Classification and detection method of Blood lancet based on VGG16 network

Abstract

Detection method based on convolutional neural network aims to classify and detect blood lancet after production. The Canny detection algorithm is used to multi-target detection on blood lancet images. The Canny algorithm extracts the blood lancet target, expands the blood lancet data set through image augmentation technology, and solves sample imbalance. The pre-trained VGG16 model is used for blood lancet classification training and fine-tuning of parameters, and the adjusted network model is used for lancet classification and detection. The comparative experiments of the VGG16 network and the two classification networks of GoogLeNet and ResNet show that the average recognition rate of the VGG16 model for blood lancet classification reaches 98.12%, and the classification and detection ability is better than the comparison algorithm. It proves that the network can complete the classification and recognition of blood lancet, and provide technical support for the intelligent classification of blood lancet after production.