3D Environment Modeling Method for UAV Based on Vision-aided LiDAR

Abstract

The invention discloses a three-dimensional environment modeling method for an unmanned aerial vehicle based on a visual aided laser radar, which includes calibrating a binocular camera based on Zhang Zhengyou's calibration method, and then using a 3D marker method to perform external parameter calibration on the camera / lidar, Finally, the binocular visual odometer is used as the high-frequency auxiliary of the lidar odometer, providing outlier point elimination and relative pose initial value for the lidar odometer, and the low-frequency operation of the lidar odometer, using the relative The pose initial value further optimizes the motion. The invention realizes the high-precision modeling of the flight environment of the UAV, which can not only lay the foundation for the UAV to realize autonomous obstacle avoidance, thereby ensuring the safe flight of the UAV, but also provides a reliable channel for the acquisition of intelligence and resource information, and improves the efficiency of combat operations. efficacy.

Description

technical field [0001] The invention relates to the field of image processing and control of unmanned aerial vehicles, in particular to a three-dimensional environment modeling method for unmanned aerial vehicles based on visual aided laser radar. Background technique [0002] The multi-UAV formation cooperative combat mode can give full play to combat effectiveness and enhance combat effectiveness, and has gradually become a hot research topic today. It is said that the U.S. Defense Advanced Research Projects Agency has planned to launch highly intelligent drones that can be used in actual combat in 2025. Intelligent drone formation requires drones to be able to identify and monitor obstacles that may be encountered on the flight path in real time, so as to formulate a reasonable avoidance strategy, and finally complete the established tasks, such as transporting materials, detecting terrain and striking targets Wait. Therefore, UAVs must have the ability to model the sur...

AI Technical Summary

Problems solved by technology

However, the binocular vision sensor also has shortcomings such as limited image accuracy, depth direction error gradually increases with distance, resulting in limited depth perception distance, and inability to work in scenes where feature points are missing.

The image feature point selection strategy of the laser sensor is relatively simple, and there may be many mismatching points, but its measurement accuracy is high, and the depth observation range is wide, and there is a strong complementarity with the visual sensor.

Images