Multiple Object Detection of Workpieces Based on Fusion of Deep Learning and Image Processing*

摘要

A workpiece detection method based on fusion of deep learning and image processing is proposed. Firstly, the workpiece bounding boxes are located in the workpiece images by YOLOv3, whose parameters are compressed by an improved convolutional neural network residual structure pruning strategy. Then, the workpiece images are cropped based on the bounding boxes with cropping biases. Finally, the contours and suitable gripping points of the workpieces are obtained through image processing. The experimental results show that mean Average Precision (mAP) is 98.60% for YOLOv3, and 99.38% for that one by pruning 50.89% of its parameters, and the inference time is shortened by 31.13%. Image processing effectively corrects the bounding boxes obtained by deep learning, and obtains workpiece contour and gripping point information.