Leukocyte recognition algorithm in leucorrhea microscopic images based on ResNet-34 neural network

Abstract

Automatic leucocyte recognition for leukorrhea microscopic images is a digital image processing technology in the field of machine learning. The existence and quantity of leukocytes in leucorrhea microscopic image is an important sign and basis to judge the inflammation of vagina or cervix. Therefore, the recognition and count of leucocyte is an effective means to evaluate the condition of the disease. To solve the problem of low efficiency of leucocyte recognition in traditional artificial microscopy, this paper proposes an automatic recognition algorithm based on ResNet-34 neural network. Firstly, Canny edge detection algorithm based on genetic algorithm is used to extract the foreground target in the leucorrhea microscopic image. Secondly, the leucocyte target is selected according to the connected region and boundary rectangle parameters of the foreground target. Finally, ResNet-34 neural network is applied for the classification of leukocytes. Experiments show that the recognition accuracy of leukocytes in leucorrhea microscopic image is 92.8%, and the recall is 97.1%, which is higher and better than other methods.