Research On Moving Object Detection And Shadow Elimination In Complex Scene

As the key component of an intelligent video surveillance system,moving target detection of the video sequence is widely used in the car automatic driving and auxiliary driving,intelligent security,human tracking,video retrieval,missile warning and other occasions.The purpose of moving target detection is to accurately extract parameters of the interesting moving region in the video sequence.However,disturbance factors,such as light mutation and shadow,is existed in real scenes,which leads to high rate of missed detection and false detection by the current proposed detection methods.As a result,application of moving target detection is hampered.Therefore,it is of great theoretical and practical significance to study the methods of moving target detection and shadow elimination.The thesis deeply studied the moving object detection of complex scenes such as illumination changes,swing of the surrounding leaves,the strong lane in the traffic lane,mutual occlusion between multiple targets.Specific work can be summarized as follows.Firstly,on the basis of the existing detection methods,to cope with cavities and false targets under dynamic backgrounds with noise and light changes,a novel moving target detection method,which is based on seven frame difference and background edge detection,is proposed.To cope with poor real-time performance of traditional Gaussian mixture background model method and low detection accuracy of frame difference method under complex scenes,a new moving target detection method,which is based on seven frame difference and improved adaptive Gaussian mixture model,is provided.And comparative experimental results indicate that the proposed methods are effective.Secondly,the shadow elimination method is deeply explored.After mastering the existing shadow elimination methods,an effective moving target detection and shadow removal method,which is based on improved adaptive Gaussian mixture model and color space,is put forward to cope with low accuracy caused by moving shadow.Experiments are carried out to verify the feasibility of the method presented in the thesis.Finally,a moving target segmentation system is constructed.With the visual studio2010 and Qt creator,OpenCv is adopted to reprogram all the proposed methods,and an improved Camshift method is used to verify that the work done in the thesis is efficient.