A vehicle detection and tracking method based on monocular vision

Abstract

The invention claims a monocular vision-based vehicle detection and tracking method, which belongs to the field of machine vision. The method is as follows: in the vehicle detection stage, use the region of interest extraction, adaptive Canny edge detection and lane line detection to determine the road driving area, and then use the local gray value statistics and the two maximum between-class variance methods to comprehensively use Thresholding method obtains the bottom shadow of the vehicle, and then puts forward a hypothetical vehicle, and then uses the co-occurrence matrix method of texture description to verify the hypothetical vehicle; 1. Whether the search is successful or not and whether the vehicle is out of the boundary or not are three major criteria. Each new target detected twice in a row will be treated as a new tracking target, and the tracking target will be updated continuously. This method can realize vehicle detection and multi-target tracking under dynamic background, and has high real-time performance, accuracy and reliability.

Description

technical field [0001] The invention relates to a method for real-time vehicle detection and tracking in an assisted driving system, belonging to the field of machine vision, in particular to a method for vehicle detection and tracking based on monocular vision. Background technique [0002] When the vehicle is driving on the road, the losses caused by rear-end collision with the vehicle in front or improper avoidance are huge. Using the safety assistance driving system can not only reduce the pressure on the driver, but also give early warning and corresponding measures before the accident occurs , thereby effectively reducing the occurrence of traffic accidents. The vehicle anti-collision warning system in the safety assisted driving system is the most important part of the system. The vehicle anti-collision warning system judges whether there is a vehicle in front by searching the area in front of the road, and locates the vehicle in front if there is a vehicle. The spee...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the problems of poor real-time performance, anti-interference performance and high false detection rate in the vehicle detection process in the prior art, the present invention provides a method to improve the real-time performance, anti-interference performance and reduce false detection rate of vehicle detection. Vehicle detection and tracking method, the technical solution of the present invention is as follows: a vehicle detection and tracking method based on monocular vision, which includes the following steps: in the detection stage of moving vehicles, 101, the road driving area extraction step: specifically, A1 1. Obtain the image information around the car, remove the region of the image part that does not contain road information, and extract the region of interest; A2, use the adaptive canny edge detection method to detect the entire region of interest extracted in step A1 The edge of the image; A3, then use the Hough transform to detect all straight lines in the image area, and vote for each straight line, and get the straight line with the largest number of votes in this area; A4, set the threshold for the angle of the left and right lane lines, and establish In the polar angle constraint area, the positions of the left and right lane lines are delineated in the polar angle constraint area, and the straight line with the largest number of votes in the polar angle constraint area is selected as the lane line, thereby locating the road driving area;

Images