and related books. Because the algorithm of this paper is mainly based on the relative fixed indoor background, putting forward a Gauss background modeling algorithm to reconstructing the background. The algorithm is complex on math, but it do better on distilling background. It could get very clear background, remove and restrain the disturbance of foreground. Because of using OpenCV, it can implement on PC easily.At last, in the process of object detecting, aim at the bug of object template area wrap necessarily of the classical MeanShift algorithm, integrate it with Kalman filter, this paper put forward a algorithm based on the MeanShift and Kalman. Tracing with using the forecasting object position of Kalman filter as the initial searching focus of MeanShift algorithm, then get the new object position by MeanShift algorithm, it can overcome the long time that Kalman filter estimates the new position of the target. In a large degree, it makes the algorithm be more real-time.