Research On Moving Target Detection Technology Based On Time And Space Saliency Features Vision

The complete detection of moving objects from video sequences has become a popular research direction in computer vision,which is the basis of intelligent surveillance video analysis.In order to optimize the detection effect of the moving target detection algorithm,considering that the visual saliency performance is enough to quickly point attention to the interested object.A moving object detection algorithm based on spatiotemporal salient features is proposed,which can accurately segment the moving target.The main work and results of this thesis are as follows:1.This paper analyzes three typical moving target detection algorithms: optical flow method,inter-frame difference method and background subtraction method.In the time domain,the optical flow method is used to extract the moving information of the moving target.The FT visual saliency model combined in the spatial domain.The images were analyzed in the Lab and RGB color space to extract the significant features of moving objects in the spatial domain,which laid the foundation for image fusion.2.Fusion the salient features of the extracted moving targets in time and space.Wavelet transform fusion,pulse-coupled neural network fusion,and adaptive image fusion are analyzed.Combined with practical application,an adaptive spatiotemporal feature image fusion method is proposed.It can adaptively adjust the weights of the time-domain saliency map and the air-space saliency map according to the severity of the moving target.Not only is the time-domain saliency feature easy to attract attention,but it can also take into account the spatially significant features according to the human visual attention mechanism.3.The moving object detection algorithm is optimized.In this paper,an improved CV model level set is used to segment the spatiotemporal saliency feature map obtained after moving target detection processing to obtain the moving target area.For the case with shadow,the shadow is removed by means of multi-feature fusion,and the target voids and redundant parts are morphologically manipulated to make the extracted moving target complete.4.Evaluate the feasibility of the algorithm in this paper.Select multiple samples in the video set.For different moving targets(size,number and speed),complex backgrounds,and different light intensity,use the moving target algorithm in this paper and the current mainstream moving target detection algorithms.Moving target detection,and evaluate the feasibility of the algorithm in this paper from both subjective and objective aspects.