基于极大化思想的无人机安全避障域识别算法  

Recognition Algorithm of Safe Obstacle Avoidance Domain for UAVs Based on Maximization Idea

作　　者：王家亮[1] 董楷 顾兆军[2] 陈辉 韩强 WANG Jialiang;DONG Kai;GU Zhaojun;CHEN Hui;HAN Qiang(School of Computer Science and Technology,Civil Aviation University of China,Tianjin 300300,China;Information Security Assessment Center of Civil Aviation University of China,Tianjin 300300,China)

机构地区：[1]中国民航大学计算机科学与技术学院,天津300300 [2]中国民航大学信息安全测评中心,天津300300

出　　处：《西南交通大学学报》2023年第6期1267-1276,共10页Journal of Southwest Jiaotong University

基　　金：天津市教委科研计划(2020KJ026)。

摘　　要：为提高四轴飞行器避障的准确性与实时性,提出一种结合LK (Lucas-Kanade)光流法和极大化思想的四轴飞行器避障算法.首先,对四轴飞行器采集的视频流进行预处理,得到图像帧;其次,通过LK光流法剔除图像帧中光流小于阈值的角点,采用基于角点距离的聚类算法对角点进行分组,并计算出每组角点的外包轮廓;然后,利用基于极大化思想的安全避障域算法计算最优通行区域,进一步根据避障域求得偏差数据;最后,将偏差数据输入比例微分(PD)控制器得到控制信息,并发送控制指令使四轴飞行器及时调整飞行姿态,完成避障飞行.通过特洛(Tello)四轴飞行器进行不同场景的实验表明,本文所提出的算法计算每帧图像最优安全避障域平均所需时间为0.17 s,既满足无人机避障实时性要求,又解决了识别障碍物区域与计算安全避障域问题.In order to improve the accuracy and real-time performance of obstacle avoidance for quad-rotor helicopters,an obstacle avoidance algorithm combining the Lucas-Kanade(LK) optical flow method and maximization idea was proposed.Firstly,the video stream collected by the quad-rotor helicopter was preprocessed to obtain the image frame.Secondly,corner points whose optical flow was less than the threshold value in the image frame were eliminated by the LK optical flow method,and corner points were grouped by a clustering algorithm based on corner point distance.In addition,the outsourcing contour of each group of corner points was calculated.Then,the safe obstacle avoidance domain algorithm based on the maximization idea was used to calculate the optimal passing domain,and the deviation data were obtained according to the obstacle avoidance domain.Finally,the deviation data were input to the proportional and differential(PD) controller to obtain the control information,and the control command was sent to make the quad-rotor helicopter adjust the flight attitude in time to complete the obstacle avoidance flight.Experiments on the Tello quad-rotor helicopter in different scenes show that the proposed algorithm takes an average of 0.17 seconds to calculate the optimal safe obstacle avoidance domain for each frame of the image,which meets the real-time requirements of unmanned aerial vehicle(UAV) obstacle avoidance and solves the problem of identifying obstacle domains and calculating safe obstacle avoidance domain.

关 键 词：极大化思想 安全避障域 四轴飞行器 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] V279[自动化与计算机技术—计算机科学与技术] V249[航空宇航科学与技术—飞行器设计]