基于假设检验的室内环境多特征检测方法

摘要

In order to solve the adaptability of indoor environmental feature extraction of mobile robots, this paper proposed a hypothesis⁃testing⁃based indoor multi⁃feature identification method. Firstly, the multi⁃dimensional space is built before processing features. Then, the distance probability function is defined, and by combination with theχ 2 hypothesis test theory the feature regions of the sampled points are estimated and segmented. Further, the corner features are recognized by the extreme value method while the features of line segments and arcs are obtained by u⁃sing the constrained least square method. Finally, the results of experiments verify that the proposed method can ex⁃tract the corner feature, the line segment feature and the arc feature steadily, and the feature recognition rate is o⁃ver 94%.%为解决移动机器人室内环境特征提取适应性问题，提出了基于假设检验的室内多特征检测方法。该方法首先构建多维数据空间。通过定义距离概率函数，结合χ2假设检验理论，在估计采样点的特征区域基础上，通过极值法进行角点检测，然后采用约束最小二乘法提取线段与圆弧特征。最后，实验验证了该方法可以提取较稳定的角点、线段以及圆弧特征，同时特征的识别率达到94％以上。