双目视觉在类人机器人测距中的应用

为实现类人机器人对复杂目标高精度识别并测量出目标实际距离，提出一种双目目标识别与测距方法. 首先利用棋盘标定法对摄像机进行标定并捕获图像；然后利用局部二值模式算子（LBP）和优化后支持向量机（SVM）对目标进行识别；在识别的基础上，再采用尺度不变特征变换（SIFT）算法对特征点进行匹配并根据三角测距原理计算出目标实际距离. 实验结果表明该方法不但减少了人为参数指定，并且提高了特征点匹配效率，测距精度能够达到94%以上，满足类人机器人高精度测距和实时性要求.

Application of Binocular Vision Range Measuring in Humanoid Robots

To realize the idea that humanoid robots can recognize complicated targets with high accuracy and measure the actual distance of the target, a binocular method was proposed for recognizing target and measuring distance. Firstly, the camera calibration was realized by the checkerboard method to capture images afterwards.Then Local Binary Pattern operator and optimized Support Vector Machine were employed to identify the target. On the basis of recognition, Scale-invariant Feature Transform algorithm was used to match feature points and to determine the actual distance based on the principle of triangulation. Experimental results show that this method can reduce the designation of artificial parameters and improve the efficiency of matching features, and the accuracy of measuring actual distance is more than 94%, satisfying real-time and high-accuracy requirements of humanoid robots.