Real-time detection method for silkworm cocoon feeding

The invention relates to a real-time detection method for silkworm cocoon feeding, and belongs to the technical field of silkworm industry processing. According to the method, a YOLOv3 deep learning target detection algorithm is improved on the basis of anchor point frame parameter presetting, channel pruning and receptive field module embedding, and then cocoon feeding is detected in real time during cocoon selection in the silkworm cocoon production and processing process. A feeding device of the silkworm cocoon sorting machine is used for feeding silkworm cocoons, a conveying device is used for conveying the silkworm cocoons, an image collecting device is used for photographing the silkworm cocoons on the conveying device and obtaining silkworm cocoon images, a lower cocoon grabbing device is used for grabbing and removing the silkworm cocoons, and an upper cocoon storage device is used for storing the silkworm cocoons. The silkworm cocoon sorting machine grabs and removes the lower cocoons according to th