基于先验知识的月球车视觉系统特征提取方法  

A Feature Extracting Algorithm for Lunar Rover Vision System Based on Prior Knowledge

作　　者：刘佳璐[1] 

机构地区：[1]北京控制工程研究所空间智能控制技术国家重点实验室,北京100080

出　　处：《计算机仿真》2009年第9期77-79,111,共4页Computer Simulation

摘　　要：为了满足月面巡视探测器的自主导航要求。提出一种新的基于先验知识的特征点提取算法。首先,根据先验知识对原始图像进行预分割成危险区域和安全区域,然后在危险区域里面进行特征点的选取工作,对特征点用于以后的匹配和目标定位,进而用于视觉系统的导航工作。与传统算法相比,提取方法可以有效减少计算量,使选取特征点更加准确,提高后续匹配阶段的处理速度和匹配精度。在模拟试验场的双目视觉照片中,通过实验取得较好的效果。The paper presents a new feature points extracting algorithm based on prior knowledge for navigation systems of lunar rover. It separates the original image into safe area and dangerous area for rover at first and then uses Harris corner detection to get the feature points in dangerous area. It searches corner in dangerous region to substitute searching feature points in original image . These points will be used for object location in next phase. This method could reduce computation and improve the speed and detection accuracy in the later matching phase compareing with traditional methods. This algorithm has been used in simulative environment experiments and has obtained good result.

关 键 词：特征提取 图像分割 角点提取 视觉导航 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]