Research On Target Detection Algorithm For Intelligent Video Surveillance

With the developments of computer vision technology and network transmission technology,the intelligent video surveillance technology has gradually replaced the traditional manual video surveillance to complete the monitor of the specified scene,which involves a lot of computer vision technologies,such as image preprocessing,target detection,pattern recognition and image understanding,etc.The target detection is an important technology in the intelligent video surveillance.A quick and effective target detection algorithm is important for the subsequent modules of the video surveillance.However,the current commonly used methods of target detection have disadvantages in the aspects of the target detection accuracy,processing speed and memory usage.In order to solve the problems above,this paper sums up some different target detection algorithms.After the experimental results of different target detection algorithms are compared,Vi Be algorithm is selected as the main research algorithm of this paper.The improved Vi Be algorithm can be achieved to solve the ghost problem and shadow problem of the traditional Vi Be algorithm with small memory occupation and high detection speed.The improved Vi Be algorithm can improve the target detection accuracy.This paper mainly includes the following contents:1.The saliency detection,the morphological filtering operation and the color space are researched.Then,this paper studies the current mainstream moving target detection algorithms.The analysis of these target detection algorithms are compared to get the advantages and disadvantages by the experiments.2.The Vi Be detection algorithm is easy to produce ghost problem.This paper proposes an ghost elimination algorithm based on saliency detection and adaptive background update strategy.(1)The saliency detection preprocess of the first frame is done before the target detection to get the saliency region.After the target segmentation of foreground and background,the high accuracy of the background set is got.The accuracy rate of background model is improved.The background model is built by using the accurate background.The Vi Be algorithm is used to complete the subsequent target detection,which can eliminate the ghost area caused by the moving object in the first frame of the monitoring video.(2)The algorithm adds the criterion of ghost in the Vi Be algorithm’s random strategy.When an area meets the criterion,the update probability of the background will be adaptive increased.This will improve the ghost elimination rate.Then the regional background updating speed is improved,speed up eliminating ghosting speed.3.In order to solve the shadow problem in the Vi Be algorithm,this paper proposes a shadow removal algorithm based on mixed color space.By constructing the mixed color space,the shadow region and the moving target region are distinguished according to the change of the color feature in the same region when the shadow is covered or not.Finally,a hybrid color space Vi Be algorithm based on saliency detection and adaptive background update strategy is proposed.The experimental results show that the proposed algorithm can quickly complete elimination of moving target and ghost shadow.The improved Vi Be algorithm improves the target detection accuracy.It lays a good foundation for the subsequent module of the intelligent video monitoring system.