Research on Recognition and Localization of Food for Automatic Grasping Operation

With the increase of peoples demand for all kinds of food, the requirements for logistics are getting higher and higher. In order to improve the accuracy of logistics robot in the process of sorting goods, an improved detection algorithm YOLOv4-MobilenetV3 is proposed in this paper. On the basis of the original YOLOv4 network, CSPdarknet53, the backbone of the original network, was replaced by Mobilenet network, which enhanced the precise positioning ability of the network. Then through the camera calibration method, we can further get the three-dimensional coordinates of all kinds of food. Finally, through experimental verification, this method has high accuracy, and has important research significance for the development of logistics and express industry.