Detection and identification method for automatically refining and labeling categories of spores

Abstract

The invention relates to a detection and identification method for automatically refining and labeling categories of spores. The method comprises the following steps: acquiring an original image and an expert labeling document; formulating a fine classification rule of the bud spores; designing a rapid and automatic classification method based on an adaptive mask according to a fine classification rule formulated by the bud spores; intercepting all marked areas of interest of the bud spores and storing the areas of interest of the bud spores; automatically classifying the intercepted and stored areas of interest of the germinated spores according to a formulated fine classification rule and an automatic classification method; constructing a training set with a small intra-class gap and a large inter-class gap; training a fine classification AI target detection model for detecting bud spores; and summarizing the bud spores of the fine classification category detected by using the AI target detection model as a final detection result. According to the method, the detection rate of the bud spores is effectively increased, and meanwhile, the false detection rate of the bud spores is reduced; in addition, by adopting a self-adaptive mask rapid automatic classification method, the time and the labor cost of manual re-labeling are greatly reduced.

Description

technical field [0001] The invention relates to the technical field of intelligent detection and identification of blastospores in microscopic images of gynecological vaginal microecology, and in particular to an intelligent detection and identification method for automatic and rapid refinement and labeling of blastospores in gynecological vaginal microecology and divide-and-conquer. Background technique [0002] Blastospore is a common pathogenic fungus of the female reproductive tract. It is a pathogenic mold in which spores germinate to form two spores. Morphological examination under a microscope is currently the gold standard for the diagnosis of gynecological reproductive tract microecology. Therefore, the morphological detection of blastospores and the improvement of the detection rate of blastospores play a vital role in the diagnosis of female fungal vaginitis. [0003] Traditional machine vision methods to detect blastospores usually use a series of methods such as...

AI Technical Summary

Problems solved by technology

[0003] Traditional machine vision methods to detect blastospores usually use a series of methods such as adaptive threshold segmentation, morphological processing, contour detection, geometric shape fitting of the contour, and screening by geometric characteristics of geometric shapes such as perimeter and area. This type of method can detect blastospores with a simple background and scattered distribution, but the blastospores have various shapes, and if the stacking distribution of blastospores or the background is complex, or when the blastospores and other types of microecology such as bacteria are cross-stacked , these conditions will affect the geometric characteristics of blastospores such as circumference, area, shape, etc. Therefore, traditional machine vision methods may easily cause missed detection of blastospores

[0004] The manual labeling of blastospores and the deep learning method of convoluti

Images