基于双目视觉的无人机编队相对定位算法

Relative Positioning Algorithm of UAV Formation Based on Binocular Vision

针对无人机编队飞行时双目视觉定位精确性差、计算量大、实时性不高的技术现状,对基于特征点的FAST定位和BRIEF旋转(Oriented fast and rotated brief,ORB)算法进行了改进,提出了一种适用于无人机双目视觉定位的算法。在改进ORB算法中,采用提取目标区域、最近邻约束和随机抽样一致(Random sampling consensus,RANSAC)方法,提高了特征点提取与匹配效率,也提高了特征点匹配质量;对于双目视觉定位,提出了适用条件更加宽泛的双目视觉定位模型,并保证了模型的定位精度;最后使用卡尔曼滤波算法对无人机的定位信息进行估计,进一步提高了无人机的定位精度。实验表明,算法具有较高的精确性和实时性,满足无人机间的相对定位要求。

Aiming at the technical situation of poor accuracy of binocular vision positioning,large amount of calculation,and low real-time performance during UAV formation flying,the oriented fast and rotated brief(ORB)algorithm based on feature points is improved,and an algorithm suitable for binocular vision positioning of UAVs is proposed.In the improved ORB algorithm,the methods of extracting the target area,nearest neighbor constraint and random sampling consensus(RANSAC)are adopted to improve the efficiency of feature point extraction and matching,and also improve the quality of feature point matching.For binocular vision positioning,a binocular vision positioning model with broader applicable conditions is proposed,and the positioning accuracy of the model is guaranteed.Finally,the Kalman filter algorithm is used to estimate the positioning information of the UAV,which further improves the positioning accuracy of the UAV.The experiments show that the algorithm has high accuracy and real-time performance,and meets the relative positioning requirements between UAVs.