Research on the Identification Method of Silkworm Cocoon Species Based on Improved YOLOv3

Silkworm cocoon species identification is a key technical link to realize the intelligent sorting of silkworm cocoons. In order to realize the rapid and accurate identification of some species of group cocoons, an improved YOLOv3 colony cocoon species detection model-CBL-YOLOv3 model is proposed. Taking macular cocoons, rotten cocoons and reelable cocoons as the research objects, the YOLOv3 model is improved: K-means clustering is adopted to obtain the aspect ratio of the prior frame of silkworm cocoons, the composition ratio in the loss function is designed, and the network structure of the model is also modified. The CBL-YOLOv3 model is tested from two aspects: visual effects and evaluation indicators. The test result shows that the detection effect of the CBL-YOLOv3 model is superior to the traditional YOLOv3 model in terms of accuracy and speed, which can meet the needs of real-time detection of group cocoons.