基于多视角特征点匹配的室外目标定位

Outdoor object localization based on multiple views of feature points matching

受室外大范围场景中景物多样性和复杂性的影响,移动机器人在进行目标检测与定位时,运动视角的改变往往导致定位成功率下降.针对此提出一种基于多视角特征点匹配的方法进行室外目标定位.该方法首先获取图像的SURF特征进行匹配,然后结合FLANN和KNN算法滤除错误匹配点,有效地提升了匹配质量,节省了运算时间.通过对不同视角下景物模板的匹配判断,经透视变换映射出目标物体,最终实现目标定位.以校园场景景物为定位目标进行实验.结果表明该方法有效提升了单视角目标定位的成功率,具有较好实时性.

Because of diversity and complexity of objects in large-scale outdoor scene,change of views always leads to the decline of success while mobile robot is detecting and locating objects.To deal with this problem,an outdoor object locating method based on multiple views of feature points matc-hing was proposed.SURF was employed to describe and match the character.Then the FLANN and ANN algorithm were applied to filter error matching points,which could effectively improve the quali-ty of matching and save the computation time.Finally the target object was mapped out through per-spective transformation,by template matching under different angles of view.Experiment results demonstrate that this approach has higher classification accuracy and better real-time performance.