An image recognition-based superfine high-fiber instant broccoli pollen particle size monitoring method
Image recognition technology enables real-time monitoring and process optimization of the processing of ultrafine, high-fiber, quick-dissolving broccoli pollen, solving the problems of complexity and low accuracy of traditional detection methods and improving product quality and production efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- JINAN INST OF FRUIT PRODS CHINA GENERAL SUPPLY & MARKETING COOP
- Filing Date
- 2026-03-05
- Publication Date
- 2026-07-03
AI Technical Summary
Existing technologies cannot achieve real-time monitoring of the processing of ultrafine, high-fiber, quick-dissolving broccoli pollen. Furthermore, traditional detection methods suffer from problems such as complex operation, long detection cycles, low accuracy, and inability to simultaneously analyze particle size distribution, making it difficult to meet the needs of process optimization.
An image acquisition system was built using an image recognition-based method. Through image preprocessing, contour extraction, and feature filtering, the particle size and distribution were calculated, and real-time monitoring and process adjustment were performed in conjunction with process thresholds.
It enables simultaneous analysis of particle size and particle size distribution, improves detection accuracy, supports real-time process optimization, enhances the quality indicators of broccoli pollen such as solubility, water retention capacity and swelling capacity, and reduces the defect rate.
Smart Images
Figure CN122335664A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of image recognition technology, and in particular to a method for monitoring the particle size of ultrafine, high-fiber, fast-dissolving broccoli pollen based on image recognition. Background Technology
[0002] The core processing of ultrafine high-fiber instant broccoli pollen is wet ultrafine grinding and enzymatic hydrolysis. Particle size and particle size distribution are the key parameters that determine its solubility, water holding capacity, swelling capacity and other key quality indicators. At the same time, particle size directly affects the stability of broccoli pollen paste. The smaller the particle size and the more uniform the distribution, the less likely the pollen paste is to float or settle. The instant solubility and dietary fiber activity of the finished powder are also higher.
[0003] Currently, particle size detection in the processing of ultrafine, high-fiber, instant broccoli pollen mainly employs laser particle size analyzer. This method requires pretreatment such as sampling, dilution, and ultrasonic dispersion, making it complex and time-consuming. It cannot achieve real-time monitoring of the processing, and the sampling process is prone to sample bias, leading to test results that do not match the actual process conditions. Furthermore, traditional detection methods can only output characteristic particle size data, failing to simultaneously analyze indicators directly related to process optimization, such as particle size distribution span and specific surface area, making it difficult to quickly provide feedback on the adjustment direction of process parameters such as wet ultrafine grinding and enzymatic hydrolysis. In addition, for in-situ detection of broccoli pollen pulp, traditional methods are affected by pulp viscosity and particle adhesion, resulting in low detection accuracy and failing to meet the need for precise monitoring of D50 in ultrafinely ground pollen pulp. Summary of the Invention
[0004] In view of the aforementioned existing problems, the inventors have proposed the present invention.
[0005] This invention provides the following technical solution: In a first aspect, the present invention provides a method for monitoring the particle size of ultrafine, high-fiber, fast-dissolving broccoli pollen based on image recognition, comprising: S1. Set up an image acquisition system. Depending on whether the object to be detected is broccoli pollen paste or dry powder, select the appropriate imaging module and shooting parameters to acquire microscopic images of the sample. S2. Preprocess the acquired microscopic images by sequentially performing image denoising, grayscale conversion, binarization, and edge enhancement to eliminate recognition errors caused by background interference and particle adhesion. S3. A contour extraction algorithm is used to identify particle contours in the preprocessed image. Feature filtering is performed by combining area, roundness, and pixel threshold to remove impurities and pseudo-particle contours, thus obtaining a valid broccoli pollen particle contour dataset. S4. Based on the effective contour dataset, calculate the equivalent particle size of a single broccoli pollen grain and statistically analyze the samples. The particle size distribution of all particles is determined, the characteristic particle sizes D10, D50, and D90 and the particle size distribution span are determined, and the specific surface area of the particles is estimated based on the particle size data. S5. Compare the test results of S4 with the process threshold for processing ultrafine high-fiber instant broccoli pollen. If the test index exceeds the preset range, output a process adjustment prompt; if it meets the preset range, generate a particle size monitoring report.
[0006] Preferably, the image acquisition system includes an industrial camera, a microscope imaging lens, a ring-shaped supplementary light source, a stage, and a data transmission module; for broccoli pollen slurry produced during wet ultrafine grinding and enzymatic hydrolysis, in-situ imaging is performed using a transparent flowing tank, with a lens magnification of 200~500X and a diffused white light source to avoid glare from the slurry; for dried broccoli powder after vacuum freeze-drying, imaging is performed using a scanning electron microscope with a matching imaging module, with a lens magnification of 500~1000X and imaging parameters of EHT-3.00K and WD8~8.5mm.
[0007] Preferably, the pixel resolution of the microscopic image is not less than 1920×1080, and the image acquisition frequency is set to 1~5 frames / second according to the processing rhythm, so as to realize the linkage acquisition with the broccoli pollen processing equipment during online monitoring.
[0008] Preferably, the image denoising uses a combination of median filtering and Gaussian filtering, with a median filtering window of 3×3 and a Gaussian filtering variance of 0.8~1.0; binarization uses the maximum inter-class variance method to determine the threshold, and edge enhancement is performed using the Sobel operator to highlight the edge features of the particles.
[0009] Preferably, the contour extraction algorithm is the cv2.findContours() algorithm based on OpenCV, the contour retrieval mode is external contour retrieval, and the approximation method is approximation polygon; the feature selection threshold is: particle contour area ≥ 50 pixels, roundness 0.3~1.0, equivalent particle size ≥ 5μm, and contours exceeding the threshold are removed.
[0010] Preferably, the equivalent particle size of a single particle is calculated using the equivalent circle area method, with the formula: d = 2√(S / π), where d is the equivalent particle size, S is the pixel area of the particle outline, and π is pi; the feature particle sizes D10, D50, and D90 are the particle size values corresponding to the cumulative percentages of 10%, 50%, and 90% after arranging the equivalent particle sizes of all effective particles in ascending order; the particle size distribution span is calculated using the formula: span = (D90 - D10) / D50.
[0011] Preferably, the specific surface area of the particles is calculated based on the particle size distribution data, and the calculation formula is: S=6 / (ρ ×D50), where S is the specific surface area, in m² / g; ρ is the bulk density of broccoli pollen, in g / mL; and D50 is the volume-weighted average particle size, in m.
[0012] Preferably, the process thresholds for processing the ultrafine high-fiber instant broccoli pollen are: pollen pulp D50 ≤ 37.66 μm and span ≤ 1.713 μm after wet ultrafine grinding; after enzymatic hydrolysis (enzymatic hydrolysis for 2.5 h, 3% enzyme addition). The dry powder has a D50 of 27.20 ± 0.01 μm, a specific surface area of 0.15 ± 0.001 m² / g, and a particle size distribution range ≤ 1.5μm.
[0013] The beneficial effects of this invention are as follows: By using image acquisition parameters and recognition algorithms, it solves the recognition error problems caused by pollen pulp reflection, powder particle adhesion, and background interference. It achieves simultaneous analysis of particle size, particle size distribution range, and specific surface area, eliminating the need for additional detection equipment. A single image acquisition can output multiple indicators directly related to broccoli pollen quality and process optimization, solving the problems of single-indicator detection and incomplete information feedback in traditional methods. It establishes a threshold system that matches the optimal process for ultrafine, high-fiber, quick-dissolving broccoli pollen. The detection results can directly guide the adjustment of process parameters, such as the gap between ultrafine grinding blades, enzymatic hydrolysis time, and enzyme addition amount, achieving closed-loop optimization of the process. This effectively improves the solubility, water holding capacity, and swelling capacity of broccoli pollen, while also improving the stability of pollen pulp and reducing the defect rate of finished products. Attached Figure Description
[0014] 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 accompanying 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.
[0015] Figure 1 is a flowchart illustrating a method for monitoring the particle size of ultrafine high-fiber instant broccoli pollen based on image recognition in Example 1. Detailed Implementation
[0016] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
[0017] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.
[0018] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that excludes other embodiments.
[0019] Example 1, referring to Figure 1, is the first embodiment of the present invention. This embodiment provides a method for monitoring the particle size of ultrafine, high-fiber, fast-dissolving broccoli pollen based on image recognition, including the following steps: Image acquisition system setup and sample shooting An image acquisition system consisting of an industrial camera, a microscopic imaging lens, a ring-shaped supplementary light source, a stage, and a data transmission module was constructed. Depending on whether the object to be detected is broccoli pollen paste (wet ultrafine grinding and enzymatic hydrolysis) or dry powder (after vacuum freeze-drying), the imaging module and shooting parameters were adjusted to achieve interference-free acquisition of microscopic images of the sample.
[0020] For broccoli pollen slurry, in-situ imaging was performed using a transparent flow tray to avoid errors caused by sample dilution. The lens magnification was 200-500X, and a diffused ring-shaped white light supplement was used to eliminate slurry reflections and background interference. For dried broccoli powder, a scanning electron microscope with an imaging module was used, with a lens magnification of 500-1000X. The imaging parameters were set to EHT-3.00K and WD8-8.5mm to clearly capture the microscopic morphology and particle outline of the powder. The image acquisition resolution was no less than 1920×1080, and the acquisition frequency during online monitoring was matched with the processing rhythm at 1-5 frames / second.
[0021] Image preprocessing The acquired microscopic images undergo multi-step preprocessing to eliminate recognition errors caused by background noise, particle adhesion, and uneven illumination, laying the foundation for subsequent contour extraction. Denoising: A 3 × 3 window mid-range filter combined with a Gaussian filter with a variance of 0.8 to 1.0 is used to remove salt-and-pepper noise and Gaussian noise from the image while preserving particle edge features; Grayscale conversion: Converting a color image to a grayscale image reduces data volume and improves processing efficiency; Binarization: The binarization threshold is automatically determined using the maximum inter-class variance method to clearly separate particles from the background; Edge enhancement: The Sobel operator is used to enhance the edges of the binarized image, highlighting the contour features of the particles and solving the problem of blurred contours caused by the adhesion of broccoli pollen particles.
[0022] Particle contour extraction and feature selection The cv2.findContours() algorithm based on OpenCV was used to extract contours from the preprocessed image. The contour retrieval mode was set to external contour retrieval, and the approximation method was approximation polygon. Only the outer contour of the broccoli pollen grains was extracted to avoid contour errors caused by internal pores.
[0023] Feature filtering was performed on all extracted contours to remove impurities, bubbles, and pseudo-particle contours, thus determining the effective broccoli pollen particle contour dataset. The filtering thresholds were: particle contour area ≥ 50 pixels, roundness 0.3~1.0, and equivalent particle size ≥ 5μm, ensuring that the filtered contours were all effective broccoli pollen particles.
[0024] Calculation of particle size and related indices Based on an effective contour dataset, the equivalent area of a single broccoli pollen grain is calculated using the equivalent circle area method. The equivalent particle size is calculated using the formula: d = 2√(S / π), where d is the equivalent particle size, S is the pixel area of the particle outline, and π is pi. The equivalent particle sizes of all effective particles are arranged in ascending order, and the characteristic particle sizes D10, D50, and D90 (particle sizes corresponding to 10%, 50%, and 90% of the total) are statistically obtained. The particle size distribution span is then calculated using the formula span = (D90 - D10) / D50. A smaller span indicates a more uniform particle size distribution.
[0025] Based on the particle size distribution data and the bulk density of broccoli pollen, the specific surface area of the particles is calculated using the formula: S = 6 / (ρ × D50), where S is the specific surface area (m² / g), ρ is the bulk density of broccoli pollen (g / mL), and D50 is the volume-weighted average particle size (m), thus achieving simultaneous analysis of particle size and specific surface area.
[0026] Test result judgment and process feedback The detection indicators (D10, D50, D90, span, specific surface area) were compared with the preset process thresholds for processing ultrafine high-fiber instant broccoli pollen. These process thresholds were determined based on the optimal process parameters for wet ultrafine grinding and enzymatic hydrolysis. Pollen paste after wet ultrafine grinding: D50≤37.66μm, span≤1.713μm; The dried powder after enzymatic hydrolysis (enzymatic hydrolysis for 2.5 h, 3% enzyme addition): D50 = 27.20 ± 0.01 μm, specific surface area = 0.15 ± 0.001 m² / g, span ≤ 1.5 μm.
[0027] If the detection indicators meet the preset threshold, the current process status is determined to be normal, and a particle size monitoring report is generated, including data such as characteristic particle size, particle size distribution, and specific surface area. If the detection indicators exceed the preset threshold, a process adjustment prompt is output in a timely manner. For example, if the particle size is too large in the ultrafine grinding stage, the prompt is to increase the grinding shear force and reduce the blade gap. If the particle size distribution is uneven in the enzymatic hydrolysis stage, the prompt is to adjust the enzymatic hydrolysis time or the amount of enzyme added. This method supports two modes: offline detection and online monitoring. Offline detection is suitable for laboratory process development and finished product sampling inspection, and outputs single test results. In online monitoring, the image acquisition system is deployed at key process nodes of the processing equipment to realize real-time image acquisition, real-time data analysis and real-time process feedback. The monitoring data is automatically stored in the database, supporting historical record traceability and process trend analysis.
[0028] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A method for monitoring the particle size of ultrafine, high-fiber, fast-dissolving broccoli pollen based on image recognition, characterized in that, include: S1. Set up an image acquisition system. Depending on whether the object to be detected is broccoli pollen paste or dry powder, select the appropriate imaging module and shooting parameters to acquire microscopic images of the sample. S2. Preprocess the acquired microscopic images by sequentially performing image denoising, grayscale conversion, binarization, and edge enhancement to eliminate recognition errors caused by background interference and particle adhesion. S3. A contour extraction algorithm is used to identify particle contours in the preprocessed image. Feature filtering is performed by combining area, roundness, and pixel threshold to remove impurities and pseudo-particle contours, thus obtaining a valid broccoli pollen particle contour dataset. S4. Based on the effective contour dataset, calculate the equivalent particle size of a single broccoli pollen grain, statistically analyze the particle size distribution of all grains in the sample, determine the characteristic particle sizes D10, D50, D90 and the particle size distribution span, and estimate the specific surface area of the grains based on the particle size data. S5. Compare the test results of S4 with the process threshold of ultrafine high fiber instant broccoli pollen processing. If the test index exceeds the preset range, output a process adjustment prompt. If the particle size meets the preset range, a particle size monitoring report will be generated.
2. The method for monitoring the particle size of ultrafine, high-fiber, fast-dissolving broccoli pollen based on image recognition as described in claim 1, characterized in that: The image acquisition system includes an industrial camera, a microscope imaging lens, a ring-shaped supplementary light source, a stage, and a data transmission module. For broccoli pollen slurry produced during wet ultrafine grinding and enzymatic hydrolysis, in-situ imaging is performed using a transparent flowing tank with a lens magnification of 200-500X and a diffused white light source to avoid glare from the slurry. For dried broccoli powder produced after vacuum freeze-drying, imaging is performed using a scanning electron microscope with a matching imaging module, with a lens magnification of 500-1000X and imaging parameters of EHT-3.00K and WD8-8.5mm.
3. The method for monitoring the particle size of ultrafine, high-fiber, fast-dissolving broccoli pollen based on image recognition as described in claim 1, characterized in that: The microscopic image pixel resolution is no less than 1920 × 1080, and the image acquisition frequency is set to 1~5 frames / second according to the processing rhythm. During online monitoring, it can realize linkage acquisition with the broccoli pollen processing equipment.
4. The method for monitoring the particle size of ultrafine, high-fiber, quick-dissolving broccoli pollen based on image recognition as described in claim 1, characterized in that: The image denoising method employs a combination of median filtering and Gaussian filtering, with a median filtering window of 3×3 and a Gaussian filtering variance of 0.8~1.0; binarization uses the maximum inter-class variance method. A threshold is set, and edge enhancement is performed using the Sobel operator to highlight the edge features of the particles.
5. The method for monitoring the particle size of ultrafine, high-fiber, fast-dissolving broccoli pollen based on image recognition as described in claim 1, characterized in that: The contour extraction algorithm is based on OpenCV's cv2.findContours() algorithm. The contour retrieval mode is external contour retrieval, and the approximation method is approximation polygon. The feature selection threshold is: particle contour area ≥ 50 pixels, roundness 0.3~1.0, equivalent particle size ≥ 5μm, and contours exceeding the threshold are removed.
6. The method for monitoring the particle size of ultrafine, high-fiber, quick-dissolving broccoli pollen based on image recognition as described in claim 1, characterized in that: The equivalent particle size of a single particle is calculated using the equivalent circle area method, with the formula: d = 2√(S / π), where d is the equivalent particle size, S is the pixel area of the particle outline, and π is pi; the characteristic particle sizes D10, D50, and D90 are the particle size values corresponding to the cumulative percentages of 10%, 50%, and 90% after arranging the equivalent particle sizes of all effective particles in ascending order; the particle size distribution span is calculated using the formula: span = (D90 - D10) / D50.
7. The method for monitoring the particle size of ultrafine, high-fiber, fast-dissolving broccoli pollen based on image recognition as described in claim 1, characterized in that: The specific surface area of the particles is estimated based on the particle size distribution data. The estimation formula is: S=6 / (ρ×D50), where S is the specific surface area, in m² / g. ρ is the bulk density of broccoli pollen, in g / mL; D50 is the volume-weighted average particle size, in m.
8. The method for monitoring the particle size of ultrafine, high-fiber, fast-dissolving broccoli pollen based on image recognition as described in claim 1, characterized in that: The process thresholds for processing ultrafine high-fiber instant broccoli pollen are as follows: After wet ultrafine grinding, the pollen paste has a D50 of ≤37.66μm and a particle size distribution of ≤1.713μm; after enzymatic hydrolysis (enzymatic hydrolysis for 2.5h, 3% enzyme addition), the dried powder has a D50 of 27.20±0.01μm, a specific surface area of 0.15±0.001m² / g, and a particle size distribution of ≤1.5μm.