Research On Moving Object Detection And Tracking Methods In Illumination Changing Environment

Moving object detection and tracking algorithm has become one of the important research projects in the current computer vision and digital image processing fields,for its widely applications in the fields of military guidance,safety monitoring,intelligent traffic and others.With the improvement of people's economic conditions and safety awareness,intelligent surveillance systems have been greatly developed.In the development of intelligent surveillance systems,surveillance cameras have spread throughout the urban and rural areas.However,problems such as the dramatic increase in the number of video caused by a complex surveillance camera system have made it a key issue for how to quickly obtain information through intelligent analysis in massive data.Because the complex environment the camera placed in such as bad weather,illumination changes and shadows of objects in the background cause the image quality to decline,leading to difficulties in establishing background models and segmenting foreground objects.The long-term stable moving object tracking becomes an urgent problem to be solved.On the basis of analyzing and summarizing the existing common algorithms,this paper focuses on the image illumination normalization algorithm under the changing illumination environment for the complex environment of illumination changes.And increase the robustness of the illumination by improving moving object detection algorithms and moving object tracking algorithms.The main research contents of this paper are as follows:1.An illumination preprocessing algorithm based on Retinex theory is proposed aiming at the illumination nonuniform image.Firstly,the essence is to predict the incident component of the image.This algorithm is different from traditional Retinex algorithm in complex operation prediction because of its fast prediction;then estimate the illumination component of the image by using bilateral filters and the input image to perform convolution operations;finally,the reflected component is normalized to obtain an enhanced image.2.Firstly,this paper briefly introduces the classic moving object detection technology.The principle and process of classical Vibe algorithm and the XCS-LBP(eXtended Center-Symmetric Local Binary Pattern)texture operator are introduced.Aiming at the incompleteness and inaccuracy of the foreground object obtained by the moving object detection algorithm under the complicated conditions of uneven illumination,this paper proposes a method of merging RGB color information and XCS-LBP texture information with confidence.The experimental results show that the fusion method proposed in this paper can effectively suppress shadows,has good robustness to indoor illumination changes,and is effective in dealing with complex backgrounds such as uneven illumination.3.Aiming at the tracking of moving objects in complex illumination scenes,an algorithm based on particle filter moving target tracking is proposed in this paper.The algorithm integrates SURF(Speeded Up Robust Features)feature points and color features under the framework of particle filter moving target tracking algorithm.This method can significantly enhance the ability of particle filter tracking algorithm to cope with illumination changes.A shadow removal method based on the HSV color space further improves the robustness of the moving objects tracking in the condition of illumination changes.And the particle filter algorithm based on multi feature fusion has obvious advantages in accuracy and robustness under the background of changing illumination.