无人飞行器远距目标识别与方位感知方法

Remote target recognition and orientation perception for unmanned aerial vehicles

关键词:  远距目标识别  视觉角度测量  灯光编码解码  紫外标记  无人飞行器  相对导航

Key words:  Remote target recognition  Visual angle measurement  Light encoding and decoding  Ultraviolet marker  Unmanned aerial vehicle  Relative navigation

摘要:

高精度的导航信息是无人飞行器编队完成各项任务的关键。针对无人飞行器利用视觉传感器对远距目标进行导航时存在目标成像特征少和多目标身份信息匹配困难的问题，提出了一种基于光学标记的远距目标识别与方位感知方法。通过分析无人飞行器实际编队飞行需求，设计了基于视觉的相对角度测量和身份ID识别的总体方案；利用光学标记的闪烁特征设计了无人飞行器目标身份的编码解码，并根据获取的灯光序列查询编码库解码识别目标ID；结合光学标记的成像特征实现标记的快速检测与跟踪，在此基础上拟合标记的像素坐标并计算相对角度信息。实验结果表明，在50~150 m的实验环境下，静态测角误差优于0.21°，动态测角误差优于0.73°，可实现10个目标的身份ID识别。

Abstract:

High-precision navigation information is key to the formation of unmanned aerial vehicle (UAV) to perform various tasks. To address the problems of few target image features and difficulties in matching multi-target identity information when using visual sensors to navigate UAVs to remote targets, a method for remote target recognition and orientation perception based on optical markers is proposed. By analyzing the actual formation flight requirements of UAVs, a comprehensive approach of relative angle measurement and unique target ID identification based on vision is designed. The scintillation characteristics of optical markers are used to encode and decode target identities for UAVs, and the encoding database is searched according to the light sequence acquired to decode and identify the target ID. The optical markers can be quickly detected and tracked by combining the imaging features of the optical markers. Based on this, the pixel coordinates of the markers are fitted and the relative angle information is calculated. The experimental results show that within the environmental range of 50~150 m, the static angle measurement error is better than 0.21°, the dynamic angle measurement error is better than 0.73°, and the identification of 10 targets can be achieved.