Research On Leukocyte Recognition Algorithm In Leucorrhea Microscopic Image

The number and status of leukocytes in leucorrhea microscopic image is an important sign and basis to judge the health of the female reproductive tract.The recognition and counting of leukocytes in the microscopic image of leucorrhea is an important part of routine examination of leucorrhea,and it is also an effective means of gynecological clinical diagnosis.To solve the problems of complicated operation,time-consuming,labor-intensive,and low detection efficiency of traditional manual microscopic examination,combined with the characteristics of leucorrhea microscopic image,this paper uses digital image processing method to complete the segmentation and recognition of leukocytes.In the process of leukocyte segmentation,a Canny edge detection algorithm which is based on genetic algorithm is proposed.In its threshold selection link,the threshold is determined by the maximum entropy,and the optimal dual threshold is selected by the genetic algorithm.Firstly,the foreground objects in the leucorrhea microscopic image is extracted by the Canny edge detection algorithm based on genetic algorithm.Secondly,leukocyte were screened out according to the connected region and external rectangle parameters of the foreground targets.Compared to other segmentation algorithms,the proposed segmentation algorithm has higher sensitivity and better segmentation.In the leukocyte recognition link,this paper proposes an improved Res2 Net network recognition algorithm based on the fusion channel attention mechanism.The improved SE-Res2 Net network integration of channel attention mechanism and multi-scale feature extraction.Experiments have proved that compared with the leukocyte recognition algorithms based on Res Net,DenseNet,Res Ne Xt and Res2 Net,the training speed and recognition accuracy of the SE-Res2 Net,which integrates the channel attention mechanism,is much batter,and the recognition accuracy rate reaches 96.4%.