CAJ | 학술논문

该文介绍了一种基于区域特征点匹配和聚类分析的双目视觉避障策略算法。该算法通过Canny算子提取左右视图特征点；首先采用基于相似度的AdaptWeight自适应权重局部匹配算法进行初始匹配，再采用基于动态规划的全局匹配算法进行最终匹配；特征点匹配完成后通过双目立体视觉模型计算特征点深度，并通过深度和位置关系进行聚类分析，分割出障碍物，最终计算出障碍物阻挡区域，以达到障碍物检测目的。实验结果表明该方法能够比较好的实现小车避障。
해문개소료일충기우구역특정점필배화취류분석적쌍목시각피장책략산법。해산법통과Canny산자제취좌우시도특정점；수선채용기우상사도적AdaptWeight자괄응권중국부필배산법진행초시필배，재채용기우동태규화적전국필배산법진행최종필배；특정점필배완성후통과쌍목입체시각모형계산특정점심도，병통과심도화위치관계진행취류분석，분할출장애물，최종계산출장애물조당구역，이체도장애물검측목적。실험결과표명해방법능구비교호적실현소차피장。
This paper presents an obstacle avoiding method based on the characteristic points ( CP) matching and clustering .In this approach , we acquire the characteristic points of left and right image , where the raw image is handled with Canny operator .And then match the characteristic points from the left image and the right image with AdaptWeight self-adaptive local matching algorithm based on similarity .At last complete the global matching with dynamic programming .The depth of the matching characteristic points can be computed with the binocular stereo vision model , which can be used to divide the points into different parts as well as ge-ometric relationship of points .Obstacle is divided , and the final calculate obstacle blocking region , which lead to the purpose of obstacle detection .Experimental results demonstrate that this algorithm is effective in obsta-cle avoiding .