基于改进SIFT的无人机双目目标识别与定位

UAV Binocular Vision Target Recognition and Positioning Based on Improved SIFT Algorithm

对无人机自主着陆系统中双目视觉采集到的地标图像进行了研究,在分析地标图像中存在模糊噪声以及大量背景干扰问题,提出一种基于改进SIFT算法的无人机双目视觉目标识别与定位方法。首先,采用基于OTSU与HSV的ROI算法对无人机双目图像进行目标识别与分割预处理操作,将目标准确识别;其次,针对双目视觉获取三维信息效率慢的问题,采用基于改进的SIFT算法对已识别的地标进行特征提取,生成二进制描述符,并采用局部敏感哈希算法对特征点进行稀疏匹配,提高目标特征匹配准确度及效率;最后,采用相似三角形原理计算每个特征匹配点的三维距离,获得无人机与目标之间的平均三维距离。实验结果表明所设计的算法相较于传统的SIFT算法更具有可行性和有效性。

Research on the landmark images collected by the binocular vision in the autonomous landing system of the UAV.In the analysis of the problem of fuzzy noise and a large amount of background interference in the landmark images,a binocular vision target recognition for the UAV based on the improved SIFT algorithm is proposed.And positioning method.First,the ROI algorithm based on OTSU and HSV is used to perform target recognition and segmentation preprocessing operations on the UAV binocular image to accurately identify the target.Then,in view of the slow efficiency of binocular vision to obtain 3D information,the improved SIFT algorithm is used to extract the features of the recognized landmarks to generate binary descriptors,and the local sensitive hash algorithm is used to sparse the feature points to improve Target feature matching accuracy and efficiency.Finally,the principle of similar triangles is used to calculate the three-dimensional distance of each feature matching point to obtain the average three-dimensional distance between the UAV and the target.The experimental results show that the designed algorithm is more feasible and effective than the traditional SIFT algorithm.