Adaptive analysis method and system for micro-particle morphology and topological characteristics

Abstract

The present application relates to the technical field of image processing and material characterization, and discloses a kind of micro-particle morphology and topological feature adaptive analysis method and system.For the problem that the accuracy of traditional threshold segmentation is low due to the serious adhesion of particles in complex multiphase system, the present application obtains a microscopic image and performs contrast-limited adaptive noise reduction;extract the global gradient features of the image, dynamically adaptively match the watershed or Hough circle segmentation model;extract the morphological features such as equivalent particle diameter of the scalar domain after segmentation, and construct a multi-dimensional feature matrix combined with topological invariants such as Euler number of connected domain;multidimensional probability density fitting is carried out by maximum likelihood estimation.The present application realizes a closed loop from image parameter adaptive optimization segmentation to physical feature extraction, effectively improves the recognition accuracy of complex overlapping particles, and is mainly used for high-throughput quantitative characterization of microstructure of high polymer, powder and porous material.

Description

Technical Field

[0001] This invention relates to the fields of image processing and material characterization technology, and more specifically, to an adaptive analysis method and system for microscopic particle morphology and topological features. Background Technology

[0002] In materials science, polymer chemistry, and powder engineering, scanning electron microscopy (SEM) and transmission electron microscopy (TEM) are core tools for revealing the microstructure of matter. The particle size distribution, morphological characteristics, and pore connectivity of microparticles directly determine the rheological properties of polymers, the permeability of porous materials, and the physical stability of multiphase flow systems. However, in practical industrial quality inspection and scientific research scenarios, existing quantitative analysis methods for particle characteristics have significant technical shortcomings: First, microscopic images are often limited by extreme imaging conditions, resulting in severe background noise and contrast degradation, and traditional global or local static threshold segmentation is prone to false positives; second, in complex multiphase flow systems or crystal self-assembly processes, particles often undergo severe aggregation and physical adhesion, and existing single morphological segmentation algorithms cannot accurately separate overlapping boundaries, leading to statistical particle sizes that are much larger than the actual physical particle sizes; finally, existing analytical methods usually only output single area or one-dimensional size features, ignoring topological invariants (such as Euler numbers) that characterize the material network structure, and cannot achieve multidimensional, comprehensive, and quantitative characterization of the microscopic physical properties of the system. Therefore, there is an urgent need for a high-throughput quantitative analysis method that can adaptively handle complex backgrounds, intelligently match segmentation models, and combine topological features. Summary of the Invention

[0003] The main objective of this invention is to provide an adaptive analysis method and system for the morphology and topological features of microscopic particles, addressing the technical problems of low segmentation accuracy and limited feature representation dimensions in existing technologies when processing overlapping particles. To achieve this objective, a first aspect of this invention provides an adaptive analysis method for the morphology and topological features of microscopic particles, comprising the following steps: acquiring a microscopic image of the complex multiphase system to be analyzed, and performing contrast-limited adaptive histogram equalization and nonlocal mean denoising preprocessing on the microscopic image; extracting global image features from the preprocessed image, and dynamically adaptively matching a target segmentation model based on the global image features, wherein the target segmentation model is either a morphological watershed model or a Hough circle detection model; using the matched target segmentation model to identify and segment the preprocessed image, obtaining the independent scalar domains of the target particles; extracting morphological features from the independent scalar domains, calculating the topological invariants of the connected components of the target particles, and fusing the morphological features and the topological invariants to construct a multidimensional feature matrix; and based on the multidimensional feature matrix, performing multidimensional probability density fitting using maximum likelihood estimation to output the physical distribution parameters of the target particles. Further, the step of extracting global image features from the preprocessed image and dynamically adaptively matching the target segmentation model based on the global image features includes: calculating the global gray-level gradient variance and local contrast of the preprocessed image; evaluating the edge curvature features and clustering overlap of particles in the image; when the clustering overlap is greater than a preset overlap threshold or the edge curvature features exhibit an irregular polygonal distribution, selecting the morphological watershed model as the target segmentation model; when the edge curvature features exhibit standard spherical symmetry and the clustering overlap is lower than the preset overlap threshold, selecting the Hough circle detection model as the target segmentation model and adaptively matching the edge detection gradient parameters and perfection parameters. Furthermore, the morphological features include the equivalent circle diameter based on projected area transformation, the true contour roundness, and the aspect ratio of the best-fit ellipse; the topological invariants include the Euler number characterizing the connectivity of micropores; and the step of fusing the morphological features and the topological invariants to construct a multidimensional feature matrix includes: globally normalizing the equivalent circle diameters of the extracted multiple independent scalar domains; and using the normalized equivalent circle diameter, the true contour roundness, the aspect ratio of the best-fit ellipse, and the Euler number as orthogonal feature dimensions to construct the multidimensional feature matrix used to characterize the physical properties of the microscopic system.Further, the step of performing multidimensional probability density fitting based on the multidimensional feature matrix using maximum likelihood estimation to output the physical distribution parameters of the target particles includes: extracting the equivalent circle diameter feature sequence from the multidimensional feature matrix; providing a log-normal distribution model, a Weiber distribution model, and a standard normal distribution model as candidate probability density functions; inputting the equivalent circle diameter feature sequence into the candidate probability density functions respectively, and calculating the corresponding shape parameters and scale parameters using the maximum likelihood estimation; and outputting the probability density distribution curve with the best fit and its corresponding physical distribution parameters. A second aspect of the present invention provides an adaptive analysis system for the morphology and topological features of microscopic particles, comprising: an image acquisition and preprocessing module for acquiring a microscopic image of a complex multiphase system to be analyzed, and performing contrast-limited adaptive histogram equalization and nonlocal mean denoising preprocessing; an adaptive matching module for extracting global image features of the preprocessed image, and dynamically adaptively matching a target segmentation model based on the global image features, wherein the target segmentation model is either a morphological watershed model or a Hough circle detection model; an image segmentation module for identifying and segmenting the preprocessed image using the matched target segmentation model to obtain the independent scalar domain of the target particles; a multidimensional feature extraction module for extracting morphological features of the independent scalar domain, calculating the topological invariants of the connected components of the target particles, and fusing the morphological features and the topological invariants to construct a multidimensional feature matrix; and a statistical fitting module for performing multidimensional probability density fitting based on the multidimensional feature matrix using maximum likelihood estimation, and outputting the physical distribution parameters of the target particles. The beneficial effects of this invention are as follows: by introducing an adaptive preprocessing and dynamic algorithm matching mechanism, the system can intelligently switch between the watershed model and the Hough transform according to the physical characteristics of the microscopic image, which greatly improves the robustness to complex adhesive systems; creatively integrating topological invariants such as Euler number with traditional geometric morphological features to construct a multidimensional feature matrix, which not only achieves high-precision extraction of physical particle size, but also provides rigorous data support for the macroscopic performance evaluation of porous materials and polymer network structures. Attached Figure Description

[0004] Figure 1 is a flowchart of an adaptive analysis method for microparticle morphology and topological features provided in an embodiment of the present invention; Figure 2 is a structural block diagram of an adaptive analysis system for microscopic particle morphology and topological features provided in an embodiment of the present invention. Detailed Implementation

[0005] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention. This invention provides an adaptive analysis method for the morphology and topological features of microscopic particles. The method mainly includes the following steps: Step S1: Acquire a microscopic image of the complex multiphase system to be analyzed, and perform contrast-limited adaptive histogram equalization (CLAHE) and non-local means (NL-Means) noise reduction preprocessing on the microscopic image. In this embodiment, the original SEM / TEM image is usually accompanied by an extremely low signal-to-noise ratio. CLAHE can significantly enhance the edge contours of the dark field image while limiting noise amplification, while combining it with non-local means noise reduction can perfectly preserve the physical hard boundaries of the particles while smoothing speckle white noise. Step S2: Extract the global image features of the preprocessed image, and dynamically adaptively match the target segmentation model based on the global image features. Specifically, the system calculates the global gray-level gradient variance and local contrast, and evaluates the degree of particle aggregation and edge curvature. When faced with systems prone to severe adhesion, such as polymer blends, where the overlap exceeds a preset threshold, the system intelligently matches a morphological watershed model. For standard spherical particles, such as emulsion polymerization products, the system switches to the Hough circle detection model and adaptively derives the optimal edge detection gradient parameters (P1) and perfection parameters (P2). Step S3: The preprocessed image is segmented using the matched target segmentation model to obtain the independent scalar domain of the target particles. If a watershed model is used, the system locates the topological center of the adhered particles by calculating Euclidean distance transformation and combining it with morphological opening and closing operations, automatically deducing and cutting out segmentation lines that closely match the actual physical boundaries. Step S4: The morphological features of the independent scalar domain are extracted, and the topological invariants of the connected domains of the target particles are calculated. The two are then fused to construct a multidimensional feature matrix. Specifically, the morphological features include the equivalent circle diameter, the true contour roundness, and the aspect ratio. The equivalent circle diameter and roundness are calculated using a formula. Simultaneously, the Euler number, representing the connectivity of spatial pores, is introduced, forming a multidimensional orthogonal feature matrix with the aforementioned morphological features to characterize the physical properties of the system. Step S5: Based on the multidimensional feature matrix, multidimensional probability density fitting is performed using maximum likelihood estimation. The system provides candidate models such as log-normal distribution, Weiber distribution, and standard normal distribution. The extracted feature sequences are input into the candidate functions, and the optimal shape and scale parameters are calculated through maximum likelihood estimation. Finally, the curve and physical distribution parameters with the best fit are output, providing accurate reference for powder failure analysis and asymmetric distribution assessment.Another embodiment of the present invention provides an adaptive analysis system for microscopic particle morphology and topological features, including: an image acquisition and preprocessing module, an adaptive matching module, an image segmentation module, a multidimensional feature extraction module, and a statistical fitting module. The specific functions of each module correspond one-to-one with the steps in the above method embodiments, and will not be repeated here. The above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An adaptive analysis method for microscopic particle morphology and topological characteristics, characterized in that, Includes the following steps: Acquire microscopic images of the complex multiphase system to be analyzed, and perform contrast-limited adaptive histogram equalization and nonlocal mean denoising preprocessing on the microscopic images. Global image features are extracted from the preprocessed image, and a target segmentation model is dynamically and adaptively matched based on the global image features. The target segmentation model is either a morphological watershed model or a Hough circle detection model. The preprocessed image is identified and segmented using the matched target segmentation model to obtain the independent scalar domain of the target particles; Extract the morphological features of the independent scalar domain and calculate the topological invariants of the connected domain of the target particle. Then, fuse the morphological features and the topological invariants to construct a multidimensional feature matrix. Based on the multidimensional feature matrix, multidimensional probability density fitting is performed using maximum likelihood estimation to output the physical distribution parameters of the target particles.

2. The adaptive analysis method for microparticle morphology and topological features according to claim 1, characterized in that, The step of extracting global image features from the preprocessed image and dynamically adaptively matching the target segmentation model based on the global image features includes: Calculate the global grayscale gradient variance and local contrast of the preprocessed image; Evaluate the edge curvature features and clustering overlap of particles in the image; When the degree of cluster overlap is greater than a preset overlap threshold or the edge curvature features present an irregular polygonal distribution, the morphological watershed model is selected as the target segmentation model. When the edge curvature feature exhibits standard spherical symmetry and the degree of clustering overlap is lower than the preset overlap threshold, the Hough circle detection model is selected as the target segmentation model and the edge detection gradient parameters and perfection parameters are adaptively matched.

3. The adaptive analysis method for microscopic particle morphology and topological features according to claim 1, characterized in that, The morphological features include the equivalent circle diameter based on projected area transformation, the true contour roundness, and the aspect ratio of the best-fit ellipse. The topological invariants include the Euler number characterizing the connectivity of micropores. The step of fusing the morphological features and the topological invariants to construct a multidimensional feature matrix includes: The equivalent circle diameters of the extracted independent scalar domains are globally normalized. The normalized equivalent circle diameter, the true contour roundness, the aspect ratio of the best-fit ellipse, and the Euler number are used as orthogonal feature dimensions to construct the multidimensional feature matrix used to characterize the physical properties of the microscopic system.

4. The adaptive analysis method for microparticle morphology and topological features according to claim 1, characterized in that, The step of performing multidimensional probability density fitting based on the multidimensional feature matrix and outputting the physical distribution parameters of the target particles using maximum likelihood estimation includes: Extract the equivalent circle diameter feature sequence from the multidimensional feature matrix; Log-normal distribution, Weiber distribution, and standard normal distribution are provided as candidate probability density functions; The equivalent circle diameter feature sequence is input into the candidate probability density function, and the corresponding shape parameters and scale parameters are calculated using the maximum likelihood estimation. Output the probability density distribution curve with the best fit and its corresponding physical distribution parameters.

5. An adaptive analysis system for microscopic particle morphology and topological characteristics, characterized in that, include: The image acquisition and preprocessing module is used to acquire microscopic images of the complex multiphase system to be analyzed, and to perform contrast-limited adaptive histogram equalization and nonlocal mean denoising preprocessing. An adaptive matching module is used to extract global image features of the preprocessed image and dynamically and adaptively match a target segmentation model based on the global image features. The target segmentation model is either a morphological watershed model or a Hough circle detection model. The image segmentation module is used to identify and segment the preprocessed image using the matched target segmentation model to obtain the independent scalar domain of the target particles; A multidimensional feature extraction module is used to extract morphological features of the independent scalar domain, calculate the topological invariants of the connected domain of the target particle, and fuse the morphological features and the topological invariants to construct a multidimensional feature matrix. The statistical fitting module is used to perform multidimensional probability density fitting based on the multidimensional feature matrix using maximum likelihood estimation, and output the physical distribution parameters of the target particles.

6. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the adaptive analysis method for microscopic particle morphology and topological features as described in any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the adaptive analysis method for microscopic particle morphology and topological features as described in any one of claims 1 to 4.

Images