基于YOLO v3的自动餐具分拣检测

Automatic tableware sorting detection based on YOLO v3

当今餐饮行业中餐具的使用愈发频繁,而现有的人工餐具分拣方法效率低、安全性差且成本过高.为此,设计了一套基于YOLO v3的自动餐具分拣系统,包括硬件数据采集、图像识别及分拣执行三部分.其中,基于机器视觉的采集系统利用摄像头将餐具图像传入计算机,并在摄像头背面配上小夜视灯以实现补光.再引入Gamma校正进行图像预处理,同时基于YOLO v3实现对餐具的识别与定位.最后通过串口传递信息以控制机械臂实现分拣.实验结果表明,利用YOLO v3能够达到较好的餐具检测效果.

The use of tableware in today's catering has become increasingly frequent.The existing manual tableware sorting method remain inefficient,lack in safety and costliness.Due to these disadvantages and based on YOLO v3,here we have an automatic tableware sorting system,which includes data acquisition,image recognition and sorting execution.In the system,images are transferred to the computer by using the camera with a small night vision light on the back for the fill light and preprocessed by Gamma correction next.Then the system implements the identification and the positioning of tableware based on YOLO v3 and finally transfers control commands through the serial port to allow the robot arm execute tableware sorting.The experimental results show that the system achieves better detection effects of tableware by using YOLO v3.