【摘要】 针对工业应用中遮挡场景下物体位姿难识别的现象，论述利用机械臂手眼系统在多种遮挡场景下实现多视角位姿估计的可行性，提出一种基于关键点数量及分布的融合多视角位姿评估结果的位姿检测方法。利用机械臂手眼系统可移动特性，根据单次识别所得关键点数量分布分割拟合网格区域，选择二次位姿观测位点组，控制手眼系统从多个视角提取RGB图像，通过位姿融合策略得到优化位姿评估结果。实验表明，提出的方法大幅提升了遮挡场景下位姿检测的准确性。更多还原

【Abstract】 Aiming at the difficulty of object pose recognition in industrial application in occlusion scenes,the feasibility of implementing pose estimation from multiple perspectives by using robotic arm eye-in-hand system in multiple occlusion scenes was discussed. A multiple-view pose detection approach was proposed based on the number and distribution of key points and the results of pose evaluation from multiple perspectives. The mobile characteristics of the eye-in-hand system of the manipulator were used to segment and fit the grid area according to the distribution of the number of key points obtained in the first single recognition. The twice position and pose observation site group was selected;and the eye-inhand system was controlled to extract RGB frames from multiple perspectives. The optimized position and pose estimation results were obtained through the pose fusion strategy. Experiments showed that the proposed method greatly improved the accuracy of pose detection in occlusion scenes.更多还原