Improvement And Prototype Design Of Road Moving Target Detection And Tracking Algorithm

In the intelligent transportation system,the detection and tracking of road moving obstacles has always been a key research content.As the motion of the on-board camera in the driving process leads to the background movement,the moving obstacles and the background move independently in the video,which leads to the difficulty in target detection and follow-up tracking.The main work of this thesis is as follows:(1)For mobile object detection,the ROI of interest is selected for the video image frame,and the median filtering,grayscale and other operations are carried out for denoising and enhancement of the ROI region;Then the feature points were extracted,the feature points were calculated by the hierarchical feature corner detection algorithm,and then the optical flow error was calculated by the improved pyramid LK optical flow method for the interested feature points,and the moving optical flow field was established by matching and tracking between the frames.(2)Based on optical flow field,an improved DBSCAN clustering segmentation algorithm based on optical flow point features is proposed in this thesis,which is used to segment regional targets in optical flow field,and the optical loss figure is obtained.The similarity measurement criterion is improved by combining the characteristic attributes of the optical flow vector points,and the similarity of the optical flow velocity vector and the coordinate density distribution of the optical flow points are included in the similarity measurement function.This improved measurement can be used to determine whether the optical flow point is in a target region,which can obtain a better segmentation effect,achieve the target segmentation of the optical flow field,locate the moving target region of the current frame,and initially realize the detection of multiple targets.(3)Design and improvement of road moving target detection and tracking algorithm.The main results are as follows: 1)A block-adaptive threshold algorithm is proposed to improve the accuracy of threshold segmentation by block-segmenting images,and the threshold segmentation graph is obtained.2)in this thesis,a moving object detection based on movement three figure fusion process,combined with the Canny edge detection to obtain segmentation image,the light loss of figure and partitioned threshold figure,figure by pressing the pixel edge fusion operation,with multiple morphological processing,region growing method to obtain accurate image binarization,improve under moving vehicle camera for the moving target detection accuracy.3)On the follow-up frame motion tracking problem,the Kalman motion model of the system is built by combining the Cam Shift algorithm,and the dynamic equation and observation equation of the target are established.The Kalman filter is used to predict the position of each target in the next frame through continuous iteration,reducing the number of iterations of the Cam Shift algorithm and increasing the tracking stability.4)Design the road moving target detection and tracking system,and complete the prototype design and implementation of the system.Experimental results show that the algorithm can detect moving targets in the process of vehicle running,and the segmentation effect is good.At the same time,the tracking algorithm also has a good effect,and has a good tracking accuracy on complex roads.