An ore granularity online analysis method and system based on visual segmentation
The online ore particle size analysis system, which combines visual segmentation methods with deep learning, solves the problems of real-time performance and accuracy in ore particle size analysis. It achieves high-precision segmentation and automated statistics in complex environments and adapts to the deployment needs of multiple mining areas.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CITIC HEAVY INDUSTRIES CO LTD
- Filing Date
- 2026-02-03
- Publication Date
- 2026-06-26
AI Technical Summary
Existing methods for ore particle size analysis are time-consuming and labor-intensive, have poor real-time performance, are difficult to quantify accurately, have poor repeatability, and in complex environments, ore image recognition results are prone to multiple ore targets sticking together, resulting in high segmentation error rates and insufficient product reliability and shock resistance.
An online ore particle size analysis method based on visual segmentation is adopted. A pre-trained dataset is established by collecting images across working conditions. The pre-trained model is trained using a deep learning semantic segmentation network model and fine-tuned at the target site. Accurate segmentation is achieved by combining OpenCV and watershed segmentation algorithm. A stable industrial camera mounting bracket is designed to realize real-time and accurate ore particle size analysis.
It enables accurate real-time analysis of ore particle size in complex environments, improving segmentation accuracy and system reliability. It can automatically count ore particle size without manual screening, and it is linked with an intelligent control system to provide real-time data and control suggestions.
Smart Images
Figure CN122289673A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of ore image processing technology, specifically to an online ore particle size analysis method and system based on visual segmentation. Background Technology
[0002] Ore particle size is a key production indicator in the grinding process of mines, referring to the relative content of ore particles of different sizes. Traditional methods for ore particle size statistics require manual screening or experience-based observation after the conveyor belt is stopped; this is time-consuming, labor-intensive, lacks real-time accuracy, is difficult to quantify accurately, and has poor repeatability, which seriously restricts the automation and stability of the production line and therefore cannot meet the usage requirements.
[0003] With the development of vision technology, ore image granularity processing technology based on industrial cameras has gradually emerged. This technology uses high-quality industrial-grade cameras to acquire images of ore on moving conveyor belts, analyzes and statistically analyzes ore distribution data of different particle sizes in real time, and has advantages such as high real-time performance and non-contact operation. However, existing ore image granularity processing still has the following problems: 1. Ore flow images have phenomena such as dust occlusion, similar gray levels, and unclear boundaries between adjacent ores. The recognition results of existing algorithms are prone to multiple ore targets sticking together, resulting in a high segmentation error rate; 2. The on-site working conditions are harsh, and the reliability and shock resistance of existing products are insufficient. Summary of the Invention
[0004] To address the shortcomings of existing technologies, this invention provides a method and system for online ore particle size analysis based on visual segmentation, which can accurately analyze ore particle size in real time under complex environments, thereby improving the accuracy and reliability of ore particle size analysis.
[0005] To achieve the above objectives, the technical solution adopted by this invention is: an online ore particle size analysis method based on visual segmentation, used to perform particle size analysis on ore flow transported on a belt conveyor, comprising the following steps: S1. Collect multiple ore images from different sites, working conditions, and time periods across different working conditions to establish a pre-training dataset; after preprocessing the images in the pre-training dataset and labeling the ore contour information, use the pre-training dataset to train a deep learning semantic segmentation network model to obtain a pre-training model. S2. Collect multiple ore images at different time periods at the target site and establish a fine-tuning dataset. The number of images in the fine-tuning dataset is less than the number of images in the pre-training dataset. After preprocessing the images in the fine-tuning dataset and labeling the ore contour information, use the fine-tuning dataset to fine-tune the pre-training model obtained in step S1. During the training process, freeze the first three network convolutional layers of the model and save the model with the highest score as the final fine-tuning model. S3. Real-time acquisition of ore images from the target site, preprocessing, and input into the final fine-tuning model obtained in step S2 to infer the preliminary edge segmentation map of the ore. S4. Perform OpenCV thresholding, morphological processing, distance transformation and threshold labeling on the preliminary edge segmentation map in step S3, and then use the watershed segmentation algorithm with label guidance to segment to obtain the final ore region segmentation map. S5. Perform statistical calculations on the final ore region segmentation map to obtain ore particle size data for each particle size class.
[0006] According to claim 1, the online ore particle size analysis method based on visual segmentation is characterized in that: the images in the fine-tuning dataset are derived from a single target site, and the number of images in the fine-tuning dataset is one-tenth of that in the pre-training dataset.
[0007] Furthermore, in steps S1, S2 and S3, the image preprocessing process is as follows: first, the image is distorted and corrected; then, the corrected RGB image is converted to grayscale and the image size is adjusted; then, smoothing filtering is performed; and finally, contrast enhancement operation is performed on the filtered image to increase the effective features of the ore image.
[0008] Furthermore, in steps S1 and S2, before loading the images from the pre-training dataset or fine-tuning dataset into the model, data augmentation functions are used to perform random rotation, stitching, cropping, and normalization operations on the images to expand the dataset size.
[0009] Furthermore, step S4 specifically includes the following steps: S4.1 Perform OpenCV thresholding and morphological processing on the preliminary edge segmentation map obtained in step S3 to obtain a binarized image; S4.2. Using distance variation and setting a distance threshold, process the obtained binary map to obtain a guide marker map of the ore region; S4.3. Use the marker-guided watershed segmentation algorithm to segment the binary map to obtain the final ore region segmentation map.
[0010] This invention discloses an online ore particle size analysis system based on visual segmentation, comprising an image acquisition module, an image preprocessing module, an algorithm processing module, a data processing module, a data communication module, a display module, and a control module. The image acquisition module includes an industrial camera and a camera fixture. The industrial camera is mounted above the ore flow via the camera fixture for dynamically acquiring image data of the ore flow. An electrical control box is installed near the camera fixture, and the image preprocessing module, algorithm processing module, and data processing module are all mounted on a computing unit within the electrical control box. The image preprocessing module is used to preprocess the ore image to enhance its feature information; the algorithm processing module is used to train and store the deep learning semantic segmentation network model, and use the trained final fine-tuned model and image segmentation method to perform edge segmentation and processing on the preprocessed ore image to obtain the final ore region segmentation map; the data processing module uses data analysis algorithms to perform statistical analysis on the final ore region segmentation map and to measure the ore particle size composition in real time. The data communication module is used to realize communication between the electrical control box and the central control room workstation, sending the images and data processed by the electrical control box's computing unit to the central control room workstation; the display module and the control module are respectively installed in the central control room workstation. The display module can process and store the received images and data, and display the data; the control module can exchange data with other intelligent control systems, analyze the ore particle size data, and issue intelligent control and adjustment suggestions.
[0011] Furthermore, the camera fixture includes a bracket, which is suspended above the ore flow by multiple steel wire ropes evenly distributed around it. A protective box with an adjustable vertical angle is installed in the middle of the bracket. The industrial camera is installed inside the protective box. The vertical center of the camera image coincides with the vertical center of the actual distribution of ore on the belt conveyor, ensuring that the incoming ore captured by the industrial camera remains horizontal. At the same time, clean high-pressure air is used to blow away the dust in front of the lens.
[0012] Furthermore, the camera fixture has a rectangular frame, with four steel wire ropes connecting to the four corners of the frame for hoisting. The height of the steel wire ropes can be adjusted.
[0013] Furthermore, the support is also equipped with strip light sources parallel to the direction of belt conveyor operation. The two strip light sources are located on both sides of the protective box and are used to provide supplementary lighting for the ore flow.
[0014] Furthermore, the display module can use a fitting formula algorithm to fit the received raw data, and then store and display the data. The displayed information includes data curves, ore segmentation images, fitting percentage values, and maximum ore particle size.
[0015] Beneficial effects: This invention uses an industrial camera to acquire real-time images of ore on a conveyor belt, and uses a combination of deep learning and traditional image processing to segment the ore edge information. By using the ore edge information to statistically determine the proportion of each particle size in the ore, it provides the staff with result data display and intelligent control suggestions, thereby improving the accuracy and reliability of ore particle size analysis.
[0016] 1. When using a deep learning model for image processing, this invention first establishes a pre-training dataset by collecting ore images from different locations and time periods at different project sites. The images in the pre-training dataset come from a wide range of sources, thus enabling the training of a pre-training model with strong generalization ability. Then, a fine-tuning dataset is established using a small number of images from the target site. Through fine-tuning training, a final fine-tuned model with high accuracy is obtained. Through two-stage training, the general model is "customized" into a site-specific model, which can improve the accuracy of image recognition, significantly reduce annotation costs, and adapt to the deployment needs of multiple mining areas with large differences in ore composition and particle size distribution.
[0017] 2. This invention integrates deep learning with traditional image processing techniques to form an image segmentation system that first intelligently perceives and then refines the image. First, the final fine-tuning model of deep learning is used to process the ore image for perception, identifying the general ore region from the complex background and solving the problem of unstructured interference that traditional methods cannot handle. Then, OpenCV and traditional image segmentation algorithms are used to refine the preliminary edge segmentation map of the ore, making up for the shortcomings of deep learning in boundary details and adhesion segmentation, and improving the accuracy of the algorithm in segmenting adjacent ore regions with similar gray levels and unclear boundaries, resulting in a more accurate final ore region segmentation map. Through statistical calculation, the ore particle size data of each level can be obtained, ensuring the statistical accuracy of ore particle size without manual screening.
[0018] 3. This invention uses OpenCV to perform thresholding, morphological processing, distance transformation and threshold marking, and guided watershed segmentation on the image to achieve accurate segmentation. Traditional watershed algorithms are prone to oversegmentation, while this invention accurately locates the central region of the ore through distance transformation and threshold marking, providing "navigation coordinates" for the watershed algorithm. This allows the algorithm to actively avoid edge noise and only converge at the true boundary, achieving separation of adhering ore and significantly improving segmentation accuracy in actual production lines.
[0019] 4. This invention provides an online analysis system consisting of an industrial camera, an electrical control box, and a hollow workstation from a hardware perspective. A special mounting bracket is designed for the industrial camera to ensure stable installation and resistance to vibration interference, ensuring stable and clear images acquired by the camera, effectively addressing the shortcomings of insufficient reliability and shock resistance in existing products. The display module and control module are installed in the central control room workstation, which can display data fitting information on ore particle size, promptly detect large ore pieces, issue early warnings, propose control suggestions, assist in decision-making, and achieve linkage with the intelligent control system. Attached Figure Description
[0020] Figure 1 This is a flowchart of the model training process of the present invention.
[0021] Figure 2 This is a flowchart of the image preprocessing process of the present invention.
[0022] Figure 3 This is a flowchart of the online particle size analysis process for ore according to the present invention.
[0023] Figure 4 This is a hardware layout diagram of the online analysis system of the present invention.
[0024] Figure 5 This is a structural diagram of the camera fixture of the present invention.
[0025] Figure 6 This is a flowchart of the online analysis system of the present invention.
[0026] Reference numerals: 1. Protective box; 2. Bracket; 3. Steel wire rope; 4. Strip light source; P1, P2, P3, P6 indicate power cable connection; P4, P7, P8 indicate Cat5e network cable connection; P5 indicates fiber optic connection. Detailed Implementation
[0027] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments; the embodiments described below are only for explaining the technical solutions of the present invention, and the content of the present invention is not limited to the following embodiments.
[0028] See appendix Figure 1-3 The specific implementation of the online ore particle size analysis method of the present invention includes steps S1-S5, wherein steps S1-S2 are preliminary work, which mainly provide a method for training an ore image segmentation model in two stages; steps S3-S4 combine the trained deep learning segmentation model and image segmentation algorithm to segment the ore images collected on site; and step S5 performs statistical analysis on the segmented images to obtain ore particle size data.
[0029] S1, such as Figure 1 As shown in the left branch, multiple ore images are collected across different sites, working conditions, and time periods to establish a pre-training dataset. After preprocessing the images in the pre-training dataset and labeling the ore contour information, a deep learning semantic segmentation network model is trained using the pre-training dataset to obtain the pre-trained model, specifically including S1.1-S1.4.
[0030] S1.1 In the early stage, multiple site inspections were conducted to determine the installation location, and a high frame rate industrial camera was installed at a suitable position at the top of the ore flow. Images of the ore under different site conditions and different external environmental influences were collected across working conditions, with one image collected every half hour as a pre-training dataset.
[0031] S1.2, Select 500 images from the acquired images to ensure coverage of all operating conditions. (Following the steps as follows...) Figure 2The process shown involves preprocessing the pre-training dataset, specifically including: correcting the images in the dataset using a distortion correction algorithm to make subsequent calculation results more accurate; converting the corrected RGB images to grayscale and adjusting the image size to 800*640 to meet the input requirements of the network model; smoothing the obtained grayscale image using the OpenCV algorithm to reduce the impact of noise; and applying contrast enhancement to the filtered image to increase the effective features of the mineral image.
[0032] In step S1.3, professional personnel complete the annotation of ore images, annotating the ore contour information in the images, and divide the pre-training dataset into training, validation, and test sets in an 8:1:1 ratio; then, data augmentation functions (performing random rotation, stitching, cropping, and normalization operations) are used to expand the dataset size to improve the generalization of the trained model.
[0033] S1.4, Model training is performed on the processed dataset using a graphics card with 32GB of VRAM, a training batch size of 16, and the Adam optimizer. The optimal pre-trained model parameters are then evaluated using the validation set. The pre-training dataset has a wide range of data sources, resulting in a pre-trained model with strong generalization ability. Subsequent fine-tuning using only a small amount of image data from other locations can yield a good model.
[0034] S2, such as Figure 1 As shown in the right branch, camera fixtures and industrial cameras are installed at the target implementation site to collect multiple ore images at different time periods. Fifty images are selected as the fine-tuning dataset. The fine-tuning dataset is obtained and processed using the same methods as in steps S1.1, S1.2, and S1.3. The pre-trained model obtained in step S1.4 is then fine-tuned using the fine-tuning dataset. During the training process, the first three convolutional layers of the model are frozen to preserve general visual features, accelerate convergence time, and capture high-level semantic features. Finally, the model with the highest score is saved as the final fine-tuned model.
[0035] The processing flow for steps S3-S4 is as follows: Figure 3 As shown.
[0036] S3. Real-time acquisition of ore images from the target site, according to... Figure 2 After preprocessing as shown, an image that meets the format requirements of a deep learning model is obtained and the ore features are enhanced. Then, it is input into the final fine-tuned model trained in step S2 to infer the preliminary edge segmentation map of the ore.
[0037] S4. Perform OpenCV thresholding, morphological processing, distance transformation, and threshold labeling on the preliminary edge segmentation map obtained in step S3. Then, use the watershed segmentation algorithm with label guidance to segment the area and obtain the final ore region segmentation map, as shown in steps S4.1-S4.4. S4.1 After selecting a suitable threshold parameter through experiments, the thresholding operation of OpenCV is used on the preliminary edge segmentation map to obtain a binary map.
[0038] S4.2 uses OpenCV's morphological operations to repair the binary image, removing abnormal regions and improving the quality of the binary image.
[0039] S4.3, using distance variation and setting a distance threshold, the obtained binary map is processed to obtain a guide marker map of the ore region, which serves as the input for the next segmentation algorithm.
[0040] S4.4, use the watershed segmentation algorithm with marker guidance to segment the binary map to obtain the final ore region segmentation map.
[0041] S5. Perform statistical calculations on the final ore region segmentation map to obtain ore particle size data for each particle size class.
[0042] See attached document Figure 4-5 This embodiment provides an online ore particle size analysis system based on visual segmentation, including an image acquisition module, an image preprocessing module, an algorithm processing module, a data processing module, a data communication module, a display module, and a control module. The data flow and workflow between each module are as follows: Figure 6 As shown.
[0043] The image acquisition module includes an industrial camera and camera fixture. The industrial camera is mounted above the ore flow using the camera fixture to ensure stable and clear images. The installation location of the camera fixture needs to be surveyed in advance to ensure minimal dust and minimal changes in lighting conditions. Figure 5 As shown, the camera fixture includes a bracket 2. In this embodiment, the bracket 2 is a rectangular frame assembled from aluminum profiles. The four corners of the bracket 2 are connected to four height-adjustable steel wire ropes 3, suspending it above the ore flow. This installation method reduces the impact of vibration on the camera. A protective box 1 is bolted to the center of the bracket 2, and the protective box 1 can be adjusted vertically. The industrial camera is installed inside the protective box 1. By adjusting the installation position and height, it is ensured that the bracket 2 is parallel to the ore flow, and the camera lens inside the protective box 1 faces the ore flow on the conveyor belt. The vertical center of the camera lens coincides with the center of the actual ore distribution on the conveyor belt, ensuring that the industrial camera lens faces the central area of the ore. Strip LED light sources are installed on both sides of the bracket 2 for supplemental lighting, highlighting the ore contour features and reducing the influence of ambient light. After the camera is installed, clean high-pressure air is used to blow away dust in front of the lens to ensure clear images.
[0044] An electrical control box is installed within 3 meters of the camera fixture. The installation of the camera fixture and electrical control box should not interfere with maintenance. The electrical control box supplies power to the camera and has a built-in computing unit. The image preprocessing module, algorithm processing module, and data processing module are all installed on the computing unit within the electrical control box. The display module and control module are installed on the central control room workstation, which is responsible for displaying the processing results, communicating with other systems, and providing control suggestions.
[0045] The image preprocessing module uses OpenCV to preprocess the ore image to enhance image feature information. The algorithm processing module is used to train and store the deep learning semantic segmentation network model, and uses the trained final fine-tuned model and image segmentation method to perform edge segmentation and processing on the preprocessed ore image to obtain the final ore region segmentation map. The data processing module uses data analysis algorithms to perform statistical analysis on the final ore region segmentation map and to measure the ore particle size composition in real time.
[0046] The data communication module is used to realize communication between the electrical control box and the central control room workstation, and sends the images and data processed by the electrical control box's computing unit to the central control room workstation; the field electrical control box and the central control room workstation use optical fiber to transmit data, and the data communication module uses the TCP / IP protocol to send the processed images and data to the central control room workstation.
[0047] The display module is installed on the workstation in the central control room. It can process and store the received images and data, and display the data. Specifically, the display module can use fitting formula algorithms to further fit the received raw data and store it in the database. The user interface of the display module can display data fitting information, including data curves, ore segmentation images, fitting percentage values, and maximum ore particle size.
[0048] The control module is installed on the workstation in the central control room. The control module can analyze the ore particle size data to obtain intelligent control and adjustment suggestions, and supports multiple communication interfaces such as TCP / IP, HTTP, and Modbus / TCP, enabling it to exchange data with other intelligent control systems.
[0049] Compared with existing technologies, the advantages of this invention are: 1. By optimizing model training and image segmentation methods, it can accurately segment the edges of ore images, improving the accuracy of segmenting adjacent ore regions with similar gray levels and indistinct boundaries; 2. It can automatically perform ore flow analysis and statistically analyze ore particle size distribution data in real time, eliminating the need for manual screening and avoiding subjective human judgment; 3. The camera fixture designed in this invention is stable and reliable, not easily affected by vibration, ensuring clear and stable acquired images; 4. This invention can analyze ore particle size data, promptly detect large ore pieces, issue early warnings, propose control suggestions, assist in decision-making, and achieve linkage with intelligent control systems.
[0050] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.
Claims
1. A visual segmentation-based online ore particle size analysis method for analyzing the particle size of ore streams transported on belt conveyors, characterized in that, Includes the following steps: S1. Collect multiple ore images from different sites, working conditions, and time periods across different working conditions to establish a pre-training dataset; after preprocessing the images in the pre-training dataset and labeling the ore contour information, use the pre-training dataset to train a deep learning semantic segmentation network model to obtain a pre-training model. S2. Collect multiple ore images at different time periods at the target site to establish a fine-tuning dataset. The number of images in the fine-tuning dataset is less than the number of images in the pre-training dataset. After preprocessing the images in the fine-tuning dataset and labeling the ore contour information, the pre-trained model obtained in step S1 is fine-tuned using the fine-tuning dataset. During the training process, the first three network convolutional layers of the model are frozen, and the model with the highest score is saved as the final fine-tuning model. S3. Real-time acquisition of ore images from the target site, preprocessing, and input into the final fine-tuning model obtained in step S2 to infer the preliminary edge segmentation map of the ore. S4. Perform OpenCV thresholding, morphological processing, distance transformation and threshold labeling on the preliminary edge segmentation map of step S3, and then use the watershed segmentation algorithm with label guidance to segment to obtain the final ore region segmentation map. S5. Perform statistical calculations on the final ore region segmentation map to obtain ore particle size data for each particle size class.
2. The online ore particle size analysis method based on visual segmentation according to claim 1, characterized in that: The images in the fine-tuning dataset are derived from a single target site, and the number of images in the fine-tuning dataset is one-tenth that of the pre-training dataset.
3. The online ore particle size analysis method based on visual segmentation according to claim 1, characterized in that: In steps S1, S2 and S3, the image preprocessing process is as follows: first, the image is distorted and corrected; then, the RGB image is converted to grayscale and the image size is adjusted; then, smoothing filtering is performed; and finally, contrast enhancement is performed on the filtered image to increase the effective features of the mineral image.
4. The online ore particle size analysis method based on visual segmentation according to claim 1, characterized in that: In steps S1 and S2, before loading the images from the pre-training dataset or fine-tuning dataset into the model, data augmentation functions are used to perform random rotation, stitching, cropping, and normalization operations on the images to expand the dataset size.
5. The online ore particle size analysis method based on visual segmentation according to claim 1, characterized in that: Step S4 Specifically, the following steps are included: S4.1 Perform OpenCV thresholding and morphological processing on the preliminary edge segmentation map obtained in step S3 to obtain a binarized image; S4.
2. Using distance variation and setting a distance threshold, process the obtained binary map to obtain a guide marker map of the ore region; S4.
3. Use the marker-guided watershed segmentation algorithm to segment the binary map to obtain the final ore region segmentation map.
6. An online ore particle size analysis system based on visual segmentation, comprising an image acquisition module, an image preprocessing module, an algorithm processing module, a data processing module, a data communication module, a display module, and a control module; characterized in that: The image acquisition module includes an industrial camera and a camera fixture. The industrial camera is mounted above the ore flow via the camera fixture to dynamically acquire image data of the ore flow. An electrical control box is installed near the camera fixture, and the image preprocessing module, algorithm processing module, and data processing module are all installed on the computing unit in the electrical control box. The image preprocessing module is used to preprocess the ore image to enhance its feature information; the algorithm processing module is used to train and store the deep learning semantic segmentation network model, and use the trained final fine-tuned model and image segmentation method to perform edge segmentation and processing on the preprocessed ore image to obtain the final ore region segmentation map; the data processing module uses data analysis algorithms to perform statistical analysis on the final ore region segmentation map and to measure the ore particle size composition in real time. The data communication module is used to realize communication between the electrical control box and the central control room workstation, sending the images and data processed by the electrical control box's computing unit to the central control room workstation; the display module and the control module are respectively installed in the central control room workstation. The display module can process and store the received images and data, and display the data; the control module can exchange data with other intelligent control systems, analyze the ore particle size data, and issue intelligent control and adjustment suggestions.
7. The online ore particle size analysis system based on visual segmentation according to claim 6, characterized in that: The camera fixture includes a bracket (2), which is suspended above the ore flow by multiple steel wire ropes (3) evenly distributed around it. A protective box (1) with adjustable vertical angle is installed in the middle of the bracket (2). The industrial camera is installed in the protective box (1). The vertical center of the camera image coincides with the vertical center of the actual distribution of ore on the belt conveyor, and ensures that the incoming ore captured by the industrial camera remains horizontal. At the same time, clean high-pressure air is used to blow away the dust in front of the lens.
8. The online ore particle size analysis system based on visual segmentation according to claim 7, characterized in that: The camera fixture bracket (2) is a rectangular frame. The four corners of the bracket (2) are connected to four steel wire ropes (3) for hoisting. The height of the steel wire ropes (3) can be adjusted.
9. The online ore particle size analysis system based on visual segmentation according to claim 7, characterized in that: The bracket (2) is also equipped with a strip light source (4) parallel to the direction of belt conveyor operation. The two strip light sources (4) are located on both sides of the protective box (1) and are used to provide supplementary lighting for the ore flow.
10. The online ore particle size analysis system based on visual segmentation according to claim 6, characterized in that: The display module can use a fitting formula algorithm to fit the received raw data, and then store and display the data. The displayed information includes data curves, ore segmentation images, fitting percentage values, and maximum ore particle size.