Research On Distance Measuring Technology Of Front Vehicle Based On Binocular Vision

How to measure the distance to the front vehicle accurately is the basis of preventing the vehicle from the rear-end collision.There are many questions in the traditional automobile-assisted driving system(ADAS),such as the real-time performance of the vehicle in front is not high and the precision is low.This paper proposes a method of vehicle detection and distance measurement based on binocular vision.The vehicle can be quickly and accurately measured by reducing the area of interest detected by the vehicle and improving the ORB feature matching algorithm that affects the accuracy of ranging.Main work and results are as follows:(1)Study the line-duality of the Hough transform algorithm and the method of the lane-line imaging slope(angle)to select the lane line.Using the lane-line information to demarcate the interest trapezoidal region of the detected vehicle and exclude the interference of the unrelated factors such as sky and trees to improve vehicle recognition rate and detection rate.(2)According to the mathematic characteristics of road surface gray histogram showing Normal distribution,this paper proposes a method of vehicle shadow segmentation based on dynamic threshold,which sets the shadow segmentation threshold to divide the gray value of the road surface to solve the traditional shadow detection problem that due to the gradation transformation of the road surface.And then according to the shadow of the vehicle bottom we delimit the vehicle candidate regions,which further improves the rates of the vehicle recognition r and the detection.(3)Using the vehicle MB-LBP feature to construct a weak classifier,then using the Adaboost machine learning algorithm to train the weak classifier off-line to form a strong classifier for the vehicle detection.Through experiments,it can be proved that the rates of the vehicle recognition and error are significantly improved compared with the traditional algorithm.(4)Proposing an improved ORB feature matching algorithm.Firstly,using the Laplacian extreme value equation to test the corner points of the traditional ORB algorithm,and eliminating the unstable corner points.It solves the problem of a large number of mis-matching problems in traditional ORB feature matching effectively.The experimental results show that the improved algorithm has a significant improvement in the accuracy of the disparity map and the anti-noise performance.