Research on vision system of intelligent sorting robot based on deep learning

Abstract

It is an important step to realize automatic and intelligent production of coal mine to use intelligent sorting robot instead of manual. The vision system, as the “eye” of the intelligent sorting robot, completes the rapid identification, positioning and grouping of the sorting target. Based on YOLOv5, the vision system uses GhostNet to carry out the lightweight design of the model, aiming to ensure the detection accuracy while making the entire model more lightweight, so as to improve the model recognition speed and reduce the operating cost. The model recognition speed of Ghost-YOLOv5 designed and developed is 33FPS, the model size is only 4,2 Mb, and the average detection accuracy is 96,7 %.

Keywords

• intelligent sorting robot
• lightweight model design
• detection model
• vision system