基于深度相机的障碍物识别  

作　　者：司徒仕忠 邱广萍[1] 王锦春 

机构地区：[1]华南农业大学珠江学院

出　　处：《科技创新与应用》2019年第27期37-40,共4页Technology Innovation and Application

基　　金：2017广东省普通高校特色创新类项目(教育科研项目)(编号:2017GXJK242);2018国家级大学生创新创业项目(编号:201812623006)

摘　　要：为实现机器视觉系统的障碍物信息进行探测和识别,文章提出利用深度相机识别障碍物的方法:以障碍物和平面的距离变化梯度不同为主要判别方式,搭配孔洞填充,中值滤波,漫水填充,数学形态学运算,帧间运算等噪声抑制算法,实现对障碍物的探测。根据周围环境信息提取具有一定形状特征的典型目标物。实验结果表明,此方法能有效分别障碍物和平面。In order to detect and recognize obstacle information in machine vision system, this paper presents a method of using depth camera to identify obstacles: taking the difference of distance gradient between obstacle and plane as the main discriminant method, matching with hole filling, median filtering, flooding filling, mathematical morphology operation, inter-frame operation, etc. Noise suppression algorithm is used to detect obstacles. According to the surrounding environment, the typical objects with certain shape characteristics are extracted. The experimental results show that this method can effectively distinguish obstacles from planes.

关 键 词：障碍识别 深度相机 无人驾驶 机器视觉 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]