Research Of Moving Objects Detection In Intelligent Video Surveiliance System

With the rapid development of information technology and more and more attention paid to public security, video surveillance system has been widely used, such as national defense and military field, traffic monitoring, important places with security sensitivity and common civil field. Because of the large-scale surveillance video, which is generated by the rapid development of video surveillance technology, is beyond the processing ability of human being, intelligence of surveillance system is becoming an inevitable trend. As the key step of intelligent surveillance system, moving objects detection is of important research significance and practical value. In view of the existing deficiency of moving objects detection algorithm, this paper has studied in the following aspects and some results are achieved. The major works include:(1) In view of the similarity of gray level between moving objects and background image, an object detection algorithm using pixel classification based on improved Gaussian mixture model is proposed. First of all, the improved Gaussian mixture model, which is able to overcome the influence of scenes change, is applied to complete background modeling. Then, the pixels of the difference image, which is obtained by subtracting the extracted background from the current frame, are classified by different thresholds. And Finally, the front image is obtained. Experimental results show that compared with traditional algorithm, the proposed one can achieve much better detection performance, while the robustness is guaranteed. It can increase the detection rates by about26.50%, decrease the false alarm rates by about2.756%, and decrease the time complexity by about27.13%.(2) In order to overcome the drawbacks of frames subtraction or background subtraction that are used separately in moving objects detection for traffic monitoring, and to achieve both accurate and robust detection results, a novel approach for moving objects detection based on frames subtraction using edge information and background subtraction is proposed. Firstly, a foreground image is obtained by three-frame-differencing based on edge information. Afterwards, background subtraction is used to obtain another foreground image, in which background image is created by improved mode method. Finally, the foreground object is extracted by applying Boolean OR operation on the two previous obtained foreground images. The simulation results show that the proposed method can increase the detection rates by about4.96%-36.01%, decrease the time complexity by about40%, and is more robust than other algorithms.(3) An algorithm of moving shadow detection and elimination is proposed and implemented in the video surveillance scene. Moving shadows in the video surveillance scene have important influence on the results of moving objects detection. And if not eliminated, their existing will have bad impact on the following processing. In order to improve the effectiveness and robustness of traditional shadow detection algorithms, an approach based on combined gray and texture features is proposed. Firstly, motion regions are extracted by background subtraction; Afterwards, fast normalized cross-correlation (FNCC) is used to obtain potential shadow regions from motion regions. Finally, the real shadow regions are obtained by doing texture analysis on the potential shadow regions, wherein Gabor Wavelet is used. The results show that the proposed method is more accurate and robust than other algorithms. The shadow detection rates and object detection rates is each increased by about5.19%and1.10%, and time complexity is decreased by about29.42%.