Image processing-based granule uniformity detection method and system

Abstract

The application belongs to the technical field of image processing, and particularly relates to a particle uniformity detection method and system based on image processing, which comprises the following steps: S1, acquiring a to-be-detected image and performing background calibration processing, determining the centroid coordinates of each particle by recognizing the particle region in the to-be-detected image, and generating a particle centroid coordinate point set; S2, based on the particle centroid coordinate point set and a preset spatial reference constant, calculating the aggregation energy of the particles in a two-dimensional space, and determining a spatial attractive potential index reflecting the global structural congestion of the particles; S3, for each particle in the particle centroid coordinate point set, constructing an interface annular region within a preset range of the particle boundary. The application can effectively identify the hidden aggregation defects of the particles, accurately evaluate the physical quality of the particle dispersion, and has very high determination robustness.

Description

Technical Field

[0001] This invention belongs to the field of image processing technology, specifically relating to a method and system for detecting particle uniformity based on image processing. Background Technology

[0002] In the processing of rubber additives and pre-dispersed masterbatches, the distribution of pigments and functional additive particles in the polymer matrix is ​​crucial. The dispersion quality of these micro-particles directly determines the color stability, anti-aging strength, and mechanical properties of the final industrial products. Whether it's core industrial components like automotive tires and seals, or everyday rubber products, their service life and safety are closely related to this dispersion quality. To ensure high-standard production quality and meet the stringent performance requirements of downstream industries, high-precision microscopic inspection of the masterbatch cross-section is necessary to accurately assess whether the particles meet the required micron-level uniform dispersion standard, thus eliminating potential quality problems caused by uneven dispersion at the source.

[0003] However, most existing image detection methods rely on global statistical indicators such as grayscale variance or average color difference, resulting in a relatively single detection dimension and a lack of in-depth analytical capabilities regarding microstructure. These methods cannot identify the spatial topological relationships of particles, and struggle to distinguish between uniform particle distribution and localized agglomeration. While uniformly distributed lattices and locally clustered agglomerations may show similar global variance indicators, localized particle agglomeration can easily disrupt the continuity of the substrate, leading to stress concentration in the product and subsequently causing serious material failures such as cracking and accelerated aging. Furthermore, existing technologies struggle to accurately assess the concentration gradient of particles diffusing into the substrate and cannot effectively distinguish between mechanical stacking and true melt dispersion of particles. This results in a lack of crucial information such as the microscopic interface bonding state and dispersion uniformity, failing to comprehensively reflect the quality of masterbatch processing.

[0004] Furthermore, traditional quality assessment methods often employ linear weighting models, resulting in a slow response to localized severe defects, making it difficult to meet the quality requirements of zero major defects in industrial production. Since product quality often has a veto-limit effect, even a small percentage of localized severe defects in the overall average can significantly reduce the mechanical strength and stability of the product, ultimately leading to product failure, economic losses, and safety risks. Therefore, there is an urgent need for a testing scheme that can deeply assess the quality of spatial structure and interface integration, filling the gaps in existing technologies and effectively ensuring the production quality of rubber additives and pre-dispersed masterbatches. Summary of the Invention

[0005] This invention provides a particle uniformity detection method and system based on image processing to solve the technical problems of difficulty in identifying particle spatial topological relationships, inability to effectively evaluate the fusion quality of micro-interfaces, and insufficient sensitivity of linear evaluation results.

[0006] In a first aspect, the present invention provides a particle uniformity detection method based on image processing, comprising the following steps: S1. Acquire the image to be detected and perform background calibration processing. Determine the centroid coordinates of each particle by identifying the particle region in the image to be detected, so as to generate a set of particle centroid coordinate points. S2, based on the set of particle centroid coordinate points and a preset spatial reference constant, calculates the particle’s aggregation energy in two-dimensional space and determines the spatial gravitational potential index that reflects the particle’s global structural crowding. S3, for each particle in the set of centroid coordinate points, construct an interface annular region within a preset range of particle boundaries, and calculate the interface diffusion flux coefficient reflecting the fusion quality between the particle and the substrate based on the activity and abruptness of grayscale changes within the interface annular region. S4 nonlinearly couples the spatial gravitational potential index with the interface diffusion flux coefficient to determine the global uniformity score, thereby evaluating the particle uniformity detection results based on image processing.

[0007] Its effects are as follows: addressing the pain points of existing image detection methods that rely solely on global grayscale variance, making it difficult to distinguish between uniform particle lattices and local implicit agglomerations, and failing to assess the interface fusion state, this method introduces physical potential field theory to calculate spatial gravitational potential index and combines it with particle edge features to calculate interface diffusion flux coefficient. This breaks through the limitations of traditional single statistical indicators and achieves dual accurate quantification of particle topological crowding structure and actual melt wetting quality in the processing of polymer materials such as rubber additives. This significantly improves the sensitivity to identify local fatal agglomeration defects and the robustness of quality judgment.

[0008] Furthermore, based on the set of particle centroid coordinates and a preset spatial reference constant, the aggregation energy of the particles in two-dimensional space is calculated, and spatial gravitational potential indices reflecting the global structural crowding of the particles are determined, including: Using the formula:

[0009] Calculate the gravitational potential index of space ; in, This represents the total number of valid particles in the set of particle centroid coordinate points. Indicates the sequence number of the current target particle. This indicates the sequence number of the remaining particles besides the target particle. Indicates the first The particle and the first The straight-line distance between particles This represents a spatial reference constant preset according to standard process parameters. This represents the natural logarithm function.

[0010] Its effect is that, through specific logarithmic and distance fraction formulas, the abstract spatial topological distribution of particles is transformed into a quantifiable gravitational potential energy superposition process. Compared with simply calculating physical straight-line distance, this mathematical model can nonlinearly amplify the gravitational interaction between particles that are extremely close together. Even under deceptive conditions where the global variance is consistent, it can keenly and accurately distinguish high-risk local clustering patterns, providing solid physical algorithm support for identifying potential stress concentration hazards in materials.

[0011] Furthermore, based on the activity and abruptness of grayscale changes within the annular region of the interface, the interfacial dispersion flux coefficient, reflecting the fusion quality between particles and the substrate, is calculated, including: Using the formula:

[0012] Calculate the interfacial diffusion flux coefficient ; in, Indicates the particle number. Indicates the first The sum of the gradient magnitudes of all pixels within the annular region of the interface corresponding to each particle. Indicates the first The standard deviation of pixel grayscale within the annular region of the interface corresponding to each particle Indicates a fixed anti-interference constant. This represents the total number of valid particles in the set of particle centroid coordinate points.

[0013] Its effect is as follows: by using a formula that includes the ratio of gradient modulus and grayscale standard deviation, it quantifies the smoothness and abruptness of the color transition at the particle edge, and can accurately distinguish whether the micro particles in the substrate are in a good physical melting and diffusion state or are just a simple mechanical dead stacking. It effectively filters out background noise interference, objectively restores the real micro-wetting state in the industrial extrusion shearing process, and fills the gap in traditional detection methods that cannot evaluate the interfacial bonding force.

[0014] Furthermore, the space gravitational potential index is nonlinearly coupled with the interface dispersion flux coefficient to determine the global uniformity score, including: Using the formula:

[0015] Calculate the global uniformity score ; in, Indices representing the gravitational potential in space. This represents the interfacial diffusion flux coefficient. Indicates the sensitivity adjustment factor. Represents the hyperbolic tangent function. This indicates the range coefficient.

[0016] Its effects are as follows: by using the hyperbolic tangent function to perform nonlinear coupling calculation of spatial gravitational potential and interface diffusion flux, it breaks the drawback of the traditional linear weighted evaluation being slow to respond to local severe defects; when severe aggregation or mechanical stacking occurs locally, the ratio quickly enters the high slope range of the function, resulting in a sharp drop in the comprehensive score, which perfectly matches the quality control logic of local severe defects leading to the failure of the overall product in industrial production, and ensures the high efficiency of defect early warning.

[0017] Further, the image to be detected is acquired and background calibration is performed, including: The original image of the cross-section of the masterbatch was acquired using an industrial microscopic imaging system. The original image was then mapped to the logarithmic domain feature space for background compensation to decouple brightness and contrast and enhance the edge contrast between the particles and the carrier substrate.

[0018] Its effects are as follows: In response to the problem of uneven illumination caused by light source aging or slight vibration during microscopic imaging in industrial production sites, the brightness and contrast are decoupled by background compensation in the digital domain feature space, eliminating the interference of illumination gradient on image segmentation. It can stably enhance the edge contrast between the masterbatch and the rubber substrate under different lighting conditions, providing unified and high-quality underlying image data for subsequent high-precision coordinate extraction and microstructure analysis.

[0019] Further, the particle regions in the image to be detected are identified to determine the centroid coordinates of each particle, thereby generating a set of particle centroid coordinate points, including: An adaptive dynamic threshold segmentation algorithm is used to identify connected components in the image to be detected, calculate the centroid coordinates of each connected component, and remove noise particles with an area smaller than a preset pixel threshold.

[0020] Furthermore, when calculating the spatial straight-line distance between any two particles, a Euclidean distance matrix is ​​established based on the set of particle centroid coordinate points.

[0021] Furthermore, when constructing the annular area of ​​the interface, a closed annular region is formed by shifting the particle boundary outward by a preset pixel width.

[0022] Furthermore, the sensitivity adjustment factor is adapted to the warning threshold based on the material category of different masterbatch products.

[0023] Secondly, the present invention provides a particle uniformity detection system based on image processing, including a memory and a processor. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned particle uniformity detection method based on image processing is implemented.

[0024] The beneficial effects are: The core innovation of this invention lies in solving the topological identification problem of particle distribution through physical potential field theory. By treating each particle as a source of gravity and calculating the superposition of its potential energy in the entire domain, it fundamentally breaks through the bottleneck that traditional statistical indicators cannot perceive the spatial arrangement pattern of particles, enabling the system to accurately identify hidden agglomeration hotspots that have a huge impact on the strength of materials.

[0025] Meanwhile, this invention utilizes an interface dispersion flux model to evaluate the physical fusion depth of pigments and carriers. By extracting the microscopic gradient characteristics of the annular region at the particle edge, it effectively filters out background noise interference and objectively restores the wetting state of particles during the extrusion shearing process. This provides a scientific basis for in-depth evaluation of process parameter adjustments in industrial production and ensures the robustness of the detection system under complex working conditions. Attached Figure Description

[0026] Figure 1 This is a flowchart of the particle uniformity detection method based on image processing in this invention.

[0027] Figure 2 This is a schematic diagram of the real-time particle distribution extraction results in this invention.

[0028] Figure 3 This is a schematic diagram of the spatial gravitational potential field distribution cloud map of particles in this invention. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] An embodiment of the particle uniformity detection method based on image processing provided by the present invention: like Figure 1 As shown, the particle uniformity detection method based on image processing includes the following steps: S1. Acquire the image to be detected and perform background calibration processing. Determine the centroid coordinates of each particle by identifying the particle region in the image to be detected, so as to generate a set of particle centroid coordinate points.

[0031] In the specific implementation process, the original image of the cross-section of the masterbatch is first captured using an industrial microscopic imaging system. Since the light source in the production site may be aging or have slight vibrations that cause uneven lighting, this step maps the image to the logarithmic domain feature space. The luminance component and contrast component are decoupled through the logarithmic background compensation algorithm, which enhances the edge contrast between the pigment particles and the carrier substrate such as EPDM or styrene-butadiene rubber, and ensures the uniformity of image quality under different lighting conditions.

[0032] Subsequently, an adaptive dynamic thresholding segmentation technique is used to binarize the processed image, identifying all connected granular regions. The system traverses each connected component and calculates its centroid coordinates to generate a point set. ,in This represents the total number of valid particles. To eliminate interference from dust or image noise, the system automatically removes tiny dots with an area smaller than 3 pixels squared, ensuring the purity of the original data.

[0033] By performing background compensation and precise segmentation on the image, a high-confidence set of particle location points can be obtained, laying a reliable data foundation for subsequent topological structure evaluation.

[0034] S2, based on the set of particle centroid coordinate points and a preset spatial reference constant, calculates the aggregation energy of particles in two-dimensional space and determines the spatial gravitational potential index that reflects the structural crowding of particles across the entire domain.

[0035] This step introduces a physical gravitational potential field model, treating each particle as a source of outward radiating gravity. Aggregation regions are identified by calculating the energy superposition across the entire domain. The specific calculation formula is as follows:

[0036] in, Indices representing the gravitational potential in space. This represents the total number of valid particles in the set of particle centroid coordinate points. Indicates the sequence number of the current target particle. This indicates the sequence number of the remaining particles besides the target particle. Indicates the first The particle and the first The straight-line distance between particles This represents a spatial reference constant preset according to standard process parameters. This represents the natural logarithm function.

[0037] This formula, based on classical gravitational potential field theory, treats each particle as an independent gravitational source and equates the spatial interaction between particles to the superposition of gravitational potential energy. This achieves the engineering transformation of physical laws into particle distribution detection, with its core function being to quantify the degree of structural crowding across the entire two-dimensional space of particles. The formula first... Construct a process-adaptive distance response function, using the ideal distance threshold preset by the process. Matching actual particle spacing This allows the fractional value to change non-linearly with the degree of particle aggregation; then nesting... This approach achieves nonlinear amplification of localized high-aggregation gravitational effects while avoiding meaningless values, highlighting high-risk aggregation regions. Subsequently, a two-layer summation is used to complete the superposition of single-particle global interactions and the gravitational potentials of all particles globally, reflecting the overall crowding state. Finally, [the following is a continuation of the previous sentence, but the translation is incomplete]. Normalization is performed to eliminate the size interference of particle number on the indicators, ensuring comparability across detection fields. The formula, through a dual nonlinear design using both fractions and logarithms, solves the problem of local defect masking by traditional linear statistics, and each module has been designed for numerical stability. The adjustability also allows the formula to be adapted to different industrial testing conditions, while giving the indicators clear physical interpretability, realizing the upgrade from simple distance statistics to the energy of aggregation potential.

[0038] To better understand the formula, here's a calculation example. Suppose there are two particles in the detection field of view. The straight-line distance between the two Pixels, preset process reference distance Pixels. First, calculate the gravitational force on particle 1: the internal summation term is... Next, take the natural logarithm. Similarly, the term corresponding to particle 2 is also... Finally, the average value was calculated. If the distance between two particles is reduced to 1 pixel, the fraction term approaches 1, reflecting an increase in the degree of aggregation.

[0039] By calculating the spatial gravitational potential index, the risk of particle aggregation in space can be nonlinearly assessed, and distribution patterns can be accurately distinguished even when the global variance is uniform.

[0040] S3. For each particle in the set of particle centroid coordinate points, construct an interface annular region within a preset range of particle boundaries. Calculate the interface diffusion flux coefficient, which reflects the fusion quality between the particle and the substrate, based on the activity and abruptness of grayscale changes within the interface annular region.

[0041] For the identified first Each particle is extended outward by 5 pixels from its boundary to construct a ring-shaped interface area surrounding the particle. Within this region, the system extracts the gradient magnitude and grayscale standard deviation of all pixels, calculated using the following formula:

[0042] in, This represents the interfacial diffusion flux coefficient. Indicates the particle number. Indicates the first The sum of the gradient magnitudes of all pixels within the annular region of the interface corresponding to each particle reflects the activity level of color changes. Indicates the first The standard deviation of pixel grayscale within the annular region of the interface corresponding to each particle reflects the abruptness of the change. This represents a fixed anti-interference constant.

[0043] The core of this formula is used to quantify the interfacial fusion quality between particles and the substrate, distinguishing between molten diffusion and mechanical stacking of particles, and providing a quantitative basis for particle uniformity detection at the interface dimension. Its construction logic integrates practical engineering needs with mathematical adaptation design, balancing accuracy and stability. The formula takes the effective particles after noise removal within the detection field as the research object, where N is the total number of effective particles. It uses single-layer summation to traverse all particles, achieving a comprehensive characterization of the interface state across the entire domain. Among these, Used to quantify the smoothness of color transition at particle edges, the larger the cumulative gradient magnitude, the more severe the transition at the particle edges and the worse the interface blending. Used to assist in characterizing the grayscale uniformity of the interface region, in conjunction with Improve the accuracy of interface state assessment. The core function of the preset anti-interference constant is to filter out the interference of background noise on the calculation results and avoid interference caused by background noise. Too small a value can cause abnormal fluctuations in the fractional value, thus ensuring numerical stability under complex industrial conditions. The entire formula, through a concise fractional design, transforms the microscopic features of the interface into quantifiable indicators, and achieves full coverage through iterative summation. This not only meets the assessment requirements of interface fusion quality in industrial testing, but also enhances the robustness of testing through anti-interference design, providing reliable interface dimension support for subsequent calculation of the full-domain uniformity score.

[0044] Here's a specific calculation example: Suppose we detect a particle, Within its annular interface region, due to the certain melting and diffusion between the particles and the substrate, the color transition is relatively smooth, and the cumulative sum of the gradient modulus values ​​is calculated. Standard deviation Substituting into the formula, we get If the particles are merely mechanically stacked, drastic color changes will occur at the boundaries, leading to a higher standard deviation. If it surges to 150, then The interface diffusion flux coefficient decreased significantly.

[0045] By calculating the interfacial dispersion flux coefficient, the physical wetting degree between particles and the substrate can be assessed, and quality defects caused by mechanical stacking can be effectively identified.

[0046] S4 nonlinearly couples the spatial gravitational potential index with the interface diffusion flux coefficient to determine the global uniformity score, thereby evaluating the particle uniformity detection results based on image processing.

[0047] The system deeply couples topological features with interface features to derive the final quality score, as shown in the following formula:

[0048] in, This represents the global uniformity score. Indices representing the gravitational potential in space. This represents the interfacial diffusion flux coefficient. This represents the sensitivity adjustment factor, used to adapt the threshold for different products. Represents the hyperbolic tangent function. This indicates the range coefficient.

[0049] The core of this formula nonlinearly couples the spatial distribution of particles with the interface fusion quality, outputting a standardized score of 0-100. This intuitively quantifies the uniformity of particles across the entire domain, aligning with the quality control logic of zero major defects in industry. Its construction logic balances quantitative accuracy, defect sensitivity, and industrial practicality. The formula uses 100 as the full score benchmark and achieves uniformity comparisons for different samples and different detection fields through standardized design. The core coupling part uses the hyperbolic tangent function tanh, whose range is limited to [-1, 1]. After nesting, the range of [1-tanh(·)] becomes [0, 2]. Multiplying by 100 completes score normalization, ensuring... It falls within a reasonable scoring range. (Molecular weight) With denominator The ratio of these two factors reflects the imbalance between the degree of aggregation and the quality of fusion. The high slope of the hyperbolic tangent function can rapidly amplify the ratio caused by severe local defects, making... The score drops sharply, solving the problem of slow response of traditional linear weighting to local defects. Ultimately, it achieves accurate quantification of uniformity in both aggregation and interface dimensions, providing an intuitive and scientific scoring basis for industrial process adjustment.

[0050] Here is a comprehensive calculation example, assuming the space gravitational potential index obtained from the aforementioned steps is... Interfacial diffusion flux coefficient Sensitivity adjustment factor ; Calculate the ratio Calculate its hyperbolic tangent value. Final score If severe aggregation occurs, the ratio will quickly enter the high slope region of the function, causing the score to drop sharply.

[0051] like Figure 2As shown, it can automatically analyze particle status, with colored dots marking the particle center and rectangular boxes automatically locking localized hotspots; as... Figure 3 The gravitational potential field distribution cloud map shown has warm-colored areas that reflect the energy peak of the index, indicating the specific location of the aggregation defects and providing a basis for quality backtracking.

[0052] An embodiment of the particle uniformity detection system based on image processing provided by the present invention: The image processing-based particle uniformity detection system includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the image processing-based particle uniformity detection method described above.

[0053] The image processing-based particle uniformity detection system also includes other components well-known to those skilled in the art, such as communication interfaces. Their settings and functions are known in the art and will not be described in detail here.

[0054] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented using computer-readable / executable instructions stored or otherwise maintained by such a computer-readable medium.

[0055] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A particle uniformity detection method based on image processing, characterized in that, Includes the following steps: S1. Acquire the image to be detected and perform background calibration processing. Determine the centroid coordinates of each particle by identifying the particle region in the image to be detected, so as to generate a set of particle centroid coordinate points. S2, based on the set of particle centroid coordinate points and a preset spatial reference constant, calculates the particle’s aggregation energy in two-dimensional space and determines the spatial gravitational potential index that reflects the particle’s global structural crowding. S3, for each particle in the set of centroid coordinate points, construct an interface annular region within a preset range of particle boundaries, and calculate the interface diffusion flux coefficient reflecting the fusion quality between the particle and the substrate based on the activity and abruptness of grayscale changes within the interface annular region. S4 nonlinearly couples the spatial gravitational potential index with the interface diffusion flux coefficient to determine the global uniformity score, thereby evaluating the particle uniformity detection results based on image processing.

2. The particle uniformity detection method based on image processing according to claim 1, characterized in that, Based on the particle centroid coordinate set and a preset spatial reference constant, the aggregation energy of the particles in two-dimensional space is calculated, and spatial gravitational potential indices reflecting the global structural crowding of the particles are determined, including: Using the formula: Calculate the gravitational potential index of space ; in, This represents the total number of valid particles in the set of particle centroid coordinate points. Indicates the sequence number of the current target particle. This indicates the sequence number of the remaining particles besides the target particle. Indicates the first The particle and the first The straight-line distance between particles This represents a spatial reference constant preset according to standard process parameters. This represents the natural logarithm function.

3. The particle uniformity detection method based on image processing according to claim 1, characterized in that, Based on the activity and abruptness of grayscale changes within the annular region of the interface, the interfacial dispersion flux coefficient, reflecting the fusion quality between particles and the substrate, is calculated, including: Using the formula: Calculate the interfacial diffusion flux coefficient ; in, Indicates the particle number. Indicates the first The sum of the gradient magnitudes of all pixels within the annular region of the interface corresponding to each particle. Indicates the first The standard deviation of pixel grayscale within the annular region of the interface corresponding to each particle Indicates a fixed anti-interference constant. This represents the total number of valid particles in the set of particle centroid coordinate points.

4. The particle uniformity detection method based on image processing according to claim 1, characterized in that, Nonlinear coupling of space gravitational potential indices with interface dispersion flux coefficients determines the global uniformity score, including: Using the formula: Calculate the global uniformity score ; in, Indices representing the gravitational potential in space. This represents the interfacial diffusion flux coefficient. Indicates the sensitivity adjustment factor. Represents the hyperbolic tangent function. This indicates the range coefficient.

5. The particle uniformity detection method based on image processing according to claim 1, characterized in that, Acquire the image to be detected and perform background calibration processing, including: The original image of the cross-section of the masterbatch was acquired using an industrial microscopic imaging system. The original image was then mapped to the logarithmic domain feature space for background compensation to decouple brightness and contrast and enhance the edge contrast between the particles and the carrier substrate.

6. The particle uniformity detection method based on image processing according to claim 1, characterized in that, Identify particle regions in the image to be detected, determine the centroid coordinates of each particle, and generate a set of particle centroid coordinate points, including: An adaptive dynamic threshold segmentation algorithm is used to identify connected components in the image to be detected, calculate the centroid coordinates of each connected component, and remove noise particles with an area smaller than a preset pixel threshold.

7. The particle uniformity detection method based on image processing according to claim 2, characterized in that, When calculating the straight-line distance between any two particles, a Euclidean distance matrix is ​​established based on the set of coordinate points of the particle's centroid.

8. The particle uniformity detection method based on image processing according to claim 3, characterized in that, When constructing the ring-shaped area of ​​the interface, the ring-shaped closed area is formed by shifting the particle boundary outward by a preset pixel width.

9. The particle uniformity detection method based on image processing according to claim 4, characterized in that, The sensitivity adjustment factor is adapted to the warning threshold according to the material category of different masterbatch products.

10. A particle uniformity detection system based on image processing, characterized in that, It includes a memory and a processor, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the particle uniformity detection method based on image processing as described in any one of claims 1-9 is implemented.

Images