基于EPF最大互相关配准的机械臂目标识别方法

针对工业机械臂在复杂背景下多目标并存识别问题,提出一种基于边缘势场函数最大互相关匹配算法。该算法引入边缘势场函数(edge potential function,EPF)理论计算目标图边缘点的势场函数,有效地利用了复杂结构中单边缘点的联合效用,并在特征点变换模型约束下采用最大互相关相似性度量完成同名控制点的建立与目标配准定位。仿真实验表明,该算法对因成像条件引起噪声旋转的复杂背景图像识别效果良好,精准度高于传统相关匹配算法。

Image registration algorithm based on EPF and maximum cross-correlation for target recognition of manipulator

Aiming at the multi-object issue of manipulator self-recognition in complex background, this paper proposed a ima-ge matching algorithm based on edge potential function and maximum cross-correlation. The algorithm integrated EPF theory into calculating the potential function of target contour points that could use the synergistic effect of the single edge points in complex structure efficiently, and established corresponding control points using maximum cross-correlation as similarity mea-surement under the constraints of feature points transformation model, then finished target registering and localization. Simulation and experiment result shows that it has good recognition effect for complex background image with noise and rotation caused by imaging conditions, and accuracy is better than traditional correlation match.