Quasi-circular particle target count method based on machine vision

Abstract

The invention discloses a quasi-circular particle target count method based on machine vision. The count method comprises the steps of carrying out image collection; carrying out image enhancement preprocessing; extracting expansion edge features of quasi-circular particles; reconstructing features of the quasi-circular particles according to the extracted expansion edge features; and segmenting the features of the quasi-circular particles, carrying out denoising, and counting quasi-circular particle targets. According to the method, on the basis of the expansion edge feature information amongthe quasi-circular particles, the features of the quasi-circular particles are reconstructed through adoption of a feature reconstruction method, so the quasi-circular particles are rapidly and accurately segmented and counted; the count algorithm is high in efficiency, good in robustness and high in accuracy; a timeliness requirement of a production line can be satisfied; an automatic level of the production line is improved; and the method has practical application value.

Description

technical field [0001] The invention belongs to the technical field of machine vision and pattern recognition, in particular to a method for counting round-like particle targets based on feature reconstruction in machine vision and image features. Background technique [0002] The image analysis and processing technology of quasi-round particles is an important research content in the field of computer vision applications. It has a wide range of application requirements in different fields such as medical health, agricultural product processing, and industrial production, such as: packaging counting of agricultural products such as apples and eggs, Counting of industrial production such as rods and mouth rods, early detection and analysis of the number and shape of cell particles in various microscopic images for diagnosis of major diseases. These round particles have characteristic information such as close arrangement, similar texture and grayscale characteristics, and the...

AI Technical Summary

Problems solved by technology

However, the particle target segmentation based on the edge analysis method cannot solve the following problems well: ①The shapes of the stacked particles are different, and the separation points at the interconnections are not clear; ②The squeezed edge of the particle target is not smooth and irregular The separation points are prone to noise interference; ③The contrast between the stacked particle regions is very small or even non-existent; ④The algorithm is complex and the scope of application is limited, so it is not suitable for large-scale industrial production applications

However, there are two problems in the segmentation method based on morphology: ① When the particles are stacked and squeezed severely, the adhesion will shrink unevenly, making it difficult to segment the stacked area into independent target nuclei even after several times of erosion; ② Corrosion Dilation and dilation are not reciprocal operators. Even after the same number of erosion and dilation operations, the original image cannot be guaranteed to be completely restored, resulting in segmentation errors.

Images