Hierarchical recognition method of biofluid images

By combining multidimensional feature vectors and classification sampling with multi-round clustering optimization, the problem of balancing speed and accuracy in biological fluid image layer recognition is solved. It achieves efficient and accurate layering under complex conditions, adapts to uneven lighting and noise interference, and has adaptive adjustment capabilities.

CN122244858APending Publication Date: 2026-06-19QINGDAO HAIER BIOMEDICAL TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
QINGDAO HAIER BIOMEDICAL TECH CO LTD
Filing Date
2026-02-24
Publication Date
2026-06-19

AI Technical Summary

Technical Problem

Existing methods for layered identification of biological fluid images suffer from the problem of balancing speed and accuracy. In particular, traditional clustering algorithms have an excessive computational burden in high-resolution image processing, making it difficult to meet the needs of high-throughput analysis. Furthermore, the layer boundaries are prone to deviation under uneven lighting and noise interference.

Method used

We employ a multidimensional feature vector construction, a classification sampling strategy, and a multi-round clustering optimization method. Combining color and location information, we perform multi-round clustering on a representative sample set using the k-means clustering algorithm. We use the median to determine the boundaries and set an adaptive mechanism to correct the stratification results.

Benefits of technology

It significantly improves the speed and accuracy of layer recognition, can stably and accurately layer under complex image conditions, adaptively adjusts the layering results, adapts to uneven lighting and noise interference, and meets the high-efficiency requirements of clinical diagnosis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122244858A_ABST
    Figure CN122244858A_ABST
Patent Text Reader

Abstract

This invention relates to the field of image processing, specifically providing a hierarchical recognition method for biological fluid images. The method includes: constructing a feature vector for each pixel in an acquired biological fluid separation image; wherein the feature vector includes color information and spatial location information of the pixel; classifying all pixels into multiple categories based on the spatial location information, and sampling from each category; running a multi-round clustering algorithm based on the samples to obtain an optimal clustering model; and segmenting the image according to the model to obtain a hierarchical result. This invention performs classification sampling based on the vertical position of pixels, avoiding the loss of key layers, and achieves an order-of-magnitude speed improvement without sacrificing accuracy by running time-consuming clustering algorithms on subsets.
Need to check novelty before this filing date? Find Prior Art