Research On Vehicle Running State Estimation Based On Machine Vision

With the popularity of high-performance camera equipment and the development of computer and pattern recognition technology,vehicle detection,vehicle tracking,vehicle positioning,trajectory analysis and other technologies based on computer vision have become a research hotspot.These technologies have wide application prospects in the field of transportation,such as vehicle trajectory prediction,auxiliary driving system,automatic driving system and so on.At present,although people have carried out some research on vehicle detection technology,but the existing algorithm is still very immature,whether in terms of accuracy or calculation speed,it can not meet the needs of the actual system.In the face of this situation,this paper uses the methods of image processing and pattern recognition to study the problems of vehicle detection,speed measurement and ranging in video images.The main work is as follows:1.A feature extraction algorithm is proposed,which uses the combination of hog feature and SIFT feature as the feature extraction algorithm of vehicle detection.Then the positive and negative samples are trained by support vector machine(SVM),and the training effect is verified under different kernel functions.Finally,the training model is applied to the video files of different environments to test the effect.This method improves the efficiency of traditional vehicle detection,the vehicle detection rate reaches98.38%,but there is also a large rate of false detection under night vision,the actual effect still needs to be improved.2.A monocular vision ranging model is proposed,which uses the camera height,pitch angle and direction angle as error compensation.The distance information of the vehicle in front is obtained and the vehicle speed in front is calculated by the geometric projection principle.The final experimental results show that the method can detect the distance and speed of the vehicle in front in real time,which provides an effective guarantee for ensuring the safe distance and speed of the vehicle in the process of driving and preventing the rear end of the vehicle.3.The method of lane change recognition based on video.Firstly,the lane line is extracted by Hough transform,and the position of the vehicle in front of the frame image is predicted by Kalman filter.Considering that the problem of lane change is mainly based on the lateral movement of the vehicle,combined with the vehicle speed change and other factors in the process of lane change,when the vehicle moves laterally in the case of continuous multi frameimages,the distance between the vehicle and the lane line changes,so as to predict and analyze the possibility of left and right lane change of the vehicle.The experimental results show that the lane change recognition method can help the driver to achieve assisted driving,and it is an effective lane change recognition method for vehicles ahead.