目的 为了解决粉体包装领域自动化程度低且工作环境对人体有较大危害的问题，设计一种阀口袋动态拾取系统。方法 首先分析阀口袋动态拾取系统软硬件设计和工作过程；然后，分析相机图像坐标到机器人世界坐标的转换关系、手眼标定过程，以及对采集到的图像进行畸变校正、滤波等图像处理过程；最后，采用基于卡尔曼滤波算法对流水线上运动的阀口袋实现定位和拾取。结果 实验表明，机器人能够对输送机上的阀口袋实现定位和动态拾取，拾取误差小于1 mm，连续运行情况下可实现每分钟5~7袋的上袋速度。结论 该拾取系统运行稳定，具有较高的运行效率和拾取精度。

The work aims to design a dynamic picking system of valve pocket, so as to solve the problem of low au-tomation in powder packaging field and reduce the great harm to human body caused by working environment. Firstly, the software and hardware design and working process of the valve pocket dynamic picking system were analyzed; secondly, the conversion relationship between camera image coordinates and robot world coordinates, hand-eye calibration process, image processing processes such as distortion correction and filtering were analyzed; finally, the valve pocket moving on the assembly line was positioned and picked up by Kalman filter algorithm. The robot could locate and pick up the valve pockets on the conveyor dynamically, and the pickup error was less than 1 mm. Moreover, the bagging speed of 5-7 bags per minute could be realized under continuous operation condition. The picking system runs stably with high efficiency and accuracy.