一种基于视觉识别的乒乓球捡球机设计与开发

针对目前乒乓球捡球机捡球机构不完善、乒乓球识别算法适应性差的问题，提出一种基于视觉识别的智能乒乓球捡球机.采用树莓派4B开发板作为控制单元，利用轻量化的you only look once （YOLO） v5s算法，对乒乓球进行识别；通过扇叶式集球机构，将乒乓球卷入收纳篮.实验结果表明：在乒乓球数小于150个的情况下，该捡球机的识别精确率与查全率均可达到95%以上，漏检率控制在7%以下.同时，集球机构结构简单、可靠、效率高，整体设计方案具有较好的实际应用价值.

Design and development of a table tennis ball picking machine based on visual recognition

In order to resolve the problems of imperfect ball picking mechanism and poor adaptability of table tennis recognition algorithm, this paper proposes an intelligent table tennis ball picking machine based on visual recognition. We use the Raspberry Pi 4B demo board as the control unit, and recognize the table tennis by the lightly modified you only look once (YOLO) v5s algorithm, and roll the table tennis into the storage basket through the fan-blade type ball collection mechanism. Experimental results show that under the condition that the number of table tennis balls is less than 150, the recognition accuracy and completion rate of the ball picking machine can reach more than 95%, and the missed detection rate is less than 7%. At the same time, the structure of the ball collection mechanism is simple, reliable and efficient, and the overall design scheme has good practical application value.