基于边缘特征的双目视觉室外地图构建方法研究  

Research on edge feature based construction method of binocular vision outdoor map

作　　者：周思达 谈海浪 唐嘉宁 蒋聪成 ZHOU Sida;TAN Hailang;TANG Jianing;JIANG Congcheng(School of Electrical and Information Engineering,Yunnan Minzu University,Kunming 650500,China)

机构地区：[1]云南民族大学电气信息工程学院,云南昆明650500

出　　处：《现代电子技术》2022年第2期37-42,共6页Modern Electronics Technique

基　　金：国家自然科学基金项目(61963038)。

摘　　要：针对基于双目深度图的室外大规模地图构建计算冗长,而在无人系统的有限算力下,计算效率需求显著的情况,文中提出一种基于双目视觉立体匹配的三维地图构建方法。首先针对由立体匹配算法及原图引入的噪声误差等问题,采用双线性插值优化视差图;其次通过中值滤波以平滑视差图;最后利用Sobel算子实现对环境物体边缘特征点的提取,并二值化边缘特征图,优化深度值计算环节,在一定程度上提高地图构建的效率。实验结果表明,通过对KITTI数据集的地图构建,可得到的良好八叉树地图,效率提高了50%,证明了文中算法的可行性与有效性,能够满足无人系统在室外路径探索对三维地图的需求。As the calculation of the outdoor large-scale map construction based on binocular depth map is tedious,and the demand for computing efficiency is significant under a condition of the limited computing power of unmanned system,a method of three-dimensional map construction based on binocular vision stereo matching is proposed.In allusion to the noise error caused by the stereo matching algorithm and original image,the bilinear interpolation is used to optimize disparity map,and then the disparity map is smoothed by the median filtering.Finally,Sobel operator is used to extract the edge feature points of environmental objects,and the edge feature map is binarized to optimize the calculation link of depth value,which improves the efficiency of map construction to a certain extent.The experimental results show that the good octree map can be obtained by the map construction of KITTI dataset,which improves the efficiency by 50%,and proves the feasibility and effectiveness of the algorithm.It can meet the needs of three-dimensional map for unmanned system in outdoor path exploration.

关 键 词：室外地图 双目视觉 地图构建 视差图优化 边缘特征提取 立体匹配 深度值计算 

分 类 号：TN911.73-34[电子电信—通信与信息系统] TP391.41[电子电信—信息与通信工程]