Road Traffic Moving Object Detection Based On ViBe Algorithm

Moving target detection technology is a research hotspot in the field of computer vision and an important technology in the field of artificial intelligence.In daily life,practical applications such as traffic flow monitoring,traffic light intelligent control and road foreign object intrusion all put forward high requirements for target detection algorithm.Based on the study of the theory of traditional moving target detection algorithm,the moving target detection in road traffic video is studied in this dissertation.Aiming at the problems of dynamic background error detection,noise interference and tracking target loss in the actual scene videos,an improved adaptive moving target detection method is proposed.Firstly,the video sequence is preprocessed.Single frame images are extracted according to the sequence of video sequences.Acquired image materials are often color images which have multi-color channels,and they need to be grayed to facilitate subsequent operations.In addition,for some images that are obviously affected by illumination,the CLAHE algorithm in the histogram equalization algorithm is used for adaptive illumination compensation to increase the image contrast,so as to highlight the target features.Secondly,the moving target detection window is set.In order to reduce the influence of irrelevant pixels in the video and reduce the amount of computer information calculation,the edge straight lines are extracted by improving Canny edge detection algorithm and Hough transform,and the image area is divided to set the region of interest for the road traffic video image.Aiming at the problems of interference and difficult edge extraction in the actual scene image,the hybrid filter of adaptive median filter and bilateral filter is selected for filtering.Four direction gradient template is employed for gradient detection,and the improved Otsu algorithm is adopted to calculate the segmentation threshold and complete the image edge extraction.On this basis,Hough transform is used to extract the longer line segments in the image,and set the region boundary and the detection window.Then,the moving object in the video frame is detected.An improved ViBe algorithm combining the knowledge of Hash algorithm,image saliency analysis,inter frame difference and two-dimensional entropy is proposed to replace ViBe algorithm which produces the ghost and makes poor performance in the static background.Hash algorithm is used to extract the key frame for differential operation.The differential image is combined with the FT significance detection image to extract the target area of the current frame,fill the background pixels and filter out the ghost area other than the real target.The complexity of the background is calculated by using the minimum distance of pixels and local standard deviation,and the radius threshold and update rate of pixels are adaptively adjusted to reduce the influence of background pixels.Through connected domain detection and morphological processing,isolated noise points are filtered out and target holes are reduced.By comparing the accuracy,recall and F-measure value of each algorithm,the effectiveness of the algorithm proposed in this dissertation is verified.Experiments show that this method can detect the moving target region well as well as reduce the influence of background pixels.Finally,the moving object in the video frame is tracked.In order to improve the accuracy of moving target tracking,the traditional Cam Shift algorithm is improved.The improved algorithm combines the Hash algorithm to judge the target tracking in real time,and uses the ViBe algorithm to generate the target image and initialize the search window when the target is lost,so as to improve the accuracy of the algorithm tracking.In addition,the Cam Shift algorithm uses the Kalman filter to predict the target position in the next frame,which reduces the number of iterations of the Cam Shift algorithm and enhances the tracking stability of the algorithm.Simulation results show that the improved method can track moving targets accurately.