关键词: 自动插件机, 机器视觉, 立体视觉, 系统标定, 特征提取 

Abstract:  At present most of the autoinserting machines adopt the vision positioning method with bottom-camera， but it is difficult for some leads in bottom images of components to be distinguished from the body． In order to improve the positioning reliability，a precise positioning method is proposed of electronic-component leads based on rotational stereo vision． In the proposed method，firstly，a rotational-axis calibration method with checkerboard is put forward，and system calibration is finished． Secondly，side images of the component leads are obtained from multiple angles，a sub-pixel feature point extracted method is proposed on the basis of spatial moment． Then，spatial positioning of leads is calculated by optimizing feature points of multiple images． A vision positioning system with backlight is built for experiment，experimental results show that with the number of images increasing， the accuracy of proposed method increases and converges; grabbing position and angle produce no effect on accuracy of proposed method; the proposed method possesses higher accuracy and is more stable than the traditional method． On the other hand，because images can be obtained when moving，the proposed method brings about higher efficiency than the traditional method． 

Key words: autoinserting machines, machine vision, stereo vision, system calibration, feature extraction