Nano-silver bristle surface property detection system and method

Abstract

The application discloses a nano-silver bristle surface characteristic detection system and method, and belongs to the technical field of visual detection. The system collects and splices a full-surface image of the bristle through a linear array camera and a rotating carrier, extracts a gray-scale image to analyze uniformity, solves the problems of a curved surface detection blind area and a single index. The bristle with qualified uniformity enters a depth detection. Fluorescent markers capture frictional and fallen particles to construct a distribution image sequence. In combination with a roundness threshold and similarity analysis, damage imaging is triggered, damage areas are quantified to track coating wear evolution. In a stable wear stage, the bristle is immersed in a simulated solution containing a fluorescent probe. Fluorescent data slope is dynamically sampled and analyzed. Agglomerated particles are intelligently distinguished from monodisperse particles. The release process is accurately evaluated, aiming at solving the problems of incomplete uniformity evaluation, insufficient wear mechanism analysis and inaccurate release performance detection in the prior art.

Description

Technical Field

[0001] This invention relates to the field of visual inspection technology, and more specifically to a system and method for detecting the surface characteristics of nano-silver brush bristles. Background Technology

[0002] Due to its unique antibacterial properties, nano-silver is increasingly used in functional brush bristles, such as oral care brushes and medical cleaning brushes. The core performance of these products depends on the uniformity of the nano-silver coating distribution on the bristle surface, its wear resistance and durability, and the silver ion release behavior. Accurate detection of these characteristics is crucial to ensuring product quality and safety.

[0003] Currently, the detection of surface characteristics of nano-silver brush bristles faces multiple technical challenges: In terms of coating uniformity detection, the cylindrical curved surface structure of the bristles makes it difficult for traditional planar imaging techniques to achieve full surface coverage, often resulting in blind spots. Furthermore, the accuracy in identifying local defects such as nano-silver particle agglomeration and substrate exposure is insufficient, failing to meet the quantitative assessment requirements of coating uniformity. In the tribological testing stage, existing methods mostly employ static friction tests under fixed conditions, which are difficult to simulate the complex and variable load conditions in actual use. They also lack real-time tracking of nano-silver particle shedding behavior and coating damage evolution during friction, leading to a one-sided analysis of wear mechanisms. Regarding silver ion release performance detection, traditional methods typically use static immersion or simple oscillation to simulate the application environment, ignoring the influence of dynamic shear forces, electrolyte environments, and other factors on release behavior in real-world scenarios. This makes it impossible to accurately distinguish between sudden and stable release, resulting in discrepancies between the release amount detection results and the performance under actual use scenarios.

[0004] In summary, existing technologies for detecting the surface characteristics of nano-silver brush bristles suffer from problems such as incomplete detection coverage, distorted simulation of working conditions, limited mechanism analysis, and lack of data correlation. Therefore, in order to overcome these limitations, this invention proposes a nano-silver brush bristle surface characteristic detection system and method. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide a nano-silver brush bristle surface characteristic detection system and method. This system addresses the technical challenges of accurately performing initial quantitative detection of the uniformity of nano-silver distribution on the nano-silver brush bristle surface, simulating actual friction processes to detect coating wear, and differentially detecting the release of nano-silver particles during the wear stabilization stage. Through image acquisition and processing, multi-dimensional detection, and correlation analysis, a comprehensive and accurate evaluation of the brush bristle surface characteristics and functional performance can be achieved.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] The nano-silver brush bristle surface characteristic detection system includes:

[0008] Linear images covering the brush sample are acquired. Feature points of adjacent linear images are extracted and screened. After calculating the homography matrix, nano-silver particles are located by template matching and Gaussian fitting. The offset of adjacent linear images is obtained and stitched together to generate a full surface image of the brush sample. The main brush area of ​​the full surface image is extracted and converted into a grayscale image for initial quantitative detection of the uniformity of nano-silver distribution on the surface of the brush sample, and to determine whether the macroscopic uniformity of the grayscale image is qualified.

[0009] If the macroscopic uniformity is deemed unqualified, the severity of the defect is determined; otherwise, the macroscopic uniformity is deemed qualified, and the process proceeds to the in-depth testing stage, which includes friction loss testing and release testing.

[0010] Friction loss detection is used to simulate the friction process of using nano-silver brush bristles. Brush bristle samples with qualified macroscopic uniformity are subjected to friction detection. After a single friction, fluorescence imaging is performed on the nano-silver particle collection device to construct a sequence of nano-silver particle distribution images. By constructing a similarity analysis model of adjacent particle distribution images, the acquisition of surface damage images of the brush bristle samples is triggered to construct a damage image sequence. Damage areas are detected by the baseline image difference method, and the area ratio and spatial distribution of suspected damage areas are quantified to determine whether the wear of the brush bristle sample coating has entered a stable stage, at which point the friction loss detection is terminated.

[0011] The release detection module is used to immerse the brush bristle sample after friction loss detection into a simulated solution containing a fluorescent probe when the friction loss detection is terminated, collect fluorescence data of the sample solution, and detect the amount of silver nanoparticles released from the brush bristle sample in the simulated solution after friction loss detection.

[0012] Specifically, the steps for obtaining the offset of adjacent linear images include:

[0013] Adaptive histogram equalization is performed on adjacent linear images to extract the feature points of silver nanoparticles in the linear images, and the search direction of the feature points is limited to the circumferential direction perpendicular to the axis of the brush sample.

[0014] Configure spatial constraint thresholds and grayscale constraint thresholds, calculate the circumferential angle difference and grayscale similarity of adjacent linear image feature point pairs, filter out feature points with angle differences less than the spatial constraint threshold and grayscale similarity greater than the grayscale constraint threshold, and mark them as matching point pairs;

[0015] The homography matrix of the matching point pairs is calculated. Based on the homography matrix, a template matching combined with a Gaussian surface fitting algorithm is used to perform sub-pixel-level localization of the nano-silver particles in the overlapping area.

[0016] Based on the homography matrix, the corresponding regions of the overlapping area in adjacent linear images are located. For the located silver nanoparticles, the corresponding points in the corresponding regions are calculated using the homography matrix to obtain the offset of the sub-pixel coordinates of the corresponding points. The offset of the overlapping area of ​​adjacent linear images is calculated by taking the average value.

[0017] Specifically, the steps for sub-pixel-level localization of silver nanoparticles in the overlapping region using template matching combined with Gaussian surface fitting algorithm include:

[0018] Based on the homography matrix, a unified coordinate system is established for adjacent linear images, homography transformation is performed on matching point pairs, and the overlapping area of ​​adjacent linear images is calculated and divided into sub-regions.

[0019] For each sub-region, the normalized cross-correlation coefficient is calculated and matched with the nano-silver template library to screen candidate sub-regions. The center of the candidate sub-region is used as the candidate point to construct a candidate point set.

[0020] Set a neighborhood window, perform Gaussian surface fitting on each candidate point within its neighborhood window, configure a fitting threshold, and if the determination coefficient of the Gaussian surface fitting within the neighborhood window of a candidate point is less than the fitting threshold, mark it as an invalid candidate point and remove it from the candidate point set.

[0021] Otherwise, the extreme points of the Gaussian surface fitting are obtained as the sub-pixel coordinates of the silver nanoparticles at the candidate point.

[0022] Specifically, the steps for determining whether the macroscopic uniformity of a grayscale image is acceptable include:

[0023] The main brush bristle region was extracted from the full surface image of the brush bristle sample and converted into a grayscale image. The pixel grayscale value of the grayscale image is positively correlated with the distribution density of surface nano-silver particles.

[0024] Configure a variation threshold to measure the dispersion of the density distribution of silver nanoparticles and a contrast threshold to evaluate the spatial distribution regularity. Extract the gray mean and standard deviation of the gray image to calculate the gray variation coefficient. Extract the image contrast through the gray co-occurrence matrix.

[0025] If the grayscale coefficient of variation is greater than the variation threshold or the contrast is greater than the contrast threshold, the macroscopic uniformity is determined to be unqualified and the potential non-uniform region segmentation process is triggered to classify the severity of defects. Otherwise, the macroscopic uniformity is determined to be qualified and the process enters the in-depth detection stage.

[0026] Specifically, the steps for classifying the severity of defects include:

[0027] Defect localization was performed on grayscale images to identify exposed substrate areas and silver nanoparticle agglomerate areas. The coordinates of potential non-uniform areas and the defect type (agglomeration or absence) were recorded. The density of silver nanoparticles in potential non-uniform areas was calculated, and the standard deviation of particle spacing was calculated.

[0028] Assign region quantity weights based on the number of potential uneven regions, and establish a mapping relationship between the number of potential uneven regions and the region quantity weights;

[0029] Compare the density of silver nanoparticles in each region with the overall average density, and assign a density deviation coefficient based on the degree of deviation.

[0030] A mapping relationship between defect types and defect type weights is constructed based on the degree of impact of defect types on bristle performance;

[0031] A multidimensional evaluation model is established by multiplying the weights of the number of regions, the density deviation coefficient, and the defect type, quantifying the defect severity score, and classifying the severity of defects according to the mapping rules between the score and the severity level.

[0032] Specifically, the steps for detecting friction loss include:

[0033] Fluorescently labeled brush bristle samples that passed the macroscopic uniformity test were collected, and linear images of the fluorescently labeled brush bristle samples were stitched together to generate a baseline image of the brush bristle samples.

[0034] Set the friction conditions, start the friction test bench to rub the brush bristle sample, and at the same time place a nano silver particle collection device below the friction area to capture the detached nano silver particles.

[0035] After each brush sample was rubbed once, fluorescence imaging was performed on the nano-silver particle collection device to acquire images of the nano-silver particle distribution and construct a sequence of particle distribution images corresponding to the number of rubs.

[0036] The particle distribution image was binarized to obtain a nano-silver particle mask, and the internal cavities of the nano-silver particles were filled by morphological operations.

[0037] Configure a roundness threshold, label all silver nanoparticles through connected component analysis, obtain their centroid and area, calculate the roundness of each silver nanoparticle, and mark silver nanoparticles with a roundness greater than the roundness threshold as monodisperse particles, otherwise mark them as aggregated particles;

[0038] A similarity analysis model for adjacent particle distribution images was constructed. The differences in the number of monodisperse and aggregated particles and the mask deviation of nano-silver particles were extracted as feature parameters. The comprehensive similarity of adjacent particle distribution images was obtained by weighted averaging.

[0039] Specifically, the steps for detecting friction loss also include:

[0040] Configure a similarity threshold. When the overall similarity of adjacent particle distribution images is less than the similarity threshold, acquire a linear image of the brush bristle sample, stitch it together to generate a surface damage image of the brush bristle sample, and construct a surface damage image sequence corresponding to the number of friction cycles.

[0041] Damage region detection is performed on surface damage images. Based on the benchmark image difference method, a damage threshold is configured, and regions with gray value changes greater than the damage threshold are extracted as suspected damage regions. The suspected damage regions are connected through morphological operations, and the area ratio and spatial distribution of the suspected damage regions are quantified.

[0042] The number of cyclic frictions since the last surface damage image was acquired is counted. When the number of cyclic frictions exceeds the preset stable detection threshold, the wear of the brush sample coating is determined to have entered a stable stage and the friction loss detection is terminated.

[0043] Configure a limit number of frictions threshold and a limit area threshold. When the cumulative number of frictions exceeds the limit number of frictions threshold, or the area of ​​the suspected damage zone exceeds the limit area threshold, terminate the friction loss detection.

[0044] Specifically, the steps for conducting release testing include:

[0045] A simulated solution containing electrolytes and a fluorescent probe was obtained. The fluorescent probe specifically bound to the silver nanoparticles and generated a fluorescence signal positively correlated with the concentration of silver nanoparticles. The simulated solution was placed in a constant-temperature oscillation device, and the temperature, oscillation frequency, and oscillation amplitude were set.

[0046] Simulated solutions that were not immersed in the sample were collected as blank controls to establish fluorescence substrate data;

[0047] The brush bristle samples that have undergone friction loss testing are immersed in a simulated solution. A dynamic time series sampling strategy is adopted, with an initial sampling interval set and the sampling interval adaptively adjusted according to the changing trend of real-time fluorescence data.

[0048] Configure the rate of change threshold and adjust the step size, calculate the rate of change of fluorescence data in real time. When the rate of change of fluorescence data is less than the rate of change threshold, increase the acquisition interval by adjusting the step size; otherwise, maintain the current sampling interval.

[0049] Specifically, the steps for conducting release testing also include:

[0050] Configure a sliding window to perform segmented analysis on the collected fluorescence data and calculate the slope of the fluorescence data within the sliding window; based on preset steep slope thresholds and gradual slope thresholds, determine the release mode according to the slope: if the slope is greater than the steep slope threshold, it is determined to be a release range dominated by the shedding of aggregated particles; if the slope is less than or equal to the steep slope threshold and greater than the gradual slope threshold, it is determined to be a release range of the dissolution and shedding of monodisperse particles.

[0051] Configure a release stop threshold. When the slope of the fluorescence data in the sliding window is less than or equal to the gradually changing slope threshold, count the number of sliding windows with a slope less than or equal to the gradually changing slope threshold. If the number is greater than the release stop threshold, then stop the release detection.

[0052] A method for detecting the surface properties of nano-silver brush bristles includes the following steps:

[0053] Step S1: Acquire linear images covering the brush sample, extract and filter feature points of adjacent linear images, and calculate the homography matrix; locate the nano-silver particles through template matching and Gaussian fitting, obtain the offset of adjacent linear images, and then stitch them together to generate a full surface image of the brush sample.

[0054] Step S2: Extract the main area of ​​the brush bristles from the full surface image and convert it into a grayscale image. Perform initial quantitative detection on the uniformity of the nano-silver distribution on the surface of the brush bristle sample. Analyze the grayscale image to determine whether the macroscopic uniformity is qualified.

[0055] Step S3: If the macroscopic uniformity is determined to be unqualified, the severity of the defect is classified; if the macroscopic uniformity is qualified, the in-depth inspection stage is entered, including friction loss detection and release detection.

[0056] Step S3.1: Simulate the friction process of using nano-silver brush bristles and perform friction detection on brush bristle samples with qualified macroscopic uniformity; after a single friction, perform fluorescence imaging on the nano-silver particle collection device to construct a nano-silver particle distribution image sequence; by constructing a similarity analysis model of adjacent particle distribution images, trigger the acquisition of surface damage images of brush bristle samples to construct a damage image sequence; use the benchmark image difference method to detect the damage area, quantify the area ratio and spatial distribution of suspected damage areas, determine whether the wear of the brush bristle sample coating has entered a stable stage, and terminate the friction loss detection;

[0057] Step S3.2: When the friction loss detection is terminated, immerse the brush bristle sample after friction into a simulated solution containing a fluorescent probe, collect the fluorescence data of the sample solution, and detect the amount of silver nanoparticles released from the brush bristle sample in the simulated solution after friction.

[0058] The beneficial effects of this invention are:

[0059] This application solves the problems of blind spots and low-contrast particle identification in the detection of nano-silver distribution on cylindrical brush bristle surfaces by using linear image stitching and sub-pixel level positioning technology, achieving quantitative assessment of uniformity of full surface coverage and significantly improving defect positioning accuracy; by using fluorescent labeling and particle morphology classification, it dynamically tracks the shedding mode of nano-silver particles during friction, and combines a similarity analysis model to accurately trigger damage image acquisition, realizing quantitative differentiation between two wear mechanisms: gradual dissolution of monodisperse particles and blocky peeling of aggregates;

[0060] By employing dynamic time-series sampling and sliding window slope analysis, this method balances rapid release capture with stable-phase efficiency optimization. It intelligently determines burst release dominated by the shedding of aggregated particles and stable release dominated by the dissolution of monodisperse particles, avoiding the confusion caused by traditional detection mechanisms. This provides a cross-scale, precise detection solution for the quality control and process optimization of nano-silver brush bristles. Attached Figure Description

[0061] Figure 1 This is a schematic diagram of the structure of the nano-silver brush bristle surface characteristic detection system of the present invention;

[0062] Figure 2 This is a flowchart illustrating the specific steps involved in obtaining the offset of the overlapping region of adjacent linear images according to the present invention.

[0063] Figure 3 A flowchart illustrating the specific steps involved in friction loss detection according to the present invention;

[0064] Figure 4 A flowchart illustrating the specific steps involved in release detection according to this invention;

[0065] Figure 5 This is a flowchart of the method for detecting the surface characteristics of nano-silver brush bristles according to the present invention. Detailed Implementation

[0066] Example 1

[0067] Please see Figure 1 This embodiment introduces a nano-silver brush bristle surface characteristic detection system, including a sample pretreatment module, a surface detection module, a friction loss module, a release detection module, and a correlation analysis module;

[0068] The sample pretreatment module is used to cut the nano-silver brush bristles to be tested into brush bristle samples of a predetermined length. The brush bristle samples are fixed by a positioning carrier so that the axis of the brush bristle samples is parallel to the reference direction of the carrier. The residual impurities and moisture on the surface of the brush bristle samples are removed by ultrasonic cleaning and nitrogen purging combined with constant temperature drying, so as to avoid surface contaminants from interfering with subsequent detection and ensure that the surface condition of the brush bristle samples meets the detection requirements.

[0069] In this embodiment, brush bristles are randomly selected from the batch of nano-silver brush bristles to be tested, and each bristle is cut into a predetermined length using micro-scissors. The cut bristle samples are then fixed onto a custom-designed quartz glass rotating carrier, which has a built-in rotating shaft and positioning slots. A micromanipulator is used to embed the bristle samples into the slots, ensuring that the axial direction of the bristle samples is completely aligned with the rotation axis of the carrier. The fixed bristle samples undergo surface cleaning: the carrier is placed in a container of deionized water and ultrasonically treated in an ultrasonic cleaner to remove residual dispersant and dust particles from the surface. After ultrasonic treatment, the surface of the bristle samples is rinsed with deionized water, then purged with nitrogen to remove surface moisture, and finally, the carrier is placed in a constant-temperature drying oven to dry. This method achieves standardized preparation and efficient cleaning of nano-silver brush bristle samples, providing qualified samples with consistent condition and no surface contamination for subsequent testing, ensuring the data reliability and repeatability of the testing system.

[0070] The surface detection module is used to perform initial quantitative detection of the uniformity of nano-silver distribution on the surface of nano-silver brush bristle samples based on industrial vision technology and electron microscopy analysis technology. Specifically, it includes: acquiring full-surface images of the fixed brush bristle sample using a linear array camera with a telecentric lens and a uniform light source, combined with a rotating carrier; generating a grayscale image of the brush bristle sample surface after image preprocessing; and calculating the grayscale variation coefficient to make a preliminary judgment on the macroscopic uniformity of the brush bristle sample.

[0071] Preferably, the specific steps for the initial quantitative detection of the uniformity of silver nanoparticle distribution on the surface of the brush sample include:

[0072] The brush sample is cylindrical and requires rotation to achieve full-surface scanning. The rotating carrier with the brush sample is placed at the detection station, the annular uniform light source is activated, and the uniformity of illumination is adjusted by the integrating sphere calibration device. The focal length of the telecentric lens is adjusted according to the brush diameter parameter. The annular uniform light source combined with the integrating sphere calibration can eliminate imaging deviations caused by uneven illumination and ensure image grayscale consistency. Adjusting the focal length of the telecentric lens according to the brush diameter avoids edge blurring caused by changes in depth of field and ensures uniform resolution across the entire surface.

[0073] The rotation speed of the rotating carrier and the frame rate of the line scan camera are set. During the uniform rotation of the brush sample, the line scan camera synchronously acquires linear images, ensuring that the image strips acquired in each rotation cycle can cover the entire circumferential area of ​​the brush. Uniform rotation and synchronous acquisition ensure that the image strips are continuous and without overlap or gaps in the circumferential direction, which meets the geometric requirements of the cylindrical surface unfolding.

[0074] When acquiring linear images, the offset of the overlapping area of ​​adjacent linear images is obtained in real time. Subpixel-level stitching technology based on feature matching is used to stitch the linear images without distortion. At the same time, combined with the cylindrical surface unfolding algorithm, the annular image data is mapped into a planar unfolded image, and finally the full surface image of the brush sample is generated.

[0075] Please see Figure 2 Specifically, the steps for obtaining the offset of the overlapping region of adjacent linear images include:

[0076] The grayscale contrast between the silver nanoparticles and the bristle substrate may be low, necessitating image feature enhancement to improve subsequent matching accuracy. Adaptive histogram equalization is applied to adjacent linear images to enhance the grayscale contrast between the silver nanoparticles and the bristle substrate, thereby improving image feature clarity. Feature extraction is used to obtain the silver nanoparticle feature points in the linear images, with the feature point search direction limited to a circumferential direction perpendicular to the bristle sample axis to reduce invalid feature points.

[0077] Directly matching all feature points can easily lead to false matches, such as noise points or brush-like textures. Therefore, it's necessary to filter true matching point pairs through spatial and gray-level constraints. Feature points in adjacent linear images are matched by configuring spatial and gray-level constraint thresholds. The circumferential angle difference between feature point pairs in adjacent linear images and the gray-level similarity of the regions containing the feature point pairs are calculated. Gray-level similarity can be quantified by calculating parameters such as the mean difference and variance of gray levels. Feature point pairs with an angle difference less than the spatial constraint threshold and a gray-level similarity greater than the gray-level constraint threshold are selected as matching point pairs.

[0078] The brush bristle surface is cylindrical. To accurately locate overlapping areas, a homography matrix is ​​needed to establish the spatial transformation relationship between adjacent images. The homography matrix of matching point pairs can be obtained using the RANSAC algorithm. Based on this matrix, a template matching algorithm combined with Gaussian surface fitting is used to perform sub-pixel-level localization of the silver nanoparticles within the overlapping area.

[0079] Based on the homography matrix, a unified coordinate system is established for adjacent linear images, homography transformation is performed on matching point pairs, the overlapping area of ​​adjacent linear images is calculated, and the area is divided into sub-regions.

[0080] For each sub-region, the normalized cross-correlation coefficient is calculated and matched with the nanosilver template library to screen candidate sub-regions. The center of the candidate sub-region is used as the candidate point to construct a candidate point set. The nanosilver template library is configured as follows: based on the characteristic absorption peaks of nanosilver particles, nanosilver templates of different sizes and shapes are constructed. Each nanosilver template is enhanced with Gaussian filtering and Laplacian operator to enhance edge features.

[0081] Set a neighborhood window. For each candidate point, perform Gaussian surface fitting within its neighborhood window and configure a fitting threshold. If the determination coefficient of the Gaussian surface fitting within the neighborhood window of the candidate point is less than the fitting threshold, it is marked as an invalid candidate point and removed from the candidate point set. Otherwise, obtain the extreme point of the Gaussian surface fitting as the sub-pixel-level center coordinates of the nano-silver particles at that candidate point.

[0082] Based on the homography matrix, the corresponding regions R1 and R2 of the overlapping region R in adjacent linear images are located. For each nano-silver particle with a determined sub-pixel center coordinate, its corresponding points r1 and r2 in the corresponding regions R1 and R2 are calculated using the homography matrix. The offsets of the sub-pixel coordinates d1 and d2 of the corresponding points r1 and r2 are obtained. The offset of the overlapping region of adjacent linear images is calculated by taking the average value, providing a high-precision data basis for uniformity detection.

[0083] A threshold segmentation algorithm is used to extract the main bristle region from the full surface image of the bristle sample as the target region to remove background interference, focus on the main bristle region, and eliminate background interference. This ensures that the subsequent statistical grayscale values ​​accurately reflect the distribution density of nano-silver, simplifying the complexity of uniformity quantification analysis. The segmented full surface image of the bristle sample is converted into a grayscale image, where the pixel grayscale value of the grayscale image is positively correlated with the distribution density of nano-silver particles on the surface.

[0084] A single gray-level mean cannot comprehensively assess uniformity; it is necessary to combine density dispersion and spatial regularity for a comprehensive judgment. Uniformity thresholds are configured, including a variation threshold and a contrast threshold. The variation threshold measures the dispersion of the density distribution of silver nanoparticles, while the contrast threshold assesses the spatial distribution regularity of silver nanoparticles. The gray-level mean and standard deviation of the gray-level image are extracted to obtain the gray-level variation coefficient. The contrast of the gray-level image is extracted by calculating the gray-level co-occurrence matrix to reflect the spatial distribution regularity of the silver nanoparticles. If the gray-level variation coefficient of the gray-level image is greater than the variation threshold or the contrast of the gray-level image is greater than the contrast threshold, the macroscopic uniformity is deemed unqualified, triggering the potential non-uniformity region segmentation process to classify the severity of defects. Otherwise, the macroscopic uniformity is deemed qualified, and the process proceeds to the deep detection stage.

[0085] For brush samples that failed the macroscopic evaluation, a target detection algorithm was used to locate defects in the grayscale image, identifying the exposed substrate area and the nano-silver agglomerate area as potential non-uniform regions. The coordinates and defect types of the potential non-uniform regions, including agglomeration and missing parts, were recorded. For each potential non-uniform region, the nano-silver particle density of the potential non-uniform region was statistically analyzed, and the standard deviation of the nano-silver particle spacing was calculated.

[0086] Based on the number of potentially uneven regions, a weight level is assigned to the number of regions, and a mapping relationship is established between the number of potentially uneven regions and the weight of the region. The more potentially uneven regions there are, the higher the weight of the region.

[0087] By comparing the density of silver nanoparticles in each potential non-uniform region with the overall average density, a density deviation coefficient is assigned based on the degree of deviation, thereby achieving a quantitative assessment of the degree of local density anomaly. The greater the degree of deviation, the higher the density deviation coefficient.

[0088] Based on the defect types in the potential non-uniform regions, a mapping relationship between defect types and defect type weights was constructed. The more severe the influence of a defect type, the higher its weight. This relationship was obtained through experiments.

[0089] A multidimensional assessment model is established by combining the weights of regional quantity, density deviation coefficient, and defect type to quantify the defect severity score. By establishing a mapping rule between the defect severity score and the severity level, the severity of defects is classified. The multidimensional assessment model can be obtained by multiplying the weights of regional quantity, density deviation coefficient, and defect type.

[0090] The friction loss module is configured to simulate the friction process of nano-silver brush bristles in actual use under controllable working conditions and perform friction loss detection. It adopts a reciprocating friction test device and is equipped with a friction head with adjustable pressure, stroke and frequency. Its material and hardness simulate the actual contact medium. The brush bristle sample that has passed the macroscopic uniformity test is fixed at a position perpendicular to the friction direction to apply a periodic friction load. A nano-silver particle collection device is set below the friction area to capture the nano-silver particles that fall off during the friction process. After the friction is completed, the collection device is separated to save the fallen particles and the brush bristle surface is cleaned. The damaged area of ​​the brush bristle surface after friction is detected to realize the dynamic tracking of the wear process of the nano-silver coating on the brush bristle surface and the fallen particles.

[0091] Please see Figure 3 Preferably, the specific steps for friction loss detection include:

[0092] Fluorescent labeling enables the specific identification of silver nanoparticles. Brush samples that meet the macroscopic uniformity criteria are obtained and fluorescently labeled. The brush samples are then immersed in a probe solution, allowing the probe to fully bind with the silver nanoparticles on the brush surface. After immersion, the brush samples are cleaned and dried. They are then vertically fixed to the friction test fixture, with the friction direction perpendicular to the brush sample axis. This establishes a basis for the visual detection of silver nanoparticles, avoids uneven force due to fixing deviations, ensures consistent directionality of friction damage, and improves experimental repeatability.

[0093] Linear images of fluorescently labeled brush bristle samples were acquired and stitched together to generate a baseline image of the brush bristle samples. The initial state of the brush bristle surface, including the distribution of the nano-silver coating and the gray value of the substrate, was recorded to provide a comparative benchmark for subsequent damage detection.

[0094] Set the friction conditions, including friction pressure and friction frequency, start the friction test bench and sample the friction, and place a nano-silver particle collection device below the friction area. This device can be an electrostatic adsorption glass slide to capture the detached nano-silver particles.

[0095] After a single rubbing of the brush sample, fluorescence imaging was performed on the nano-silver particle collection device to acquire images of the nano-silver particle distribution and construct a particle distribution image sequence corresponding to the number of rubbing cycles. Imaging was performed immediately after a single rubbing cycle to avoid secondary contamination or loss of particles. The dynamic changes in particle detachment under different rubbing cycles were tracked through the time-series image sequence.

[0096] The particle distribution image is binarized to obtain a nano-silver particle mask, and morphological operations are used to fill the voids inside the nano-silver particles. The binarization process enhances the contrast between the nano-silver particles and the background, and the morphological operations eliminate image noise to ensure the integrity of the particle outline.

[0097] The shedding mechanisms of monodisperse particles and agglomerated particles are different. The former is a gradual dissolution, while the latter is a blocky peeling. Classification and labeling can distinguish different wear mechanisms. By configuring a roundness threshold, all nano-silver particles are labeled through connected region analysis to obtain their centroid and area. The roundness of each nano-silver particle is calculated, and nano-silver particles with a roundness threshold are screened and labeled as monodisperse particles, otherwise they are labeled as agglomerated particles.

[0098] To reduce the impact of random perturbations of silver nanoparticles on detection results, a similarity analysis model for adjacent particle distribution images is constructed. This model extracts the differences in the number of monodisperse and aggregated particles, as well as the mask deviation of silver nanoparticles, as feature parameters. The mask deviation of silver nanoparticles can be calculated based on the Dice coefficient, and a weighted average is used to obtain the comprehensive similarity of adjacent particle distribution images. Since the distribution of silver nanoparticles may be affected by random perturbations, the similarity is calculated by weighting the differences in the number of particles in adjacent images and the mask deviation, thus reducing the interference of random errors.

[0099] When the particle shedding pattern changes significantly, it indicates that the coating damage may be aggravated. A similarity threshold is configured. When the overall similarity of adjacent particle distribution images is less than the similarity threshold, surface damage image acquisition is triggered. That is, linear images of the brush sample are acquired and stitched together to generate surface damage images of the brush sample, and a surface damage image sequence corresponding to the number of frictions is constructed.

[0100] Damage regions are detected in surface damage images. Based on the difference method of the benchmark image, a damage threshold is configured, and regions with gray value changes greater than the damage threshold are extracted as suspected damage regions. Furthermore, morphological operations are used to connect the suspected damage regions, and the area ratio and spatial distribution of the suspected damage regions are quantified.

[0101] Otherwise, the number of cyclic frictions since the last surface damage image acquisition is counted, and the brush sample is stably tested. When the number of cyclic frictions exceeds the preset stable detection threshold, the wear of the brush sample coating is determined to have entered a stable stage, and the friction loss detection is terminated. This indicates that the wear has entered a stable period, and further detection has no significant value, thus avoiding waste of resources.

[0102] Otherwise, the cumulative number of friction cycles is calculated, and a limit number of cycles and a limit area threshold are configured. When the cumulative number of friction cycles exceeds the limit number of cycles threshold, or the proportion of the suspected damaged area exceeds the limit area threshold, the friction loss detection is terminated. Protect the experimental equipment and samples, ensuring that the detection terminates within a reasonable range, while also covering durability assessments for extreme usage scenarios.

[0103] The release detection module is used to detect the release amount of silver nanoparticles in the simulated solution of the brush bristle sample after friction loss detection is terminated. The brush bristle sample after friction is immersed in a fluorescent probe solution. The probe generates a fluorescence signal positively correlated with the silver ion concentration at a specific excitation wavelength. The release of silver nanoparticles is simulated under isothermal oscillation conditions. Samples are taken at predetermined time intervals and the fluorescence data of the solution is measured to achieve quantitative analysis of the release amount of silver nanoparticles.

[0104] Please see Figure 4 Preferably, the specific steps for performing release detection include:

[0105] A simulated solution was prepared to mimic the actual application environment of the silver nanoparticles in a brush. This solution contained an electrolyte and a fluorescent probe, which specifically bound to the silver nanoparticles and generated a fluorescence signal positively correlated with the silver nanoparticle concentration. The simulated solution was placed in a constant-temperature oscillation apparatus, with appropriate temperature, oscillation frequency, and amplitude set to ensure uniform stress on the brush sample within the solution, simulating the dynamic release environment in actual use.

[0106] A simulated solution that was not immersed in the sample was collected as a blank control to establish fluorescence substrate data, effectively eliminating background interference from probe fluorescence and solution impurities.

[0107] The brush bristle samples that have undergone friction loss testing are immersed in a simulated solution. A dynamic time series sampling strategy is adopted, with an initial sampling interval set to be small to capture the rapid release process. The sampling interval is then adaptively adjusted according to the changing trend of real-time fluorescence data. To balance data integrity and detection efficiency, sampling is performed intensively when changes are drastic and reduced when the data is stable.

[0108] By calculating the rate of change of fluorescence data in real time, the dynamics of the release process are quantified. A rate of change threshold is configured, the rate of change of fluorescence data is acquired in real time, and an adjustment step size is set. When the rate of change is less than the rate of change threshold, the release process is determined to have entered a relatively stable stage. The acquisition interval is gradually increased according to the adjustment step size. Otherwise, the current sampling interval is maintained to capture signal details in sudden release events.

[0109] Each time fluorescence data is collected, the oscillation is paused to avoid data fluctuations caused by fluid disturbance. The supernatant is taken to measure the fluorescence data and the fluorescence substrate data is subtracted to obtain the fluorescence data.

[0110] Real-time trend analysis was performed on the collected fluorescence data to extract the temporal characteristics of the fluorescence data, so as to comprehensively determine the release mode and divide the agglomeration particle detachment interval into the monodisperse detachment interval. When the fluorescence data slope is steep, it is identified as a sudden release dominated by agglomeration particle detachment. If the fluorescence data shows a gradual upward trend and the damage rate is lower than the critical value, it is identified as a stable release dominated by monodisperse particle dissolution.

[0111] Specifically, the steps for comprehensively determining the release mode include:

[0112] The slope of fluorescence data is calculated in segments using a sliding window, which smooths short-term fluctuations and highlights trend changes. By configuring a sliding window, the slope of fluorescence data is calculated within the sliding window, and the slope reflects the rate of change of fluorescence data per unit time.

[0113] Based on the preliminary experimental data, steep slope threshold and gradual slope threshold were set. In the real-time trend analysis, if the slope of the fluorescence data in the sliding window is greater than the steep slope threshold, it is determined to be the agglomeration particle shedding interval dominated by agglomeration particle shedding. At this time, the release signal is mainly caused by the shedding of agglomerates at the coating microcracks or interface peeling.

[0114] If the slope of the fluorescence data in the sliding window is less than or equal to the steep slope threshold, and the slope of the fluorescence data in the sliding window is greater than the gradual slope threshold, then the monodisperse detachment interval of the monodisperse particles dissolving and detaching is identified.

[0115] Based on the release mode determination, a release stop monitoring mechanism is set up, configuring a release stop threshold. If the slope of the fluorescence data in the sliding window is less than or equal to the gradually changing slope threshold, release detection monitoring is stopped. This involves counting the number of sliding windows with slopes less than or equal to the gradually changing slope threshold; if the number exceeds the release stop threshold, release detection is stopped. Through precise control of the simulated environment, dynamic sampling strategies, and intelligent recognition of release modes, differentiated detection of the release behavior of nano-silver bristles in practical applications is achieved. This effectively distinguishes between the sudden release of aggregates and the stable dissolution of monodisperse particles, providing targeted data support for coating durability assessment and process optimization.

[0116] The correlation analysis module is configured to deeply integrate and analyze the multi-dimensional data output from the sample preprocessing module, surface detection module, friction loss module, and release detection module. By constructing a brush bristle sample database, it achieves cross-module indexing, correlation, and standardization of basic information, surface uniformity parameters, friction conditions, damage evolution data, and release performance indicators of brush bristle samples. It establishes an initial uniformity and friction durability prediction model, uses structural equation modeling to verify the causal path between coating damage degree, release mode, and release rate, and constructs a release kinetics correction model. Through principal component analysis and clustering algorithms, it constructs a three-dimensional quality assessment system including uniformity, durability, and release stability, classifies sample quality levels, and establishes a mapping relationship between detection results and manufacturing process parameters. This enables cross-scale correlation analysis from surface characteristic detection to functional performance evaluation, providing data-driven decision support for the quality control and process optimization of nano-silver brush bristles.

[0117] Preferably, the specific steps for deep fusion and correlation analysis include:

[0118] A unified database is constructed, which includes basic information of brush samples output by the sample preprocessing module, surface uniformity parameters generated by the surface detection module, including gray-scale coefficient of variation, nano-silver particle density, and spacing standard deviation; friction condition and damage evolution data recorded by the friction loss module, including damage area ratio and exposed area location; and fluorescence intensity data and release mode characteristics obtained by the release detection module.

[0119] The cross-module data is standardized and indexed using a data cleaning algorithm. Based on the macroscopic grayscale variation coefficient and microscopic particle density distribution of the surface detection module, combined with the particle shedding rate and damage propagation curve of the friction loss module, a partial least squares regression algorithm is used to construct an initial surface uniformity and friction durability prediction model.

[0120] By integrating the damage area ratio and damage location data from the friction loss module with the fluorescence data from the release detection module, the causal path of coating damage degree, release mode, and release rate is verified through structural equation modeling. The release modes include the shedding of agglomerated particles and the dissolution of monodisperse particles.

[0121] Principal component analysis was used to reduce the dimensionality of multi-dimensional parameters such as surface uniformity, wear degree, and release performance. Key feature vectors were extracted to construct a three-dimensional quality assessment system including uniformity index, durability index, and release stability coefficient. Clustering algorithm was used to classify the brush samples into different quality grades and establish a mapping relationship between the test results and the preparation process parameters of the nano-silver coating.

[0122] Example 2

[0123] Please see Figure 5This embodiment describes a method for detecting the surface characteristics of nano-silver brush bristles, including the following steps:

[0124] Step S1: Acquire linear images covering the brush sample, extract and filter feature points of adjacent linear images, and calculate the homography matrix; locate the nano-silver particles through template matching and Gaussian fitting, obtain the offset of adjacent linear images, and then stitch them together to generate a full surface image of the brush sample.

[0125] Step S2: Extract the main area of ​​the brush bristles from the full surface image and convert it into a grayscale image. Perform initial quantitative detection on the uniformity of the nano-silver distribution on the surface of the brush bristle sample. Analyze the grayscale image to determine whether the macroscopic uniformity is qualified.

[0126] Step S3: If the macroscopic uniformity is determined to be unqualified, the severity of the defect is classified; if the macroscopic uniformity is qualified, the in-depth inspection stage is entered, including friction loss detection and release detection.

[0127] Step S3.1: Simulate the friction process of using nano-silver brush bristles and perform friction detection on brush bristle samples with qualified macroscopic uniformity; after a single friction, perform fluorescence imaging on the nano-silver particle collection device to construct a nano-silver particle distribution image sequence; by constructing a similarity analysis model of adjacent particle distribution images, trigger the acquisition of surface damage images of brush bristle samples to construct a damage image sequence; use the benchmark image difference method to detect the damage area, quantify the area ratio and spatial distribution of suspected damage areas, determine whether the wear of the brush bristle sample coating has entered a stable stage, and terminate the friction loss detection;

[0128] Step S3.2: When the friction loss detection is terminated, immerse the brush bristle sample after friction into a simulated solution containing a fluorescent probe, collect the fluorescence data of the sample solution, and detect the amount of silver nanoparticles released from the brush bristle sample in the simulated solution after friction.

[0129] Preferably, the specific steps for obtaining the offset of adjacent linear images include:

[0130] Adaptive histogram equalization is performed on adjacent linear images to extract the feature points of silver nanoparticles in the linear images, and the search direction of the feature points is limited to the circumferential direction perpendicular to the axis of the brush sample.

[0131] Configure spatial constraint thresholds and grayscale constraint thresholds, calculate the circumferential angle difference and grayscale similarity of adjacent linear image feature point pairs, filter out feature points with angle differences less than the spatial constraint threshold and grayscale similarity greater than the grayscale constraint threshold, and mark them as matching point pairs;

[0132] The homography matrix of the matching point pairs is calculated. Based on the homography matrix, a template matching combined with a Gaussian surface fitting algorithm is used to perform sub-pixel-level localization of the nano-silver particles in the overlapping area.

[0133] Based on the homography matrix, the corresponding regions of the overlapping area in adjacent linear images are located. For the located silver nanoparticles, the corresponding points in the corresponding regions are calculated using the homography matrix to obtain the offset of the sub-pixel coordinates of the corresponding points. The offset of the overlapping area of ​​adjacent linear images is calculated by taking the average value.

[0134] Preferably, the specific steps for sub-pixel-level localization of silver nanoparticles within the overlapping region using a template matching combined with Gaussian surface fitting algorithm include:

[0135] Based on the homography matrix, a unified coordinate system is established for adjacent linear images, homography transformation is performed on matching point pairs, and the overlapping area of ​​adjacent linear images is calculated and divided into sub-regions.

[0136] For each sub-region, the normalized cross-correlation coefficient is calculated and matched with the nano-silver template library to screen candidate sub-regions. The center of the candidate sub-region is used as the candidate point to construct a candidate point set.

[0137] Set a neighborhood window, perform Gaussian surface fitting on each candidate point within its neighborhood window, configure a fitting threshold, and if the determination coefficient of the Gaussian surface fitting within the neighborhood window of a candidate point is less than the fitting threshold, mark it as an invalid candidate point and remove it from the candidate point set.

[0138] Otherwise, the extreme points of the Gaussian surface fitting are obtained as the sub-pixel coordinates of the silver nanoparticles at the candidate point.

[0139] Preferably, the specific steps for friction loss detection include:

[0140] Fluorescently labeled brush bristle samples that passed the macroscopic uniformity test were collected, and linear images of the fluorescently labeled brush bristle samples were stitched together to generate a baseline image of the brush bristle samples.

[0141] Set the friction conditions, start the friction test bench to rub the brush bristle sample, and at the same time place a nano silver particle collection device below the friction area to capture the detached nano silver particles.

[0142] After each brush sample was rubbed once, fluorescence imaging was performed on the nano-silver particle collection device to acquire images of the nano-silver particle distribution and construct a sequence of particle distribution images corresponding to the number of rubs.

[0143] The particle distribution image was binarized to obtain a nano-silver particle mask, and the internal cavities of the nano-silver particles were filled by morphological operations.

[0144] Configure a roundness threshold, label all silver nanoparticles through connected component analysis, obtain their centroid and area, calculate the roundness of each silver nanoparticle, and mark silver nanoparticles with a roundness greater than the roundness threshold as monodisperse particles, otherwise mark them as aggregated particles;

[0145] A similarity analysis model for adjacent particle distribution images was constructed, and the differences in the number of monodisperse and aggregated particles and the mask deviation of nano-silver particles were extracted as feature parameters. The comprehensive similarity of adjacent particle distribution images was obtained by weighted averaging.

[0146] Configure a similarity threshold. When the overall similarity of adjacent particle distribution images is less than the similarity threshold, acquire a linear image of the brush bristle sample, stitch it together to generate a surface damage image of the brush bristle sample, and construct a surface damage image sequence corresponding to the number of friction cycles.

[0147] Damage region detection is performed on surface damage images. Based on the benchmark image difference method, a damage threshold is configured, and regions with gray value changes greater than the damage threshold are extracted as suspected damage regions. The suspected damage regions are connected through morphological operations, and the area ratio and spatial distribution of the suspected damage regions are quantified.

[0148] The number of cyclic frictions since the last surface damage image was acquired is counted. When the number of cyclic frictions exceeds the preset stable detection threshold, the wear of the brush sample coating is determined to have entered a stable stage and the friction loss detection is terminated.

[0149] Configure a limit number of frictions threshold and a limit area threshold. When the cumulative number of frictions exceeds the limit number of frictions threshold, or the area of ​​the suspected damage zone exceeds the limit area threshold, terminate the friction loss detection.

[0150] Preferably, the specific steps for performing release detection include:

[0151] A simulated solution containing electrolytes and a fluorescent probe was obtained. The fluorescent probe specifically bound to the silver nanoparticles and generated a fluorescence signal positively correlated with the concentration of silver nanoparticles. The simulated solution was placed in a constant-temperature oscillation device, and the temperature, oscillation frequency, and oscillation amplitude were set.

[0152] Simulated solutions that were not immersed in the sample were collected as blank controls to establish fluorescence substrate data;

[0153] The brush bristle samples that have undergone friction loss testing are immersed in a simulated solution. A dynamic time series sampling strategy is adopted, with an initial sampling interval set and the sampling interval adaptively adjusted according to the changing trend of real-time fluorescence data.

[0154] Configure the rate of change threshold and adjust the step size, calculate the rate of change of fluorescence data in real time, and increase the acquisition interval by adjusting the step size when the rate of change of fluorescence data is less than the rate of change threshold; otherwise, maintain the current sampling interval.

[0155] Configure a sliding window to perform segmented analysis on the collected fluorescence data and calculate the slope of the fluorescence data within the sliding window; based on preset steep slope thresholds and gradual slope thresholds, determine the release mode according to the slope: if the slope is greater than the steep slope threshold, it is determined to be a release range dominated by the shedding of aggregated particles; if the slope is less than or equal to the steep slope threshold and greater than the gradual slope threshold, it is determined to be a release range of the dissolution and shedding of monodisperse particles.

[0156] Configure a release stop threshold. When the slope of the fluorescence data in the sliding window is less than or equal to the gradually changing slope threshold, count the number of sliding windows with a slope less than or equal to the gradually changing slope threshold. If the number is greater than the release stop threshold, then stop the release detection.

[0157] Working principle and its effects:

[0158] The nano-silver brush bristle surface characteristic detection system provided in this application achieves accurate detection of the nano-silver coating on the brush bristle surface throughout the entire process through multi-module collaboration, solving key problems in uniformity detection, wear analysis and release assessment of traditional technologies.

[0159] In the surface uniformity detection stage, a linear scan camera and a rotating carrier are used to acquire linear images covering the circumference of the brush bristles. Adaptive histogram equalization is used to enhance the contrast between the nano-silver particles and the substrate. Combined with feature point screening, homography matrix calculation, and sub-pixel-level positioning technology, the linear images are stitched together without distortion to form a full-surface unfolded map, achieving coverage detection of the cylindrical brush bristle surface. After extracting the main bristle region and converting it into a grayscale image, multidimensional analysis of grayscale variation coefficient and contrast is used to comprehensively evaluate the density dispersion and spatial regularity of the nano-silver particle distribution. This accurately identifies defects such as aggregation and missing particles and classifies their severity, avoiding the one-sidedness of evaluation by a single indicator.

[0160] In tribological testing, fluorescent labeling enables the specific identification of silver nanoparticles. Electrostatic adsorption on a glass slide efficiently captures friction-shedded particles, constructing a sequence of particle distribution images. A roundness threshold distinguishes between the detachment morphology of monodisperse and aggregated particles. Combined with a similarity analysis model of adjacent images, damage image acquisition is automatically triggered when the particle detachment pattern changes significantly. The area and distribution of the damaged region are quantified using a baseline image difference method. The correlation between coating damage evolution and particle detachment mechanisms during friction is dynamically tracked, accurately capturing key stages of wear.

[0161] The release detection process employs a dynamic time-series sampling strategy. Initial dense sampling captures the rapid release process, and subsequent intervals are adaptively adjusted based on fluorescence data trends to improve detection efficiency. By analyzing the slope of fluorescence data through a sliding window and combining it with preset thresholds, the system intelligently distinguishes between burst releases dominated by the shedding of aggregated particles and stable releases dominated by the dissolution of monodisperse particles. This avoids the general judgments of traditional detection methods and automatically stops detection when release stabilizes, thus optimizing the process.

[0162] Overall, it has overcome bottlenecks such as blind spots in the detection of complex curved surfaces, ambiguity in the distinction between multiple mechanisms, and lack of data correlation, providing a systematic solution for the quality control and process optimization of nano-silver brush bristles. It has significantly improved the comprehensiveness, accuracy, and intelligence of detection, and promoted the performance optimization and reliability assessment of functional brush bristle products.

[0163] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A nano-silver brush bristle surface characteristic detection system, characterized in that, include: Linear images covering the brush sample are acquired. Feature points of adjacent linear images are extracted and screened. After calculating the homography matrix, nano-silver particles are located by template matching and Gaussian fitting. The offset of adjacent linear images is obtained and stitched together to generate a full surface image of the brush sample. The main brush area of ​​the full surface image is extracted and converted into a grayscale image for initial quantitative detection of the uniformity of nano-silver distribution on the surface of the brush sample, and to determine whether the macroscopic uniformity of the grayscale image is qualified. If the macroscopic uniformity is deemed unqualified, the severity of the defect is determined; otherwise, the macroscopic uniformity is deemed qualified, and the process proceeds to the in-depth testing stage, which includes friction loss testing and release testing. The friction loss detection is used to simulate the friction process of using nano-silver brush bristles. Brush bristle samples with qualified macroscopic uniformity are subjected to friction detection. After a single friction, fluorescence imaging is performed on the nano-silver particle collection device to construct a nano-silver particle distribution image sequence. By constructing a similarity analysis model of adjacent particle distribution images, the acquisition of surface damage images of brush bristle samples is triggered to construct a damage image sequence. Damage areas are detected by the benchmark image difference method, and the area ratio and spatial distribution of suspected damage areas are quantified to determine whether the wear of the brush bristle sample coating has entered a stable stage, and the friction loss detection is terminated. The release detection is used to immerse the brush bristle sample after friction loss detection into a simulated solution containing a fluorescent probe when the friction loss detection is terminated, collect the fluorescence data of the sample solution, and detect the amount of silver nanoparticles released from the brush bristle sample in the simulated solution after friction loss detection. The specific steps for detecting friction loss include: Fluorescently labeled brush bristle samples that passed the macroscopic uniformity test were collected, and linear images of the fluorescently labeled brush bristle samples were stitched together to generate a baseline image of the brush bristle samples. Set the friction conditions, start the friction test bench to rub the brush bristle sample, and at the same time place a nano silver particle collection device below the friction area to capture the detached nano silver particles. After each brush sample was rubbed once, fluorescence imaging was performed on the nano-silver particle collection device to acquire images of the nano-silver particle distribution and construct a sequence of particle distribution images corresponding to the number of rubs. The particle distribution image was binarized to obtain a nano-silver particle mask, and the internal cavities of the nano-silver particles were filled by morphological operations. Configure a roundness threshold, label all silver nanoparticles through connected component analysis, obtain their centroid and area, calculate the roundness of each silver nanoparticle, and mark silver nanoparticles with a roundness greater than the roundness threshold as monodisperse particles, otherwise mark them as aggregated particles; A similarity analysis model for adjacent particle distribution images was constructed, and the differences in the number of monodisperse and aggregated particles and the mask deviation of nano-silver particles were extracted as feature parameters. The comprehensive similarity of adjacent particle distribution images was obtained by weighted averaging. Configure a similarity threshold. When the overall similarity of adjacent particle distribution images is less than the similarity threshold, acquire a linear image of the brush bristle sample, stitch it together to generate a surface damage image of the brush bristle sample, and construct a surface damage image sequence corresponding to the number of friction cycles. Damage region detection is performed on surface damage images. Based on the benchmark image difference method, a damage threshold is configured, and regions with gray value changes greater than the damage threshold are extracted as suspected damage regions. The suspected damage regions are connected through morphological operations, and the area ratio and spatial distribution of the suspected damage regions are quantified. The number of cyclic frictions since the last surface damage image was acquired is counted. When the number of cyclic frictions exceeds the preset stable detection threshold, the wear of the brush sample coating is determined to have entered a stable stage and the friction loss detection is terminated. Configure a limit number of frictions threshold and a limit area threshold. When the cumulative number of frictions exceeds the limit number of frictions threshold, or the area of ​​the suspected damage zone exceeds the limit area threshold, terminate the friction loss detection.

2. The nano-silver brush bristle surface characteristic detection system as described in claim 1, characterized in that, The specific steps for obtaining the offset of adjacent linear images include: Adaptive histogram equalization is performed on adjacent linear images to extract the feature points of silver nanoparticles in the linear images, and the search direction of the feature points is limited to the circumferential direction perpendicular to the axis of the brush sample. Configure spatial constraint thresholds and grayscale constraint thresholds, calculate the circumferential angle difference and grayscale similarity of adjacent linear image feature point pairs, filter out feature points with angle differences less than the spatial constraint threshold and grayscale similarity greater than the grayscale constraint threshold, and mark them as matching point pairs; The homography matrix of the matching point pairs is calculated. Based on the homography matrix, a template matching combined with a Gaussian surface fitting algorithm is used to perform sub-pixel-level localization of the nano-silver particles in the overlapping area. Based on the homography matrix, the corresponding regions of the overlapping area in adjacent linear images are located. For the located silver nanoparticles, the corresponding points in the corresponding regions are calculated using the homography matrix, and the offset of the sub-pixel coordinates of the corresponding points is obtained. The offset of the overlapping area of ​​adjacent linear images is calculated by taking the average value.

3. The nano-silver brush bristle surface characteristic detection system as described in claim 2, characterized in that, The specific steps for sub-pixel-level localization of silver nanoparticles within the overlapping region using a template matching combined with Gaussian surface fitting algorithm include: Based on the homography matrix, a unified coordinate system is established for adjacent linear images, homography transformation is performed on matching point pairs, and the overlapping area of ​​adjacent linear images is calculated and divided into sub-regions. For each sub-region, the normalized cross-correlation coefficient is calculated and matched with the nano-silver template library to screen candidate sub-regions. The center of each candidate sub-region is used as a candidate point to construct a candidate point set. Set a neighborhood window, perform Gaussian surface fitting on each candidate point within its neighborhood window, configure a fitting threshold, and if the determination coefficient of the Gaussian surface fitting within the neighborhood window of a candidate point is less than the fitting threshold, mark it as an invalid candidate point and remove it from the candidate point set. Otherwise, the extreme points of the Gaussian surface fitting are obtained as the sub-pixel coordinates of the silver nanoparticles at the candidate point.

4. The nano-silver brush bristle surface characteristic detection system as described in claim 1, characterized in that, The specific steps for determining whether the macroscopic uniformity of a grayscale image is satisfactory include: Extract the main brush bristle region from the full surface image of the brush bristle sample and convert it into a grayscale image. The pixel grayscale value of the grayscale image is positively correlated with the distribution density of surface nano-silver particles. Configure a variation threshold to measure the dispersion of the density distribution of silver nanoparticles and a contrast threshold to evaluate the spatial distribution regularity. Extract the gray mean and standard deviation of the gray image to calculate the gray variation coefficient. Extract the image contrast through the gray co-occurrence matrix. If the grayscale coefficient of variation is greater than the variation threshold or the contrast is greater than the contrast threshold, the macroscopic uniformity is determined to be unqualified and the potential non-uniform region segmentation process is triggered to classify the severity of defects. Otherwise, the macroscopic uniformity is determined to be qualified and the process enters the in-depth detection stage.

5. The nano-silver brush bristle surface characteristic detection system as described in claim 1, characterized in that, The specific steps for classifying the severity of defects include: Defect localization was performed on grayscale images to identify exposed substrate areas and silver nanoparticle agglomerate areas. The coordinates of potential non-uniform areas and the defect type (agglomeration or absence) were recorded. The density of silver nanoparticles in potential non-uniform areas was calculated, and the standard deviation of particle spacing was calculated. Assign region quantity weights based on the number of potential uneven regions, and establish a mapping relationship between the number of potential uneven regions and the region quantity weights; Compare the density of silver nanoparticles in each region with the overall average density, and assign a density deviation coefficient based on the degree of deviation. A mapping relationship between defect types and defect type weights is constructed based on the degree of impact of defect types on bristle performance; A multidimensional evaluation model is established by multiplying the weights of the number of regions, the density deviation coefficient, and the defect type to quantify the defect severity score. The severity of defects is then classified according to the mapping rules between the score and the severity level.

6. The nano-silver brush bristle surface characteristic detection system as described in claim 1, characterized in that, The specific steps for the release detection include: A simulated solution containing an electrolyte and a fluorescent probe is obtained, wherein the fluorescent probe specifically binds to the silver nanoparticles and generates a fluorescence signal positively correlated with the concentration of the silver nanoparticles; the simulated solution is placed in a constant temperature oscillation device, and the temperature, oscillation frequency, and oscillation amplitude are set. Simulated solutions that were not immersed in the sample were collected as blank controls to establish fluorescence substrate data; The brush bristle samples that have undergone friction loss testing are immersed in a simulated solution. A dynamic time series sampling strategy is adopted, with an initial sampling interval set and the sampling interval adaptively adjusted according to the changing trend of real-time fluorescence data. Configure the rate of change threshold and adjust the step size, calculate the rate of change of fluorescence data in real time. When the rate of change of fluorescence data is less than the rate of change threshold, increase the acquisition interval by adjusting the step size; otherwise, maintain the current sampling interval.

7. The nano-silver brush bristle surface characteristic detection system as described in claim 6, characterized in that, The specific steps of the release detection also include: Configure a sliding window to perform segmented analysis on the collected fluorescence data and calculate the slope of the fluorescence data within the sliding window; based on preset steep slope thresholds and gradual slope thresholds, determine the release mode according to the slope: if the slope is greater than the steep slope threshold, it is determined to be a release interval dominated by the shedding of aggregated particles; if the slope is less than or equal to the steep slope threshold and greater than the gradual slope threshold, it is determined to be a release interval of the dissolution and shedding of monodisperse particles. Configure a release stop threshold. When the slope of the fluorescence data in the sliding window is less than or equal to the gradually changing slope threshold, count the number of sliding windows with a slope less than or equal to the gradually changing slope threshold. If the number is greater than the release stop threshold, then stop the release detection.

8. A method for detecting the surface characteristics of nano-silver brush bristles, implemented based on the nano-silver brush bristle surface characteristic detection system according to any one of claims 1-7, characterized in that, Includes the following steps: Step S1: Acquire linear images covering the brush sample, extract and filter feature points of adjacent linear images, and calculate the homography matrix; locate the nano-silver particles through template matching and Gaussian fitting, obtain the offset of adjacent linear images, and then stitch them together to generate a full surface image of the brush sample. Step S2: Extract the main area of ​​the brush bristles from the full surface image and convert it into a grayscale image. Perform initial quantitative detection on the uniformity of the nano-silver distribution on the surface of the brush bristle sample. Analyze the grayscale image to determine whether the macroscopic uniformity is qualified. Step S3: If the macroscopic uniformity is determined to be unqualified, the severity of the defect is classified; if the macroscopic uniformity is qualified, the in-depth inspection stage is entered, including friction loss detection and release detection. Step S3.1: Simulate the friction process of using nano-silver brush bristles and perform friction detection on brush bristle samples with qualified macroscopic uniformity; after a single friction, perform fluorescence imaging on the nano-silver particle collection device to construct a nano-silver particle distribution image sequence; by constructing a similarity analysis model of adjacent particle distribution images, trigger the acquisition of surface damage images of brush bristle samples to construct a damage image sequence. Damaged areas are detected using the baseline image difference method, the area ratio and spatial distribution of suspected damaged areas are quantified, and it is determined whether the wear of the brush sample coating has entered a stable stage, thus terminating the friction loss detection. Step S3.2: When the friction loss detection is terminated, immerse the brush bristle sample after friction into a simulated solution containing a fluorescent probe, collect the fluorescence data of the sample solution, and detect the amount of silver nanoparticles released from the brush bristle sample in the simulated solution after friction. The specific steps for detecting friction loss include: Fluorescently labeled brush bristle samples that passed the macroscopic uniformity test were collected, and linear images of the fluorescently labeled brush bristle samples were stitched together to generate a baseline image of the brush bristle samples. Set the friction conditions, start the friction test bench to rub the brush bristle sample, and at the same time place a nano silver particle collection device below the friction area to capture the detached nano silver particles. After each brush sample was rubbed once, fluorescence imaging was performed on the nano-silver particle collection device to acquire images of the nano-silver particle distribution and construct a sequence of particle distribution images corresponding to the number of rubs. The particle distribution image was binarized to obtain a nano-silver particle mask, and the internal cavities of the nano-silver particles were filled by morphological operations. Configure a roundness threshold, label all silver nanoparticles through connected component analysis, obtain their centroid and area, calculate the roundness of each silver nanoparticle, and mark silver nanoparticles with a roundness greater than the roundness threshold as monodisperse particles, otherwise mark them as aggregated particles; A similarity analysis model for adjacent particle distribution images was constructed, and the differences in the number of monodisperse and aggregated particles and the mask deviation of nano-silver particles were extracted as feature parameters. The comprehensive similarity of adjacent particle distribution images was obtained by weighted averaging. Configure a similarity threshold. When the overall similarity of adjacent particle distribution images is less than the similarity threshold, acquire a linear image of the brush bristle sample, stitch it together to generate a surface damage image of the brush bristle sample, and construct a surface damage image sequence corresponding to the number of friction cycles. Damage region detection is performed on surface damage images. Based on the benchmark image difference method, a damage threshold is configured, and regions with gray value changes greater than the damage threshold are extracted as suspected damage regions. The suspected damage regions are connected through morphological operations, and the area ratio and spatial distribution of the suspected damage regions are quantified. The number of cyclic frictions since the last surface damage image was acquired is counted. When the number of cyclic frictions exceeds the preset stable detection threshold, the wear of the brush sample coating is determined to have entered a stable stage and the friction loss detection is terminated. Configure a limit number of frictions threshold and a limit area threshold. When the cumulative number of frictions exceeds the limit number of frictions threshold, or the area of ​​the suspected damage zone exceeds the limit area threshold, terminate the friction loss detection.

Images