An Automatic Cell Counting Method Based on Dynamic Learning for Microscope

Abstract

The invention discloses an automatic cell counting method based on dynamic learning for microscopes. By learning the mapping relationship between pixel-level features of cells and distribution functions, new cell image distribution results at the pixel level are obtained. In the present invention, firstly, the cells and the background of the image to be recognized are marked manually, each cell is represented by a point, and a marked area centered on the point is obtained. Then construct a density function, act on the marked area of ​​each cell, and get the corresponding density matrix centered on the marked point. The size of each element value in the density matrix represents the density distribution of cells in pixels, and the total number of cells is obtained by summing the elements of the entire matrix. By building a model and learning the relationship between the feature vector of each pixel in the original image and the corresponding element value in the density matrix, the mapping from feature to density can be determined. Therefore, the estimation result of cell number can be obtained from this mapping relationship.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to an automatic cell counting method based on dynamic learning for a microscope. Background technique [0002] At present, the world's scientific research technology is becoming more and more developed, and the research and utilization of microorganisms has gradually become feasible and important. In the study of microorganisms, it is often necessary to count the number of cells. It is a time-consuming and laborious work to identify cells and count the number only by naked eyes, which greatly reduces the research efficiency. [0003] Most of the traditional cell counting methods rely on the means of cell detection to locate and count individual cells. In addition, industrial cameras are also used instead of traditional microscope eyepieces to magnify and image the cells in the culture dish, and then use image processing to count the number of cells in the imaging area. ...

AI Technical Summary

Problems solved by technology

In the case of incomplete understanding of the data, it is generally difficult to directly select important features

If the number of features is too small, the learning will be under-fitting; if the number of features is too large, some irrelevant features will also be added to the learning range, which will easily lead to over-fitting, which will reduce the learning effect of the model and slow down the calculation speed.

[0008] (2) When repeatedly marking and modifying the pixels of the image, it is necessary to retrain the samples each time. Excessive feature calculation will cause the program to run very slowly, and real-time dynamic learning cannot be realized.

[0009] (3) When estimating the density, if the edge of the cell is blurred due to imaging conditions or staining, it is easy to estimate the edge as the background density, resulting in a decrease in the sum of the density values ​​​​of the area, and the occurrence of cell omission

Images