Research And Application Of Video Foreground Detection Technology With Complex Scenes

In intelligent video surveillance,the purpose of foreground object detection is to extract interested foreground objects and use them in the middle and high level video analysis technology.Therefore,video foreground detection has important research value and is also the core content of intelligent video surveillance.In view of the unsolved or partially solved challenges of foreground detection,the shortcomings of current video foreground detection methods are analysed,effective video foreground detection algorithms are proposed,and a video foreground detection system is designed and implemented.The main works of this thesis are as follows:1.A modified SuBSENSE algorithm via adaptive distance threshold based on background complexity is proposed.In order to solve the problem of missing and wrong foreground detection caused by inaccurate initialization model and slow updating of distance threshold in complex dynamic scene,an improved initialization method and self-adaptive distance threshold modification policy based on background complexity are proposed.In the initial stage of the algorithm,the background model is created by the improved three-frame difference method to improve the authenticity of the model.In the model updating stage,we measure the background complexity via temporal and spatial consistency,and the more accurate distance threshold is obtained for pixel classification by combining the distance threshold modified with the defined background complexity.The experiment results show that the overall detection effects of the algorithm have been improved to a certain extent compared with the original algorithm.2.An algorithm of foreground detection based on fusing confidence by weight and visual attention mechanism is proposed.In order to solve the problem of color camouflage and still foreground,a new algorithm based on fusing confidence by weight and visual attention mechanism is proposed.Firstly,the moving foreground is determined by the sum of fusing with color confidence and texture confidence based on weight.Secondly,the confidence and weight of the two dimensions are updated adaptively.Then,the visual attention mechanism is constructed by combining the saliency and texture difference to judge and detect the still foreground target.Finally,the fusion of the detected still foreground and the moving foreground target is used to do the final detection result.The experiment results show that the algorithm can effectively deal with the challenges brought by color camouflage and static foreground objects,so as to improve the foreground detection effects.3.A video foreground detection system is designed and implemented.Based on the above research content and the proposed algorithm as the core function module,the video foreground detection prototype system is designed and implemented,and the data storage function was added to improve the system.Finally,the effectiveness and application value of the system are guaranteed through the system experiment.