A method for detecting in the production of fertilizers

Abstract

This invention discloses a detection method for fertilizer production, relating to the field of fertilizer detection technology. The method includes the following steps: installing a near-infrared multispectral visual inspection system in the pre-packaging section of the fertilizer production line; adjusting the camera and illumination source positions to acquire multispectral images of fertilizer particles; sequentially performing dark current correction and flat field correction on the acquired images to eliminate noise and uneven illumination, followed by median filtering to remove random noise; selecting the band image with the best contrast, separating the particles from the background through threshold segmentation, and locating each individual particle after morphological boundary optimization; based on the multi-band reflectivity differences, analyzing the particle surface material pixel-by-pixel, statistically analyzing the coating area ratio of each particle, and obtaining the coating coverage rate of a single particle. This invention, by introducing near-infrared multispectral visual inspection technology and a quantitative analysis model for coating uniformity, enables non-destructive, high-precision online detection of fertilizer coating coverage.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of fertilizer detection, and particularly relates to a detection method for fertilizer production. BACKGROUND

[0002] Slow-release fertilizer is an advanced fertilizer that can delay the release rate of nutrients and match it with the growth demand of crops through specific technical processing. Its core principle is to use coating materials such as polymers to wrap the fertilizer particles, forming a semi-permeable or degradable film, thereby controlling the slow and stable release of internal nutrients (such as nitrogen, phosphorus, and potassium) to the soil through diffusion or osmosis. This controllable release feature aims to improve nutrient utilization, reduce losses and environmental pollution caused by leaching, volatilization, etc., and achieve the purpose of prolonging the fertilizer effect and reducing the frequency of fertilization.

[0003] In the production process of slow-release fertilizer, coating uniformity is a key indicator that determines product quality and directly affects the release rate of nutrients and fertilizer efficiency. Currently, machine vision technology based on visible light is mainly used for detecting the coating uniformity on the production line. However, due to the close color and texture between the fertilizer core and the coating material, and the complex luster reflection on the particle surface, it is difficult to stably and accurately distinguish the boundary between the coating and the core only relying on image information in the visible light band. This leads to a large error in the calculation of coating coverage, making the existing visual detection system unable to reliably quantitatively evaluate the coating uniformity, and severely relying on manual sampling inspection, which is not only low in efficiency but also difficult to ensure the consistency of product quality. Therefore, the following solutions are proposed to solve the above problems. SUMMARY

[0004] The purpose of the present application is to provide a detection method for fertilizer production, which can realize non-destructive and high-precision online detection of the coating coverage of fertilizer by introducing near-infrared multispectral vision detection technology and a coating uniformity quantitative analysis model, solving the problem of inaccurate recognition and poor evaluation reliability caused by the similar optical properties of the coating and the core material in the existing visual detection method based on visible light.

[0005] To solve the above technical problems, the present application is realized by the following technical solutions:

[0006] The present application is a detection method for fertilizer production, which comprises the following steps:

[0007] Step S1, system setting and image acquisition: install a near-infrared multispectral vision detection system at the pre-packaging section of the fertilizer production line, adjust the positions of the camera and the illumination source, and collect multispectral images of the fertilizer particles;

[0008] Step S2, image preprocessing: the collected images are sequentially corrected for dark current and flat field, to eliminate noise and uneven illumination, and then median filtered to remove random noise;

[0009] Step S3, particle segmentation and positioning: select the waveband image with the best contrast, separate the particles from the background by threshold segmentation, and position each independent particle after morphological processing to optimize the boundary;

[0010] Step S4, coating coverage calculation: based on the difference in multi-waveband reflectivity, analyze the surface material of each particle pixel by pixel, and calculate the coating area ratio of each particle to obtain the coating coverage of a single particle;

[0011] Step S5, uniformity index calculation and quality judgment: calculate the uniformity index according to the coating rate data of the whole batch of particles, compare it with the preset threshold, and output the quality judgment result to complete the sorting decision.

[0012] Further, the step S1, system setting and image acquisition specifically includes the following steps:

[0013] Step S11: install a multi-spectral visual detection system in the pre-packaging section of the slow-release fertilizer production line, which includes a near-infrared camera, a uniform illumination source, a conveyor belt for conveying fertilizer particles, and a control unit; the camera is installed vertically to the conveyor belt, with a distance of 30-50 cm from the surface of the conveyor belt to ensure that the field of view covers the entire width of the conveyor belt; the illumination source is installed at an angle of 45 degrees to reduce specular reflection;

[0014] Step S12: after the system is started, the control unit triggers the camera to collect multi-spectral images of fertilizer particles at a rate of 10 frames per second, and each image contains gray value information of multiple wavebands; the image resolution is set to 2048×1536 pixels, corresponding to an actual field of view range of 50 cm×37.5 cm, and the spatial resolution is about 0.24 mm / pixel.

[0015] Further, the step S2, image preprocessing specifically includes the following steps:

[0016] Step S21: the control unit pre-processes the collected multi-spectral images to eliminate noise and environmental influences; first, apply dark current correction to the image of each waveband: use a previously collected dark field image to subtract the current image to eliminate the camera's background noise;

[0017] Step S22: perform flat field correction: use a previously collected uniform white board image to normalize the current image to compensate for uneven illumination, and the correction formula is:

[0018]

[0019] In the formula, This represents the corrected intensity value of the pixel located at image coordinates (x, y) at wavelength λ. This represents the original intensity value of the pixel located at image coordinates (x, y) at wavelength λ. Let λ be the dark field noise intensity value of the pixel located at image coordinates (x,y) at wavelength λ. Let λ be the intensity value of the pixel located at image coordinates (x,y) on an ideal diffuse white board at wavelength λ.

[0020] Step S23: Perform median filtering on the corrected image to remove salt-and-pepper noise.

[0021] Further, in step S3, particle segmentation and localization specifically involves: the control unit extracting an 800 nm band image from the preprocessed multispectral image and applying a threshold segmentation algorithm to separate fertilizer particles from the background; specifically, using the Otsu adaptive thresholding method to calculate a global threshold T, and binarizing the image: pixels with values ​​greater than T are considered particle regions, otherwise they are background; morphological operations are performed on the binary image to smooth particle boundaries and fill small holes; each independent particle is labeled using connected component analysis, and the centroid coordinates and bounding box of each particle are calculated.

[0022] Furthermore, step S4, the calculation of coating coverage, specifically includes the following steps:

[0023] Step S41: For each segmented particle, the control unit calculates the coating coverage based on the multispectral reflectance difference; and extracts the pixel intensity value of the particle in all bands.

[0024] Step S42: For each pixel within a particle, calculate the reflectivity ratio R:

[0025]

[0026] In the formula, This represents the intensity value of a pixel in the 800nm ​​band. This represents the intensity value of a pixel in the 700 nm band.

[0027] Step S43: Since the coating material has a higher reflectivity in the 800nm ​​band and the core fertilizer has a higher reflectivity in the 700nm band, the R value can be used to distinguish between the coating and the core: if R>1.2, the pixel is classified as coating; otherwise, it is core.

[0028] Step S44: For each particle, coating coverage Calculated as the ratio of the number of coated pixels to the total number of pixels in the grain:

[0029]

[0030] In the formula, For the first The coating coverage of each fertilizer granule The total number of pixels classified as "coating" out of all pixels in a fertilizer granule; This represents the total number of pixels contained in a single fertilizer particle.

[0031] Furthermore, step S5, the uniformity index calculation and quality judgment, specifically includes the following steps:

[0032] Step S51: The control unit calculates the overall uniformity index U based on the coating coverage of all particles, first calculating the average coverage of all particles. and standard deviation :

[0033]

[0034] In the formula, The coating coverage of all fertilizer particles in the entire image or a single batch of samples. The average value; The coating coverage of all fertilizer particles in the entire image or a single batch of samples. The standard deviation is used to measure the dispersion of data. The total number of fertilizer particles identified and analyzed in the entire image or in a detection batch; For indexing; For the first The coating coverage of each fertilizer granule;

[0035] Step S52: The uniformity index U is defined as:

[0036]

[0037] In the formula, It is the uniformity index;

[0038] Step S53: Compare U with a preset threshold: If U ≥ 0.85, the batch of fertilizer is deemed qualified, the control unit outputs a "pass" signal and allows it to enter the packaging section; otherwise, a "fail" signal is output and an alarm is triggered, and the data is recorded in the database for traceability.

[0039] The present invention has the following beneficial effects:

[0040] 1. This invention analyzes the reflectance characteristics of fertilizer granules in the near-infrared band, enabling quantitative detection of coating coverage without contact or damage to the sample. This non-contact measurement completely avoids sample loss associated with traditional sampling methods, ensuring the integrity of the tested products, allowing all products to proceed to the subsequent packaging process. Furthermore, due to the specific response characteristics of near-infrared spectroscopy to coating materials, this method can penetrate the fine surface structure, effectively distinguishing the differences between the coating and the core material, achieving a more accurate and objective assessment of the coating's intrinsic quality, and improving the reliability of quality control.

[0041] 2. This invention transforms subjective qualitative judgments that originally relied on human experience into objective, quantifiable numerical indicators by defining and calculating the coating uniformity index. This quantitative assessment makes product quality standards more unified and accurate, eliminating judgment biases between different operators. The entire inspection process, from image acquisition and processing to final quality judgment, is completed automatically by the system without human intervention. This not only reduces labor costs but also enables full inspection of every batch of products on the production line, avoiding the inherent random errors of sampling inspection.

[0042] 3. This invention employs a multispectral information fusion strategy and utilizes intrinsic optical properties such as reflectivity ratios for judgment. This method enhances robustness to changes in the external environment and effectively suppresses measurement errors caused by factors such as ambient light fluctuations and complex reflections on particle surfaces. Furthermore, the detection logic based on spectral characteristics makes it highly adaptable to coating materials of different colors and formulations. By simply adjusting the corresponding characteristic wavelengths and judgment thresholds, it can be applied to different product lines, thereby improving the versatility of the detection system and its stable performance under different production conditions, and reducing the cost of replacing the detection system due to product upgrades.

[0043] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0044] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 This is a schematic flowchart of a detection method for fertilizer production according to the present invention. Detailed Implementation

[0046] 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 embodiments of the present invention, and not all embodiments. 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.

[0047] Please see Figure 1 As shown, the present invention is a detection method for fertilizer production, comprising the following steps:

[0048] Step S1, System Setup and Image Acquisition: Install a near-infrared multispectral vision inspection system in the pre-packaging section of the fertilizer production line, adjust the position of the camera and the lighting source, and acquire multispectral images of fertilizer particles;

[0049] Step S1, System Settings and Image Acquisition, specifically includes the following steps:

[0050] Step S11: Install a multispectral vision inspection system in the pre-packaging section of the slow-release fertilizer production line. The system includes a near-infrared camera, a uniform illumination source, a conveyor belt for conveying fertilizer granules, and a control unit. The camera is installed perpendicular to the conveyor belt and 30-50 cm away from the surface of the conveyor belt to ensure that the field of view covers the entire width of the conveyor belt. The illumination source is installed at a 45-degree angle to reduce specular reflection.

[0051] Step S12: After the system starts, the control unit triggers the camera to acquire multispectral images of fertilizer particles at a rate of 10 frames per second. Each image contains grayscale information of multiple bands. The image resolution is set to 2048×1536 pixels, corresponding to an actual field of view of 50 cm×37.5 cm, and the spatial resolution is approximately 0.24 mm / pixel.

[0052] Step S2, Image Preprocessing: The acquired image is subjected to dark current correction and flat field correction in sequence to eliminate noise and uneven illumination, and then median filtering is performed to remove random noise;

[0053] Step S2, image preprocessing specifically includes the following steps:

[0054] Step S21: The control unit preprocesses the acquired multispectral images to eliminate noise and environmental influences; firstly, dark current correction is applied to the images of each band: the current image is subtracted from the pre-acquired dark field image to eliminate camera background noise;

[0055] Step S22: Perform flat field correction: Normalize the current image using a pre-acquired uniform white board image to compensate for illumination non-uniformity. The correction formula is:

[0056]

[0057] In the formula, This represents the corrected intensity value of the pixel located at image coordinates (x, y) at wavelength λ. This represents the original intensity value of the pixel located at image coordinates (x, y) at wavelength λ. Let be the dark field noise intensity value of the pixel located at image coordinates (x,y) at wavelength λ; Let λ be the intensity value of the pixel located at image coordinates (x,y) on an ideal diffuse white background at wavelength λ.

[0058] Step S23: Perform median filtering on the corrected image to remove salt-and-pepper noise.

[0059] Step S3, Particle Segmentation and Localization: Select the band image with the best contrast, separate the particles from the background through threshold segmentation, and locate each independent particle after optimizing the boundary through morphological processing.

[0060] Step S3, particle segmentation and localization, specifically involves: the control unit extracting the 800nm ​​band image from the preprocessed multispectral image and applying a threshold segmentation algorithm to separate fertilizer particles from the background; specifically, using the Otsu adaptive thresholding method to calculate the global threshold T, and binarizing the image: pixels with values ​​greater than T are considered particle regions, otherwise they are background; morphological operations are performed on the binary image to smooth particle boundaries and fill small holes; each independent particle is labeled using connected component analysis, and the centroid coordinates and bounding box of each particle are calculated.

[0061] Step S4: Coating coverage calculation: Based on the multi-band reflectivity difference, analyze the particle surface material pixel by pixel, and calculate the coating area ratio of each particle to obtain the coating coverage of a single particle.

[0062] Step S4, the calculation of coating coverage specifically includes the following steps:

[0063] Step S41: For each segmented particle, the control unit calculates the coating coverage based on the multispectral reflectance difference; and extracts the pixel intensity value of the particle in all bands.

[0064] Step S42: For each pixel within a particle, calculate the reflectivity ratio R:

[0065]

[0066] In the formula, This represents the intensity value of a pixel in the 800nm ​​band. This represents the intensity value of a pixel in the 700 nm band.

[0067] Step S43: Since the coating material has higher reflectivity in the 800nm ​​band and the core fertilizer has higher reflectivity in the 700nm band, the R value can be used to distinguish between the coating and the core: if R > 1.2 (the threshold R > 1.2 can be determined, for example, by acquiring and analyzing multispectral images of standard samples with different coating degrees, establishing a model of the reflectivity difference between the coating material and the core fertilizer in the 700nm and 800nm ​​bands, and determining the optimal classification threshold through statistical methods or machine learning algorithms), then the pixel is classified as coating; otherwise, it is core.

[0068] Step S44: For each particle, coating coverage Calculated as the ratio of the number of coated pixels to the total number of pixels in the grain:

[0069]

[0070] In the formula, For the first The coating coverage of each fertilizer granule The total number of pixels classified as "coating" out of all pixels in a fertilizer granule; This represents the total number of pixels contained in a single fertilizer particle.

[0071] Step S5: Uniformity Index Calculation and Quality Judgment: Calculate the uniformity index based on the coverage data of the entire batch of particles, compare it with the preset threshold, and output the quality judgment result to complete the sorting decision.

[0072] Step S5, the calculation of uniformity index and quality judgment, specifically includes the following steps:

[0073] Step S51: The control unit calculates the overall uniformity index U based on the coating coverage of all particles, first calculating the average coverage of all particles. and standard deviation :

[0074]

[0075] In the formula, The coating coverage of all fertilizer particles in the entire image or a single batch of samples. The average value; The coating coverage of all fertilizer particles in the entire image or a single batch of samples. The standard deviation is used to measure the dispersion of data. The total number of fertilizer particles identified and analyzed in the entire image or in a detection batch; For indexing; For the first The coating coverage of each fertilizer granule;

[0076] Step S52: The uniformity index U is defined as:

[0077]

[0078] In the formula, It is the uniformity index;

[0079] Step S53: Compare U with a preset threshold (the preset threshold U is, for example, 0.85, which can be determined based on the specific national standard, industry standard or internal quality control standard of the slow-release fertilizer product, combined with statistical analysis and risk assessment of actual production data): If U≥0.85, the batch of fertilizer is deemed qualified, the control unit outputs a "pass" signal and allows it to enter the packaging section; otherwise, a "fail" signal is output and an alarm is triggered, and the data is recorded in the database for traceability.

[0080] One specific application of this embodiment is:

[0081] Background: On a slow-release fertilizer production line, a fertilizer company uses a near-infrared multispectral visual inspection method to inspect the quality of polymer-coated urea fertilizer. The production line has a conveyor belt width of 50cm and an operating speed of 0.5m / s. The fertilizer granules consist of spherical urea cores with a diameter of 2-4mm coated with a polymer coating. The system is equipped with a near-infrared camera vertically mounted 40cm above the conveyor belt surface, and the illumination source projects near-infrared light at a 45-degree angle. The following is a complete application process of this inspection method in an actual production run:

[0082] S1. System acquires multispectral images.

[0083] As fertilizer particles pass through the detection area, the camera acquires a multispectral image at a rate of 10 frames per second. The image contains four wavelengths: 700nm, 800nm, 900nm, and 1000nm. The image covers an area of ​​50cm × 37.5cm, with a resolution of 2048 × 1536 pixels, clearly displaying approximately 500 individual fertilizer particles.

[0084] S2, Image Preprocessing

[0085] The control unit performs dark current and flat field corrections on the original image. For example, for a pixel (x=100, y=200) in an 800 nm band image, its original intensity value... Dark field intensity value Whiteboard strength value The corrected strength is calculated as follows:

[0086]

[0087] The entire image is processed using this formula, and then a medium filter is applied to eliminate noise.

[0088] S3, Particle Segmentation and Positioning

[0089] From the preprocessed 800nm ​​band image, the global threshold T=0.65 (normalized intensity value) was calculated using the Otsu thresholding method. After binarization, 512 independent particles were identified through morphological operations and connected component analysis. For example, the centroid coordinates of particle No. 1 are (255,388), and the bounding box size is 15×14 pixels.

[0090] S4. Calculation of Coating Coverage

[0091] For particle number 1, extract the intensity values ​​of all pixels within it across multiple wavelengths. (Using a specific pixel...) For example, its intensity in the 700nm band Intensity at 800nm Calculate the reflectivity ratio:

[0092]

[0093] Since R > 1.2, this pixel is identified as a coated region. Count the total number of pixels for particle #1. The number of pixels in the coating Calculate coverage:

[0094]

[0095] Repeat this process for all 512 particles to obtain the coverage set, for example: .

[0096] S5. Calculation of Uniformity Index and Quality Judgment:

[0097] Calculate the average coverage based on the coverage values ​​of all particles. and standard deviation :

[0098]

[0099] Then the uniformity index is calculated:

[0100]

[0101] The system compares U=0.953 with the preset threshold of 0.85. Since 0.953>0.85, the system determines that the coating uniformity of the batch of fertilizer is qualified, outputs a "pass" signal, and allows it to enter the packaging section.

[0102] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0103] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A detection method for fertilizer production, characterized in that, The detection method includes the following steps: Step S1, System Setup and Image Acquisition: Install a near-infrared multispectral vision inspection system in the pre-packaging section of the fertilizer production line, adjust the position of the camera and the lighting source, and acquire multispectral images of fertilizer particles; Step S2, Image Preprocessing: The acquired image is subjected to dark current correction and flat field correction in sequence to eliminate noise and uneven illumination, and then median filtering is performed to remove random noise; Step S3, Particle Segmentation and Localization: Select the band image with the best contrast, separate the particles from the background through threshold segmentation, and locate each independent particle after optimizing the boundary through morphological processing. Step S4, Coating Coverage Calculation: Based on the multi-band reflectivity difference, the surface material of the particles is analyzed pixel by pixel, the proportion of the coating area of ​​each particle is counted, and the coating coverage of a single particle is obtained. Step S5: Uniformity Index Calculation and Quality Judgment: Calculate the uniformity index based on the coverage data of the entire batch of particles, compare it with the preset threshold, and output the quality judgment result for sorting decision.

2. The detection method for fertilizer production according to claim 1, characterized in that, Step S1, system settings and image acquisition, specifically includes the following steps: Step S11: Install a multispectral vision inspection system in the pre-packaging section of the slow-release fertilizer production line. The system includes a near-infrared camera, a uniform illumination source, a conveyor belt for conveying fertilizer granules, and a control unit. The camera is installed perpendicular to the conveyor belt and 30-50 cm away from the surface of the conveyor belt to ensure that the field of view covers the entire width of the conveyor belt. The illumination source is installed at an angle to reduce specular reflection. Step S12: After the system starts, the control unit triggers the camera to acquire multispectral images of fertilizer particles at a rate of 10 frames per second. Each image contains grayscale information of multiple bands.

3. The detection method for fertilizer production according to claim 1, characterized in that, Step S2, image preprocessing, specifically includes the following steps: Step S21: The control unit preprocesses the acquired multispectral images to eliminate noise and environmental influences; firstly, dark current correction is applied to the images of each band: the current image is subtracted from the pre-acquired dark field image to eliminate camera background noise; Step S22: Perform flat field correction: Normalize the current image using a pre-acquired uniform white board image to compensate for illumination non-uniformity. The correction formula is: In the formula, This represents the corrected intensity value of the pixel located at image coordinates (x, y) at wavelength λ. This represents the original intensity value of the pixel located at image coordinates (x, y) at wavelength λ. Let λ be the dark field noise intensity value of the pixel located at image coordinates (x,y) at wavelength λ. Let λ be the intensity value of the pixel located at image coordinates (x,y) on an ideal diffuse white board at wavelength λ. Step S23: Perform median filtering on the corrected image to remove salt-and-pepper noise.

4. The detection method for fertilizer production according to claim 1, characterized in that, Step S3, particle segmentation and localization, specifically involves: the control unit extracting the 800 nm band image from the preprocessed multispectral image and applying a threshold segmentation algorithm to separate fertilizer particles from the background; specifically, using the Otsu adaptive thresholding method to calculate the global threshold T, and binarizing the image: pixels with values ​​greater than T are considered particle regions, otherwise they are background; morphological operations are performed on the binary image to smooth particle boundaries and fill small holes; each independent particle is labeled using connected component analysis, and the centroid coordinates and bounding box of each particle are calculated.

5. The detection method for fertilizer production according to claim 1, characterized in that, Step S4, the calculation of coating coverage rate, specifically includes the following steps: Step S41: For each segmented particle, the control unit calculates the coating coverage based on the multispectral reflectance difference; and extracts the pixel intensity value of the particle in all bands. Step S42: For each pixel within a particle, calculate the reflectivity ratio R: In the formula, This represents the intensity value of a pixel in the 800nm ​​band. This represents the intensity value of a pixel in the 700 nm band. Step S43: Since the coating material has a high reflectivity in the 800nm ​​band and the core fertilizer has a high reflectivity in the 700nm band, the R value can be used to distinguish between the coating and the core: when R is greater than a threshold pre-calibrated or determined by calibration based on the spectral characteristics of the coating material and the core fertilizer, the pixel is classified as a coating. Step S44: For each particle, coating coverage Calculated as the ratio of the number of coated pixels to the total number of pixels in the grain: In the formula, For the first The coating coverage of each fertilizer granule The total number of pixels classified as "coating" out of all pixels in a fertilizer granule; This represents the total number of pixels contained in a single fertilizer particle.

6. The detection method for fertilizer production according to claim 1, characterized in that, Step S5, the uniformity index calculation and quality judgment, specifically includes the following steps: Step S51: The control unit calculates the overall uniformity index U based on the coating coverage of all particles, first calculating the average coverage of all particles. and standard deviation : In the formula, The coating coverage of all fertilizer particles in the entire image or a single batch of samples. The average value; The coating coverage of all fertilizer particles in the entire image or a single batch of samples. The standard deviation is used to measure the dispersion of data. The total number of fertilizer particles identified and analyzed in the entire image or in a detection batch; For indexing; For the first The coating coverage of each fertilizer granule; Step S52: The uniformity index U is defined as: In the formula, It is the uniformity index; Step S53: Compare U with a threshold predetermined according to fertilizer product quality standards or production requirements: when U is greater than or equal to the threshold, the batch of fertilizer is deemed qualified, the control unit outputs a pass signal and allows it to enter the packaging section; otherwise, a fail signal is output and an alarm is triggered, and the data is recorded in the database for traceability.

Images