Industrial Parts Detection Based on Deep Learning

Object detection is the core technique of industrial sorting based on machine vision. However, traditional object detection algorithm is difficult to solve sorting tasks in complex industrial scenarios. To this end, this paper proposes an industrial parts detection network based on deep learning. This method is to use the object detection network to detect the parts on the conveyor belts and obtain the category and location information of the parts. In order to improve the network's detection accuracy of multi-scale parts, this paper use the K-Means algorithm to redesign the size of the anchor. In addition, we construct a private dataset for model training and use dataset augmentation to expand our dataset, then using the pre-trained weights based on the VOC dataset for transfer learning. Experiments based on self-constructed dataset show that the mAP of our method achieves 97.03%, which can satisfy the requirements of practical applications.