Vehicle obstacle avoidance method based on binocular vision and deep learning and electronic equipment

Abstract

The invention discloses a vehicle obstacle avoidance method based on binocular vision and deep learning. The method comprises the following steps: obtaining an RGB image in front of a vehicle; acquiring a depth information map in front of the vehicle; predicting a drivable area segmentation result; carrying out information fusion to optimize a drivable area; removing the depth information of a drivable area part, and generating a depth obstacle front view; acquiring an obstacle distribution condition in a three-dimensional space in front of the vehicle, and obtaining an aerial view obstacle scatter diagram according to the obstacle distribution condition; performing density clustering on the aerial view obstacle scatter diagram, and removing noise; performing euclidean distance transformation on the aerial view obstacle scatter diagram, setting an adaptive threshold value, and dividing a front map into a safe driving area and a dangerous area; constructing a map for path planning through the aerial view safe driving area map and the field angle boundary information; performing obstacle avoidance path planning by using a dynamic window method in combination with a map for path planning; and calculating an expected speed and an expected angle according to the obstacle avoidance path track and issuing the same to a control system. The invention further provides the corresponding electronic equipment.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and more specifically relates to a vehicle obstacle avoidance method based on binocular vision and deep learning. Background technique [0002] Navigation is one of the most important basic functions in the autonomous driving decision-making system, and it is also an important bridge between environmental perception and vehicle control. It determines how the vehicle uses environmental information to make scientific judgments with the goal of safely reaching the destination. . More specifically, navigation can be divided into global navigation and local navigation. [0003] The mainstream method of global navigation is to obtain location information based on GPS and other positioning satellites, and use the global path planning algorithm to comprehensively consider road conditions, path length and other conditions to comprehensively plan the road from the starting point to the ending poin...

AI Technical Summary

Problems solved by technology

[0007] The existing obstacle recognition technology based on the binocular camera has the following problems: the traditional ground filtering algorithm is usually complicated and the effect is not good, the generalization is poor, and it cannot cope with the ups and downs of the road, and the ground is misidentified as an obstacle, which will cause the vehicle to run unstable state

If only obstacles are extracted from the field of view of the camera and the ground level, a lot of detection information on low obstacles will be lost, and the semantic information of the generated obstacle avoidance map will be weak; errors in the distance measurement of the depth camera will lead to inaccurate judgments on the position of the edge of the obstacle. Easy to be disturbed by noise, direct use will lead to mislocation of obstacles, etc.

Images