Machine Vision-based Front Cross Collision Detection And Safety Status Assessment

In order to avoid or reduce the occurrence of cross-collision accidents when the car is driving at an unsignalized intersection,this article uses a monocular camera to build a single-binocular vision conversion system to perform vehicle-side obstacle detection on the vehicle side based on machine vision.Forecast the movement status and driving trajectory of obstacle cars,obtain the intersection position of two vehicle trajectories through trajectory prediction,and further establish a cross conflict judgment model based on driving trajectory and collision time and a safe state assessment model based on safety distance to carry out obstacle car cross conflict Judgment and safety status assessment,and simulation verification.The main research work of this article is as follows:1.Detection and tracking of obstacles in front of the car based on the single-binocular vision conversion systemA single-binocular vision conversion strategy based on an improved artificial potential field is proposed,and a single-binocular vision conversion system is established to realize vehicle-side front environment perception based on machine vision.Obstacle car detection,obstacle car tracking,obstacle car ranging and speed measurement,etc.Collect real intersection driving video and use MATLAB software to verify the real-time and accuracy of the single-binocular vision conversion system.2.Obstacle car trajectory prediction based on extended Kalman filter algorithmBased on the extended Kalman filter algorithm,the vehicle motion state is estimated;on the basis of the motion state estimation,the obstacle trajectory is predicted.Using MATLAB and Car Sim software to verify the rationality of the state estimation and trajectory prediction results.3.Safety status assessment based on driving trajectory,collision time and safe distanceThrough trajectory prediction,determine trajectory intersections,establish a cross trajectory model,calculate the collision time,and determine the occurrence of cross conflicts;establish a safe distance model,compare the determined distance thresholds,and classify conflict levels.Pre Scan software is used to set different driving conditions for self-carrying and handicapped cars,and simulation verification of safety assessment is carried out.The results show that the monocular and binocular vision conversion systemconstructed in this paper achieves 95% recognition accuracy for obstacle cars,and the accuracy of ranging and speed measurement is 93% and 95%,respectively,and the measurement speed meets the needs of intersection driving;The prediction result of vehicle trajectory coincides with the actual trajectory up to 93%;the safe state assessment model of the collision domain and collision time is conducive to provide the vehicle with accurate driving status of the front side vehicle in time under the premise of reasonable threshold.The effective recognition rate exceeds 97%,and the accuracy rate of conflict level recognition exceeds 95%.