Rotating box object detection method for female vaginal pathogens based on r3det

Abstract

The invention relates to an R3Det-based method for detecting a rotating frame target of female vaginal pathogens, comprising the following steps: preparing an image to be trained, marking a target object in the image to be trained in the form of a circumscribed upright rectangular bounding box; The rectangular bounding box is converted into a circumscribed rotating rectangular bounding box; the training set of the circumscribed rotating rectangular bounding box of the detection object is constructed; based on R3Det and combined with ResNet152, the constructed training set is trained to obtain the pathogen's rotating frame target detection model; Input the image to be detected into the rotating frame target detection model for detection, and obtain the detection result of the rotating frame of the pathogen. The invention effectively improves the detection rate of the pathogen in the complex scene where the pathogen crosses and stacks with other objects, and reduces the missed detection rate and false detection rate. R3Det can meet the detection accuracy and fast detection speed to meet the high-efficiency needs of medical diagnosis, thereby further improving the detection accuracy and detection rate.

Description

technical field [0001] The invention relates to the technical field of intelligent detection and recognition of pathogens in microscopic images of gynecological reproductive tract microecology, in particular to a method for detecting a rotating frame target of female vaginal pathogens based on R3Det. Background technique [0002] Blastospores and pseudohyphae are common pathogenic pathogens in the female reproductive tract, which often cause female mycotic vaginitis. Therefore, improving the detection rate of blastospores and pseudohyphae plays a vital role in the diagnosis of female mycotic vaginitis. Blastospores are pathogenic molds with two spores formed by spore budding. If blastospores continue to grow, grow into germ tubes, and then gradually grow into filaments, they become pseudohyphae. Currently, blastospores and pseudohyphae Morphological examination under a microscope is still the gold standard for microecological examination of the female reproductive tract. ...

AI Technical Summary

Problems solved by technology

However, the two types of technologies used in the prior art have some disadvantages in the detection and identification of microecological pathogens in the gynecological reproductive tract

[0004] First of all, the scene of the microscopic image of the gynecological reproductive tract microecology is very complicated. Blastospores and pseudohyphae often cross-stack with the target objects of this category or other categories, which affects their morphology. Therefore, it is very difficult to adopt traditional machine vision technology It is easy to cause missed detection of these two types of pathogens

[0005] In addition, the upright frame target detection technology based on deep learning (such as Faster-RCN, SSD, YOLO, etc.) can improve the detection rate of these two types of pathogens compared with traditional machine vision technology, but it has the following problems: First, this type of The vertical frame target detection technology relies on the dataset of pathogens marked by vertical rectangular frames, because pathogens often cross-stack with other targets, so there will be other targets in the background of pathogens marked by vertical rectangular frames, which will affect the Subsequent identification of pathogens will cause interference, which will affect the accuracy of the target detection model and easily cause missed detection of pathogens. For example, when there are many and dense pseudohyphae, it is easy to cause multiple pseudohyphae on the background of the vertical rectangular labeling frame. Affect the accuracy of pseudohyphae detection, and it is easy to miss detection; when blastospores are stacked with blastospores or other types of targets, there will be other targets in the upright rectangular labeling frame, and blastospores themselves are small, so the blastospores are very small. The morphology of blastospores has a great impact, which widens the intra-class gap in the training set, affects the accuracy of blastospore detection, and is easy to miss. The pathogens marked by the upright rectangular frame cannot quantify the geometric characteristics of these two types of pathogens, such as width, length, etc., so that it is impossible to screen out interference targets similar to pathogens, which may easily cause false detection of pathogens

Images