Moving Target Detection Based On Saliency In The Surveillance Video

The real-time enhancement of moving target detection module plays a key role in the intelligent monitoring of security system.Considering the common points of the color feature and luminance feature in the static target saliency detection based on the visual attention mechanism and the dynamic target saliency detection in time-domain moving target detection,how to effectively use the common point and visual significance for moving target detection has become a necessary research hotspot and difficulty.This paper through in-depth study on the static saliency algorithm based on visual attention mechanism and the time domain algorithm for moving target detection,the moving target in the surveillance video as the object,the initial motion target is detected by the improved frequency-tuned(FT)algorithm at first.Then,the GMM algorithm is used to further detect and complete the whole moving target detection system.In this paper,the method of moving target detection ensures that the detection efficiency is not sacrificed and the detection real-time performance is improved effectively.In this paper,the main work is illustrated as follows:Establishing relation between frequency domain and time domain algorithm in the algorithm of moving target detection in this paper and the frequency-time combination point cut was found.First,the color features,the luminance features and the motion information of moving targets were analyzed.Secondly,the algorithm of saliency detection in frequency domain based on color features and luminance features and moving target detection algorithm based on motion information in time domain were further studied.The GMM algorithm for moving target detection in time domain has the problem of low detection efficiency and computational complexity,a method of frequency time saliency target detection algorithm combined with improved FT and GMM was proposed.In order to get more significant targets FT algorithm was improved from the two aspects of distance measurement and fusion mode.The improved FT algorithm was used to detect the initial moving targets,and then the moving target was further detected by the time domain GMM algorithm,which improved the real-time performance of the moving target detection module effectively.Two surveillance videos are taken to experience.Considering from the frequency domain and the time domain two angles,compare the algorithm in this paper with the GMM algorithm,the SR+GMM algorithm and the improved FT with based block matching optical flow method were compared and analyzed.Algorithm substitution experiment with two angles,both in effect maps and quantitative performance analysis,it can be verified that this algorithm can better ensure the accuracy of detection results and improve the speed of moving target detection.