Method for automatically detecting and avoiding obstacles for unmanned aerial vehicle under complicated environments

Abstract

The invention relates to a method for automatically detecting and avoiding obstacles for an unmanned aerial vehicle under complicated environments. The method comprises the following steps of utilizing a double-lens camera arranged on the unmanned aerial vehicle to collect images in real time, and according to the matching of collected left and right images, calculating to generate a disparity map, and pretreating; according to the information of the disparity map, clustering and classifying the contour and gray-scale value of the disparity map, so as to obtain a zone block with a clear structure; eliminating a noise interference of the zone block, so as to obtain a potential obstacle area; comparing the potential obstacle areas in two disparity maps of the front and back frames, and according to the amplifying characteristics of the obstacles, the distance between the unmanned aerial vehicle and the potential obstacle area, and the area of the potential obstacle area, screening to obtain a final obstacle area; according to the obtained final obstacle area, outputting an action instruction for the unmanned aerial vehicle to avoid the obstacles. The method has the advantage that under the complicated environment with multiple hidden obstacles, the excellent obstacle detection and avoidance ability is realized.

Description

technical field [0001] The invention relates to the fields of computer vision and unmanned aerial vehicle control, in particular to an automatic obstacle detection and obstacle avoidance method for an unmanned aerial vehicle in a complex environment. Background technique [0002] With the extensive research and application of intelligent systems such as drones, obstacle detection and obstacle avoidance has gradually become a research hotspot. Security is a prerequisite for the widespread application of intelligent systems such as drones. Obtaining the position of obstacles by analyzing the images of complex scenes captured by the camera during the real-time movement of the UAV is still a very challenging research problem. [0003] After years of development, vision-based obstacle detection methods have become increasingly mature. Vision-based obstacle detection methods can be roughly divided into monocular vision and binocular vision. The typical method in monocular visio...

AI Technical Summary

Problems solved by technology

[0005] There are also some obstacle detection methods based on ultrasonic, sonar and other sensors. These sensors can detect the environment around the agent, but in a complex environment, multiple ultrasonic information is intertwined and easily confused.

And any sensing device is not perfect, there will always be blind spots or interference from external noise, which leads to a reduction in accuracy

[0006] Looking at the existing obstacle detection methods, although some progress has been made in solving some specific problems, they all start from the surface of the problem and design a scene suitable for this type of problem to detect and test, and have not gone deep into it. Understand the essential characteristics of obstacles so that a certain method can only solve one problem

Images