Unmanned aerial vehicle sensing system based on three-dimensional vision

Abstract

The invention discloses an unmanned aerial vehicle perception system based on three-dimensional vision, and the system comprises a vision obtaining unit which is used for a vision system of an unmanned aerial vehicle to obtain the image information of the flight environment of the unmanned aerial vehicle, and carries out the processing of the obtained image information, so as to obtain the image information of an obstacle; a distance acquisition unit which is used for acquiring distance information from the unmanned aerial vehicle to the obstacle by an annular sensor of the unmanned aerial vehicle; a three-dimensional terrain construction unit which is used for constructing a depth channel of a three-dimensional space coordinate graph by using the image information obtained by the vision obtaining unit to obtain an obstacle depth graph of the image; and an obstacle sensing unit which is used for fusing the acquired image information of the obstacle and the distance information by an information fusion processor of the unmanned aerial vehicle, acquiring sensing information of the obstacle, and performing obstacle detection and recognition on the image information to obtain the pixel position, the pixel size and the obstacle type information of the obstacle in the image.

Description

technical field [0001] The invention relates to an unmanned aerial vehicle perception system based on three-dimensional vision. Background technique [0002] Obstacle perception applications in vision-based flight environments can use monocular or binocular vision, among which binocular vision can obtain higher precision than monocular vision, and is more widely used. Monocular vision uses an airborne camera to obtain flight images, but the 3D information of the flight environment will be lost during the image projection process. Although the depth information of the environment can be restored by using multi-frame images or offline training methods, the processing process is complicated, resulting in The onboard embedded processor is difficult to achieve real-time processing. At present, the only way to achieve obstacle avoidance control is to transmit the image back to the ground station for calculation and then return to the UAV. The binocular vision is based on the prin...

AI Technical Summary

Problems solved by technology

Monocular vision uses an airborne camera to obtain flight images, but the 3D information of the flight environment will be lost during the image projection process. Although the depth information of the environment can be restored by using multi-frame images or offline training methods, the processing process is complicated, resulting in The onboard embedded processor is difficult to achieve real-time processing. At present, the only way to achieve obstacle avoidance control is to send the image back to the ground station for calculation and then return to the UAV.

Images