Method and device for detecting barrier in traffic lane

Abstract

The invention discloses a method and a device for detecting a barrier in a traffic lane. The method comprises steps of obtaining a current video frame; carrying out key point detection of a barrier based on the current video frame; dividing an area according to the area to which the key points belong, and selecting observation points from key points in each subarea after division; tracking the observation points and selecting an observation point of the best response performance in each subarea; obtaining the center of mass of each subarea according to the observation point of the best response performance in each subarea; and calculating a time to collision (TTC) of the barrier according to a distance between the center of mass of each subarea in the current frame and the center of mass of each subarea in a former frame. Under a condition with a single camera, the TTC of the barrier can be accurately and timely calculated without acquiring vehicle speed information.

Description

technical field [0001] The invention relates to the technical field of intelligent driving, in particular to a method and device for detecting obstacles in driving. Background technique [0002] With the continuous development of science and technology, intelligent navigation and driving technology have begun to be developed in the driving of motor vehicles. How to detect forward obstacles is a very important technical branch of intelligent navigation and driving. In the prior art, there are two schemes. One scheme is to use ranging sensors such as laser radar and millimeter wave radar. The advantage of this scheme is that it can directly obtain the precise direct distance between the obstacle and the driving vehicle. It is simple and has high measurement accuracy, but the equipment is expensive and the installation is complicated, and it will change the appearance of the vehicle. Another solution is a vision-based ranging solution, in which vision is divided into a monocu...

AI Technical Summary

Problems solved by technology

The advantage of vision-based monocular FCWS (forward collision warning system) is that it only needs an ordinary camera, so it is cheap, easy to install, and does not change the appearance of the vehicle, etc., but its disadvantage is that in order to obtain a more accurate relative distance Or TTC (time to contact) requires relatively high requirements on the algorithm. For example, the existing algorithm has Mobileye's FCW, which requires more accurate spatial positioning of obstacles or pedestrians, and also requires vehicle speed information.

The advantage of the binocular vision-based solution is that the algorithm is relatively intuitive, that is, the distance between obstacles and vehicles is calculated by calculating the disparity between the two video inputs of the two cameras read in, and the TTC is calculated by the distance change between the two frames before and after. , but its disadvantage is that the calculation inspection algorithm is complicated, real-time calculation cannot be achieved, and special computing equipment is required

Images