Intelligent visual analysis and management system for picking and selecting process before iron making

Abstract

An intelligent visual analysis and management system for a picking and selecting process before iron making belongs to the technical field of picking and selecting management systems before iron making and comprises an image acquisition layer, a data storage layer, a preprocessing layer, an analysis model layer and a system management layer which are connected in sequence. According to the method, the production process state, the equipment state and the production operation safety in the production process are taken as the demand starting point, an intelligent visual analysis and management and control system is established, a specific management module and an algorithm analysis model for the pre-iron selection process are planned, and the pre-iron selection process, an artificial intelligence visual algorithm and a computer technology are combined; and intelligent visual analysis management of the selection process is realized.

Description

Technical Field

[0001] This invention belongs to the technical field of iron ore processing and beneficiation management systems, specifically relating to an intelligent visual analysis and management system for iron ore processing and beneficiation procedures. Background Technology

[0002] In the iron ore beneficiation process of the steel industry, there are very strict requirements for the production process status, equipment status, and production operation safety. The level of monitoring and management directly affects the quality of steel beneficiation products, production efficiency, equipment maintenance efficiency, and the safety of beneficiation workers.

[0003] Currently, the following challenges exist in the iron ore mining and beneficiation process: 1. Many production process conditions, such as the analysis of flotation foam conditions, usually rely on manual observation and identification by operators, and it is impossible to achieve comprehensive 24-hour uninterrupted monitoring.

[0004] 2. Some production process parameters, such as the particle size of the ore and the location information of the mining car, cannot be monitored and statistically analyzed using traditional measurement methods.

[0005] 3. Some operations related to iron ore processing and beneficiation cannot identify customized dangerous actions. Summary of the Invention

[0006] To address the aforementioned problems, this invention proposes an intelligent visual analysis and management system for iron ore processing, comprising an image acquisition layer, a data storage layer, a preprocessing layer, an analysis model layer, and a system management layer, which are sequentially connected.

[0007] Furthermore, the image acquisition layer is categorized by application type into process type, equipment status type, and production safety type; and according to specific operating conditions, it includes: image acquisition equipment, supplementary lighting equipment, and protective devices. Image acquisition equipment: Select appropriate frame rate and resolution specifications according to the working conditions. When selecting equipment, it is necessary to ensure the compatibility between accuracy and image data processing capabilities, and ensure the stable operation of the software platform. Supplemental lighting equipment: Select supplemental lights with appropriate illuminance according to the working conditions to ensure the light intensity for image acquisition; Cooling and protective devices: Select air cooling, water cooling, and waterproof, dustproof, and impact-resistant devices according to the working conditions.

[0008] Furthermore, the data storage layer includes the acquisition and storage of image data and text data; ①Image data Image data includes two types of services: surveillance images and alarm images; Image acquisition methods: camera video stream, subsystem video stream; Surveillance images: Data is read into the system via video stream; Alarm image: An image with marked boxes or marked text generated after being analyzed by the algorithm model; Both monitoring images and alarm images are broadly categorized into three types: process images, equipment status images, and production safety images. They are then further categorized based on specific functional algorithm models. Finally, the images are stored in the image database according to the image storage time requirements of the selected process.

[0009] Furthermore, the text data includes: alarm information from the image, and data read from other systems; Alarm information from images: The text or data information associated with the alarm image, and the text or data information parsed from the alarm image are formed after being analyzed by an algorithm model; Data read from other systems: Production-related text descriptions or data information read from other systems; Text or data information read from other systems is obtained through a data acquisition gateway. By configuring data acquisition items and protocols, it achieves interface integration with various field devices, PLCs, and DCS systems in the acquisition workshop, including the integration of heterogeneous systems, and performs unified and formatted encapsulation and storage of the acquired data. ③ Data Engine: Using the data bus as the engine, it connects and binds business image model data with actual collected data. It is responsible for storing the collected real-time data and the issued control command data. It is equipped with an SQL database for storing real-time data and a relational database for storing business process data. The system provides data backup, compression, and restoration services, a model data bus service that binds data to business models, and a data retrieval service for historical data.

[0010] Furthermore, after image information acquisition, the preprocessing layer employs distortion and aberration correction, grayscale transformation, exposure fusion network, Gaussian filtering, and contour extraction methods to preprocess the image according to the characteristics of various types of images. (1) Distortion and aberration correction Distortion correction is performed on multimodal images, while removing optical aberration pixel areas such as coma and spherical aberration at the edges of the original image, providing basic graphic conditions for subsequent recognition tasks; (2) Grayscale transformation A grayscale transformation algorithm is used to convert the image to grayscale. The algorithm aims to extract edges and contours, and color information is not used as a recognition target. The target image is first converted to grayscale to reduce the amount of computation and improve recognition accuracy and speed. (3) Dark light enhancement based on multi-exposure fusion network By using neural networks to generate and fuse multiple exposure images, and by enhancing the input image in low light conditions, the influence of the complex environment in the workshop on the recognition results is mitigated, and the detection accuracy is effectively improved. (4) Gaussian filtering The Gaussian function smooths the image while preserving its overall grayscale distribution characteristics. The Gaussian filtering algorithm is used for noise reduction. When Gaussian filtering smooths pixels in the neighborhood of an image, pixels at different positions in the neighborhood are assigned different weights. (5) Contour extraction The gradient of each pixel in the smoothed image is obtained by the convolution operator, and the gradient sum along the horizontal and vertical directions is obtained by using the operator.

[0011] Furthermore, the analytical model layer designs and constructs artificial intelligence visual algorithm models based on process categories, equipment status categories, and production safety categories, including providing an algorithm development environment and an algorithm integration environment; (1) Provide standard algorithm service packages It provides a standard iron ore processing algorithm model package library. Users can load the corresponding services by adjusting the process parameters to achieve adaptive applications. In addition, it provides algorithm development environments and mainstream tool services in languages ​​such as Python, C, and Java, so that processing engineers can easily develop specific algorithms within this system platform. The provided standard iron ore pre-harvesting algorithm model package includes: ① Mining and beneficiation process: Ore quality analysis model: Using a high frame rate and high resolution camera, ore images during the mining and beneficiation process are acquired and analyzed in real time. Based on the physical appearance conditions of the ore, such as color and texture, the physicochemical properties of the ore are analyzed, and the ore composition is analyzed based on the images. Ore particle size analysis model: A high-frame-rate, high-resolution camera is used to collect and analyze the ore particle size on the conveyor belt in the mining and beneficiation process in real time. The model is based on the physical appearance conditions of the ore, such as size and shape, and the proportion of different particle sizes of ore is statistically analyzed. Ore magnetic analysis model: A high-frame-rate, high-resolution camera is used to collect and analyze the magnetic properties of the ore on the conveyor belt in the mining and beneficiation process in real time. Based on the color distribution conditions of the ore, magnetic analysis and statistics are performed. Flotation foam analysis model: Using a high-definition camera, the foam status of the flotation machine is collected in real time, and information such as the foam generation rate, bubble size, bubble color, and movement trajectory direction is analyzed. The dosing and pulp information of the flotation process are analyzed, and then the flotation effect is comprehensively analyzed. Filtration effect analysis model: Using a high-definition camera, the state of the ore cake after filtration is collected in real time. Based on the physical appearance conditions of the filter cake, such as color and brightness, the dehydration rate of the filtration equipment is analyzed, and then the moisture content of the ore cake is analyzed. ② Device status parsing class: Vehicle Status Analysis Model: Employs high-speed, high-definition cameras to obtain real-time information on mining vehicles, including model, specifications, location, speed, and load capacity, and analyzes vehicle behavior such as driving, stopping, loading, and unloading. Visual positioning analysis model: Using a high-speed, high-definition camera, the surrounding field of vision of the mining vehicle is analyzed in real time during the forward or backward movement of the mining vehicle, and a model of the surrounding environment of the vehicle is established to achieve real-time visual positioning and environmental status modeling. Equipment operation analysis model: High-definition cameras are used to collect the operating status of the mining and sorting equipment in real time. Based on the operating characteristics of various mining and sorting equipment, the rotation, displacement, flipping, vibration, lifting and other operating actions of the equipment are analyzed, and the rationality of the equipment operating status is analyzed. Monitoring model for conveyor belt sorting machine: Real-time images of the conveyor belt sorting machine, real-time analysis of fault information such as tearing, deviation, slippage, and foreign objects on the belt, and real-time alarms; Inspection model for power distribution rooms: Real-time acquisition of images of each power distribution room, analysis of data from the acquisition panel instruments, information on the environment of the power distribution room, worker operations, etc., and real-time alarms; ③ Production Safety Analysis: Violation Analysis Model: Real-time acquisition of images of critical production environments; based on the characteristics of iron ore processing, analysis of violation information such as personnel accidentally entering the wrong area, misuse of tools, crossing equipment, and live operation, and real-time alarms are generated. Miner identification model: Through facial recognition, the information of miners is monitored in real time, and real-time alarms are triggered for abnormal information related to miners and work locations; Miner safety analysis model: Through intelligent vision, it monitors the wearing information of miners' safety helmets, work clothes and personal protective equipment in real time and issues real-time alarms; Fire identification and analysis model: Through intelligent vision technology, it can monitor the mining workshop, external network cable trays, and fire points in real time, analyze the fire situation in real time, and issue real-time alarms.

[0012] Furthermore, (2) Algorithm integration service: Build an algorithm running platform with mainstream programming languages ​​such as Python, C, and Java as development prototypes to provide import and application of externally written algorithms; (3) Algorithm call service The algorithm service layer provides an algorithm invocation service, with the following specific steps: The first step is to configure the algorithm. After the backend service starts, it will generate the calling thread, algorithm calling cycle, return value processing method, etc. for each algorithm according to the algorithm list configured in the business interface, and start the engine of each algorithm. The second step is that within each algorithm engine, the timer engine will periodically start an algorithm call cycle and call the corresponding method of the algorithm service; The third step is to send the result feedback value of the algorithm service to the algorithm recycling module. The algorithm recycling module will automatically update the data bus according to the data bus point corresponding to the result value and record the sent result.

[0013] Furthermore, the system management layer serves as the interactive layer of the system platform, providing business support for the system management and mining workers. The system provides a comprehensive real-time video interface for mining, a comprehensive image library interface for mining, a comprehensive alarm management interface for mining, and a mining interlocking control interface for mining. (1) Selecting a comprehensive real-time video interface The integrated real-time video interface provides a complete real-time video stream system for the ironmaking process, and enables smooth playback within a BS architecture. Functionally, it is divided into: a process vision system, an ironmaking vision system, and a safety vision system. In addition to providing video streams processed by intelligent vision algorithms, it also interfaces with the industrial television system in the ironmaking process and provides configurable window interface services to enable customized calling and display of the entire plant's video within this system. (2) Selecting the comprehensive image library interface Provides regular qualitative query services for all images; The application provides two drop-down lists. The first drop-down list allows you to select a workshop, and the second drop-down list displays the corresponding equipment. Selecting a device in the drop-down list will display images of normal production and malfunctions related to the selected device in the list. For common image faults, we provide standard fault analysis and handling suggestions packages, and offer automatic diagnostic services. It also provides data entry services, and users can also compile fault cause analysis and handling procedures based on their own work experience, so that similar faults can be handled in accordance with regulations in the future. (3) Select the integrated alarm management interface The application provides two drop-down lists. The first drop-down list selects the workshop, and the second drop-down list displays the corresponding equipment. Selecting a device in the drop-down list will display all the maintained equipment parameters under the selected device in the list. Real-time curves or historical curves of the relevant equipment parameters can be displayed. (4) Select the interlocking control management interface This interface parses image data, analyzes and generates optimized operations or necessary interlocking shutdown controls during the parsing process, provides feedback control opinions on production parameters, implements data communication services in the background, and provides emergency handling services in the event of a production accident. ④ Mining and beneficiation process: Ore quality control strategy: Analyze the physical and chemical properties of ore, analyze ore composition based on images, and provide mining process parameter adjustment strategy services; Ore particle size control strategy: Analyze the ore particle size on the conveyor belt in the mining and beneficiation process, and statistically analyze the proportion of different particle sizes of ore, and provide crushing process parameter adjustment strategy services; Ore Magnetic Control Strategy: Analyze the magnetic properties of ore on conveyor belts in mining and beneficiation processes, and provide mining process parameter adjustment strategy services; Flotation foam control strategy: Analyze information such as foam generation speed, size, and direction of movement, and then comprehensively analyze the flotation effect, and provide services for adjusting mineral processing equipment parameters and reagent dosage; Filtration effect analysis model: Analyzes the filtration dehydration rate, then analyzes the moisture content of the ore cake, and provides filtration equipment parameter adjustment strategy services; ⑤ Device status parsing class: Vehicle status analysis and dispatch: Real-time monitoring of mining vehicle model, specifications, vehicle location, speed, and load information, and analysis of vehicle behavior to provide data support for intelligent vehicle dispatching; Visual positioning and analysis control: Analyze the surrounding field of vision of mining vehicles when they move forward or backward, establish real-time visual positioning and models, and provide data support for autonomous driving of vehicles; Equipment operation analysis and control: High-definition cameras are used to collect the operating status of the screening equipment in real time, analyze the rationality of the operating status, and provide parameter adjustment strategy services for the filtration equipment. ① Production Safety Analysis: The system analyzes and alarms images related to safety violations, and can implement a virtual electronic fence function according to production management needs; the system is interconnected and organically integrated with the video surveillance system; it interlocks corresponding processing methods, and tracks the responsible person and the cause of the accident through video.

[0014] Furthermore, the system management layer provides a system management module, which includes a unified entry management application, an account and role management application, and a system operation log application; The unified entry management application provides a third-party single sign-on link setting. After logging in through the iron ore processing and control platform, users can directly click to enter the third-party system from the entry interface. The account role management application can maintain account information, including adding, modifying, and deleting account information, and can maintain role information. Role information determines permissions. After binding an account with a role, permission restrictions on the account can be implemented. The relationship between roles and accounts is one-to-many. The system operation log application saves logs generated from both automatic and manual operations in the system. Automatic operations include call information for external and internal interfaces, including whether the operation was successful and error messages after failure. Manual operations include event records triggered by all buttons.

[0015] The beneficial effects of this invention are as follows: An intelligent visual analysis and management system for iron ore beneficiation processes is established based on the needs of three aspects: production process status, equipment status, and production operation safety. The system plans specific management modules and algorithm analysis models for iron ore beneficiation processes, and combines iron ore beneficiation processes, artificial intelligence visual algorithms, and computer technology to achieve intelligent visual analysis and management of the beneficiation process.

[0016] Its advantages are: 1. Achieve comprehensive analysis and control of three types of vision: process, equipment, and safety. 2. The system's functions are deeply integrated with iron ore processing technology, making it highly targeted and practical. 2. Modular design, supporting the integration of other vision systems and image data storage, with unified management; 3. Supports data communication with other management and control systems, breaking down data silos. Attached Figure Description

[0017] Figure 1 This is a diagram of the intelligent vision platform architecture of the present invention. Detailed Implementation

[0018] To make the technical means and objectives of this invention easier to understand, the invention is further described below with reference to specific embodiments, such as an intelligent visual analysis and management system for iron ore pre-mining and beneficiation processes. Figure 1 As shown, the system framework includes an image acquisition layer, a data storage layer, a preprocessing layer, a parsing model layer, and a system management layer.

[0019] 1) Image acquisition layer The image acquisition layer is categorized by application type into process type, equipment status type, and production safety type. Based on specific practical operating conditions, it includes: image acquisition equipment, supplementary lighting equipment, and protective devices, etc.

[0020] Image acquisition equipment: Select appropriate frame rate and resolution specifications according to the working conditions. When selecting equipment, it is necessary to ensure the compatibility between accuracy and image data processing capabilities, and ensure the stable operation of the software platform. Supplemental lighting equipment: Select supplemental lights with appropriate illuminance according to the working conditions to ensure the light intensity for image acquisition; Cooling and protective devices: Select air cooling, water cooling, and waterproof, dustproof, and impact-resistant devices according to the working conditions.

[0021] 2) Data storage layer The data storage layer includes the acquisition and storage of image data and text data.

[0022] ①Image data Image data includes two types of services: surveillance images and alarm images.

[0023] Image acquisition methods: camera video stream, subsystem video stream.

[0024] Surveillance images: Data is read into the system via video stream; Alarm image: An image with marked boxes or marked text generated after being analyzed by the algorithm model.

[0025] Both monitoring and alarm images are broadly categorized into three types: process images, equipment status images, and production safety images. Further subcategorization is performed based on specific functional algorithm models. Finally, images within a certain timeframe are stored in the image database according to the image storage time requirements of the selected process. Alarm images have longer storage times, while general images have shorter storage times.

[0026] ②Text data Text data includes: alarm information from images, and data read from other systems. Alarm information from images: The text or data information associated with the alarm image, and the text or data information parsed from the alarm image are formed after being analyzed by an algorithm model.

[0027] Data read from other systems: Production-related text descriptions or data information read from other systems. This text or data information is obtained through a data acquisition gateway. By configuring data acquisition items and protocols, it achieves interface integration with various field devices, PLCs, and DCS systems in the acquisition workshop, including integration with heterogeneous systems. The acquired data is then uniformly and formatted for encapsulation and storage.

[0028] ③ Data Engine Using a data bus as the engine, it connects and binds business image model data with actual collected data, and is responsible for storing the collected real-time data and the issued control command data. It is equipped with an SQL database for storing real-time data and a relational database for storing business process data.

[0029] The system provides data backup, compression, and restoration services, a model data bus service that binds data to business models, and a data retrieval service for historical data.

[0030] 3) Pre-processing layer After image information is acquired, distortion and aberration correction, grayscale transformation, exposure fusion network, Gaussian filtering contour extraction and other methods are used to preprocess the images according to the characteristics of the acquired images. The preprocessing method is selected based on the different processing techniques.

[0031] (1) Distortion and aberration correction Distortion correction is performed on multimodal images, while removing optical aberration pixel areas such as coma and spherical aberration at the edges of the original image, providing basic graphic conditions for subsequent recognition tasks.

[0032] (2) Grayscale transformation A grayscale transformation algorithm is used to convert the image to grayscale. The algorithm aims to extract edges and contours, and color information is not used as a recognition target. Therefore, the target image is first converted to grayscale to reduce the amount of computation and improve recognition accuracy and speed.

[0033] (3) Dark light enhancement based on multi-exposure fusion network By using neural networks to generate and fuse multiple exposure images, and by enhancing the input image in low light conditions, the influence of the complex environment in the workshop on the recognition results is mitigated, effectively improving the detection accuracy.

[0034] (4) Gaussian filtering Edge detection is susceptible to noise in images, thus requiring noise reduction processing. The Gaussian function smooths the image, preserving more of its overall grayscale distribution characteristics. When using a Gaussian filtering algorithm for noise reduction, different weights are assigned to pixels at different locations within the image's neighborhood as the smoothing process.

[0035] (5) Contour extraction The location of the strongest grayscale intensity change in an image is the gradient direction. The gradient of each pixel in the smoothed image can be obtained by the convolution operator, and it is advisable to use the operator to obtain the gradient sum along the horizontal and vertical directions.

[0036] 4) Parsing Model Layer The described analytical model layer designs and constructs artificial intelligence visual algorithm models based on process categories, equipment status categories, and production safety categories, including providing algorithm development and integration environments.

[0037] (1) Provide standard algorithm service packages It provides a standard iron ore processing algorithm model package library, and users can load the corresponding services by adjusting the process parameters to achieve adaptive applications. In addition, it provides algorithm development environments and mainstream tool services in languages ​​such as Python, C, and Java, so that application processing workers can easily develop specific algorithms within this system platform.

[0038] The provided standard iron ore pre-harvesting algorithm model package includes: ① Mining and beneficiation process: Ore quality analysis model: Using a high-frame-rate, high-resolution camera (starlight level), ore images during the mining and beneficiation process are acquired and analyzed in real time. Based on the physical appearance conditions such as the color and texture of the ore, the physicochemical properties of the ore are analyzed, and the ore composition is analyzed based on the images.

[0039] Ore particle size analysis model: A high-frame-rate, high-resolution camera (starlight level) is used to collect and analyze the ore particle size on the conveyor belt in the mining and beneficiation process in real time. The model is based on the physical appearance conditions of the ore, such as size and shape, and the proportion of different particle sizes of ore is statistically analyzed.

[0040] Ore magnetic analysis model: A high-frame-rate, high-resolution camera (starlight level) is used to collect and analyze the magnetic properties of the ore on the conveyor belt in the mining and beneficiation process in real time. The magnetic analysis and statistics are performed based on the color distribution conditions of the ore.

[0041] Flotation foam analysis model: Using a high-definition camera, the foam status of the flotation machine is collected in real time, and information such as foam generation rate, bubble size, bubble color, and movement trajectory direction is analyzed. The dosing and pulp information of the flotation process are analyzed, and then the flotation effect is comprehensively analyzed.

[0042] Filtration effect analysis model: Using a high-definition camera, the state of the filtered cake is collected in real time. Based on the physical appearance conditions of the filter cake, such as color and brightness, the dehydration rate of the filtration equipment is analyzed, and then the moisture content of the cake is analyzed.

[0043] ② Device status parsing class: Vehicle Status Analysis Model: Employs high-speed, high-definition cameras to collect real-time information on mining vehicles, including model, specifications, location, speed, and load capacity, and analyzes vehicle behavior such as driving, stopping, loading, and unloading.

[0044] Visual positioning analysis model: Using a high-speed, high-definition camera, the surrounding field of vision of the mining vehicle is analyzed in real time during the forward or backward movement of the mining vehicle, and a model of the surrounding environment of the vehicle is established to achieve real-time visual positioning and environmental status modeling.

[0045] Equipment operation analysis model: High-definition cameras are used to collect the operating status of the mining and sorting equipment in real time. Based on the operating characteristics of various mining and sorting equipment, the rotation, displacement, flipping, vibration, lifting and other operating actions of the equipment are analyzed, and the rationality of the equipment operating status is analyzed.

[0046] The monitoring model for the conveyor belt sorting machine provides real-time images of the machine, analyzes fault information such as tearing, misalignment, slippage, and foreign objects on the belt, and issues real-time alarms.

[0047] Inspection model for power distribution rooms: Real-time acquisition of images of each power distribution room, analysis of data from the acquisition panel instruments, information on the environment of the power distribution room, worker operations, etc., and real-time alarms.

[0048] ③ Production Safety Analysis: Violation Analysis Model: Real-time acquisition of images of critical production environments; analysis of violation information such as unauthorized entry of vehicles and personnel, misuse of tools, crossing of equipment, and operation with live wires, based on the characteristics of iron ore processing; and real-time alarms.

[0049] Miner identification model: Using facial recognition technology, the system monitors miners' information in real time and issues real-time alerts for any abnormalities in miners' or their work locations.

[0050] Miner safety analysis model: Through intelligent vision technology, it monitors the wearing information of miners' safety helmets, work clothes and personal protective equipment in real time and issues real-time alarms.

[0051] Fire identification and analysis model: Through intelligent vision technology, it can monitor the mining workshop, external network cable trays, and fire points in real time, analyze the fire situation in real time, and issue real-time alarms.

[0052] (2) Algorithm Integration Service We will build an algorithm execution platform that uses mainstream programming languages ​​such as Python, C, and Java as development prototypes to provide import and application of externally written algorithms.

[0053] (3) Algorithm call service The algorithm service layer provides an algorithm invocation service, with the following specific steps: The first step is to configure the algorithm. After the backend service starts, it will generate the calling thread, algorithm calling cycle, return value processing method, etc. for each algorithm based on the algorithm list configured in the business interface, and start the engine of each algorithm.

[0054] The second step involves the timer engine within each algorithm engine periodically initiating an algorithm call cycle and calling the corresponding method of the algorithm service.

[0055] The third step is to send the result feedback value of the algorithm service to the algorithm recycling module. The algorithm recycling module will automatically update the data bus according to the data bus point corresponding to the result value and record the sent result.

[0056] 5) System Management Layer The system management layer, as the interactive layer that provides business support for the system management and mining workers, provides a comprehensive real-time video interface for mining, a comprehensive image library interface for mining, a comprehensive alarm management interface for mining, and a mining interlocking control interface for mining.

[0057] (1) Selecting a comprehensive real-time video interface The integrated real-time video interface provides a complete real-time video stream system for the ironmaking process, and enables smooth playback within a BS architecture. Functionally, it is divided into: an ironmaking process vision system, an ironmaking vision system, and an ironmaking safety vision system. In addition to providing video streams processed by intelligent vision algorithms, it also interfaces with the industrial television system in the ironmaking process and provides configurable window interface services to enable customized calling and display of the entire plant's video within this system.

[0058] (2) Selecting the comprehensive image library interface Provides regular qualitative query services for all images.

[0059] The application provides two drop-down lists. The first drop-down list allows you to select a workshop, and the second drop-down list displays the corresponding equipment. Selecting a device in the drop-down list will display images of its normal production and malfunctions. For common image faults, we provide standard fault analysis and handling suggestions, and offer automated diagnostic services. It also provides data entry services, and users can also compile fault cause analysis and handling procedures based on their own work experience, so that similar faults can be handled in accordance with regulations in the future.

[0060] (3) Select the integrated alarm management interface The application provides two drop-down lists. The first drop-down list allows you to select a workshop, and the second drop-down list displays the corresponding equipment. Selecting a device in the drop-down list will display all the maintained equipment parameters for the selected device in the list, and can display real-time or historical curves of the relevant equipment parameters.

[0061] (4) Select the interlocking control management interface This interface parses image data, generates optimized operations or necessary interlocking shutdown controls during the parsing process, provides feedback control opinions on production parameters, implements data communication services in the background, and provides emergency handling services in the event of a production accident.

[0062] ④ Mining and beneficiation process: Ore quality control strategy: Analyze the physical and chemical properties of ore, analyze ore composition based on images, and provide mining process parameter adjustment strategies.

[0063] Ore particle size control strategy: Analyze the ore particle size on the conveyor belt in the mining and beneficiation process, statistically analyze the proportion of different particle sizes, and provide crushing process parameter adjustment strategy services.

[0064] Ore Magnetic Control Strategy: Analyze the magnetic properties of ore on conveyor belts in mining and beneficiation processes, and provide mining process parameter adjustment strategies.

[0065] Flotation foam control strategy: Analyze information such as foam generation speed, size, and movement direction to comprehensively analyze the flotation effect and provide services for adjusting mineral processing equipment parameters and reagent dosage.

[0066] Filtration effect analysis model: Analyzes the filtration dehydration rate, then analyzes the moisture content of the ore cake, and provides filtration equipment parameter adjustment strategy services.

[0067] ⑤ Device status parsing class: Vehicle status analysis and scheduling: Real-time monitoring of information such as the model, specifications, location, speed, and load of mining vehicles, and analysis of vehicle behavior to provide data support for intelligent vehicle scheduling.

[0068] Visual positioning and analysis control: Analyze the surrounding field of vision of mining vehicles when they move forward or backward, establish real-time visual positioning and models, and provide data support for autonomous driving.

[0069] Equipment operation analysis and control: Utilizing high-definition cameras, the system collects real-time data on the operating status of the filtration equipment, analyzes the rationality of the operating status, and provides parameter adjustment strategies for the filtration equipment.

[0070] ① Production Safety Analysis: The system analyzes and alarms images related to safety violations, such as identifying those not wearing safety helmets, intrusion into areas, not wearing work clothes, smoking in the factory, working while talking on the phone, working at heights without safety belts, hazardous operations, and flame detection. It can also implement a virtual electronic fence function according to production management needs. The system is interconnected and integrated with the video surveillance system, interlocking corresponding handling methods to track responsible persons and the cause of accidents via video.

[0071] (5) System permissions The system management layer provides a system management module, which includes a unified entry management application, an account and role management application, and a system operation log application.

[0072] The unified entry management application provides a third-party single sign-on link setting. After logging in through the iron ore processing and control platform, users can directly click to enter the third-party system from the entry interface.

[0073] The account role management application can maintain account information, including adding, modifying, and deleting account information, and can also maintain role information. Role information determines permissions. After binding an account with a role, permission restrictions on the account can be implemented. There is a one-to-many relationship between roles and accounts.

[0074] The system operation log application saves logs generated from both automatic and manual operations in the system. Automatic operations include call information for external and internal interfaces, including whether the operation was successful and error messages after failure. Manual operations include event records triggered by all buttons.

[0075] 6) System hardware requirements The main computing power consumption of this system is in the operation of video data processing and video parsing related algorithms. It is advisable to use a GPU-based server. The GPU computing power should be determined according to the capacity of the model and the video stream. It is advisable to use one front-end operation station. If independent expansion is carried out, the graphics card processing power of the independent operation station should be increased accordingly.

[0076] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An intelligent visual analysis and management system for iron ore pre-mining and beneficiation processes, characterized in that, It includes an image acquisition layer, a data storage layer, a preprocessing layer, a parsing model layer, and a system management layer, which are connected sequentially.

2. The intelligent visual analysis and management system for iron ore pre-mining and beneficiation processes as described in claim 1, characterized in that, The image acquisition layer is categorized by application type into process type, equipment status type, and production safety type; and according to specific operating conditions, it includes: image acquisition equipment, supplementary lighting equipment, and protective devices. Image acquisition equipment: Select appropriate frame rate and resolution specifications according to the working conditions. When selecting equipment, it is necessary to ensure the compatibility between accuracy and image data processing capabilities, and ensure the stable operation of the software platform. Supplemental lighting equipment: Select supplemental lights with appropriate illuminance according to the working conditions to ensure the light intensity for image acquisition; Cooling and protective devices: Select air cooling, water cooling, and waterproof, dustproof, and impact-resistant devices according to the working conditions.

3. The intelligent visual analysis and management system for iron ore pre-mining and beneficiation processes as described in claim 1, characterized in that, The data storage layer includes the acquisition and storage of image data and text data; ①Image data Image data includes two types of services: surveillance images and alarm images; Image acquisition methods: camera video stream, subsystem video stream; Surveillance images: Data is read into the system via video stream; Alarm image: An image with marked boxes or marked text generated after being analyzed by the algorithm model; Both monitoring images and alarm images are broadly categorized into three types: process images, equipment status images, and production safety images. They are then further categorized based on specific functional algorithm models. Finally, the images are stored in the image database according to the image storage time requirements of the selected process.

4. The intelligent visual analysis and management system for iron ore pre-mining and beneficiation processes as described in claim 1, characterized in that, The text data includes: alarm information from the image and data read from other systems; Alarm information from images: The text or data information associated with the alarm image, and the text or data information parsed from the alarm image are formed after being analyzed by an algorithm model; Data read from other systems: Production-related text descriptions or data information read from other systems; Text or data information read from other systems is obtained through a data acquisition gateway. By configuring data acquisition items and protocols, it achieves interface integration with various field devices, PLCs, and DCS systems in the acquisition workshop, including the integration of heterogeneous systems, and performs unified and formatted encapsulation and storage of the acquired data. ③ Data Engine: Using the data bus as the engine, it connects and binds business image model data with actual collected data. It is responsible for storing the collected real-time data and the issued control command data. It is equipped with an SQL database for storing real-time data and a relational database for storing business process data. The system provides data backup, compression, and restoration services, a model data bus service that binds data to business models, and a data retrieval service for historical data.

5. The intelligent visual analysis and management system for iron ore pre-mining and beneficiation processes as described in claim 1, characterized in that, After image information is acquired, the preprocessing layer performs image preprocessing based on the characteristics of various types of images by employing distortion and aberration correction, grayscale transformation, exposure fusion network, Gaussian filtering, and contour extraction methods. (1) Distortion and aberration correction Distortion correction is performed on multimodal images, while removing optical aberration pixel areas such as coma and spherical aberration at the edges of the original image, providing basic graphic conditions for subsequent recognition tasks; (2) Grayscale transformation A grayscale transformation algorithm is used to convert the image to grayscale. The algorithm aims to extract edges and contours, and color information is not used as a recognition target. The target image is first converted to grayscale to reduce the amount of computation and improve recognition accuracy and speed. (3) Dark light enhancement based on multi-exposure fusion network By using neural networks to generate and fuse multiple exposure images, and by enhancing the input image in low light conditions, the influence of the complex environment in the workshop on the recognition results is mitigated, and the detection accuracy is effectively improved. (4) Gaussian filtering The Gaussian function smooths the image while preserving its overall grayscale distribution characteristics. The Gaussian filtering algorithm is used for noise reduction. When Gaussian filtering smooths pixels in the neighborhood of an image, pixels at different positions in the neighborhood are assigned different weights. (5) Contour extraction The gradient of each pixel in the smoothed image is obtained by the convolution operator, and the gradient sum along the horizontal and vertical directions is obtained by using the operator.

6. The intelligent visual analysis and management system for iron ore pre-mining and beneficiation processes as described in claim 1, characterized in that, The analytical model layer designs and constructs artificial intelligence visual algorithm models based on process categories, equipment status categories, and production safety categories, including providing an algorithm development environment and an algorithm integration environment; (1) Provide standard algorithm service packages It provides a standard iron ore processing algorithm model package library. Users can load the corresponding services by adjusting the process parameters to achieve adaptive applications. In addition, it provides algorithm development environments and mainstream tool services in languages ​​such as Python, C, and Java, so that processing engineers can easily develop specific algorithms within this system platform. The provided standard iron ore pre-harvesting algorithm model package includes: ① Mining and beneficiation process: Ore quality analysis model: Using a high frame rate and high resolution camera, ore images during the mining and beneficiation process are acquired and analyzed in real time. Based on the physical appearance conditions of the ore, such as color and texture, the physicochemical properties of the ore are analyzed, and the ore composition is analyzed based on the images. Ore particle size analysis model: A high-frame-rate, high-resolution camera is used to collect and analyze the ore particle size on the conveyor belt in the mining and beneficiation process in real time. The model is based on the physical appearance conditions of the ore, such as size and shape, and the proportion of different particle sizes of ore is statistically analyzed. Ore magnetic analysis model: A high-frame-rate, high-resolution camera is used to collect and analyze the magnetic properties of the ore on the conveyor belt in the mining and beneficiation process in real time. Based on the color distribution conditions of the ore, magnetic analysis and statistics are performed. Flotation foam analysis model: Using a high-definition camera, the foam status of the flotation machine is collected in real time, and information such as the foam generation rate, bubble size, bubble color, and movement trajectory direction is analyzed. The dosing and pulp information of the flotation process are analyzed, and then the flotation effect is comprehensively analyzed. Filtration effect analysis model: Using a high-definition camera, the state of the ore cake after filtration is collected in real time. Based on the physical appearance conditions of the filter cake, such as color and brightness, the dehydration rate of the filtration equipment is analyzed, and then the moisture content of the ore cake is analyzed. ② Device status parsing class: Vehicle Status Analysis Model: Employs high-speed, high-definition cameras to collect real-time information on mining vehicles, including model, specifications, location, speed, and load capacity, and analyzes vehicle behavior such as driving, stopping, loading, and unloading. Visual positioning analysis model: Using a high-speed, high-definition camera, the surrounding field of vision of the mining vehicle is analyzed in real time during the forward or backward movement of the mining vehicle, and a model of the surrounding environment of the vehicle is established to achieve real-time visual positioning and environmental status modeling. Equipment operation analysis model: High-definition cameras are used to collect the operating status of the mining and sorting equipment in real time. Based on the operating characteristics of various mining and sorting equipment, the rotation, displacement, flipping, vibration, lifting and other operating actions of the equipment are analyzed, and the rationality of the equipment operating status is analyzed. Sampling conveyor belt monitoring model: Real-time images of the sampling conveyor belt, real-time analysis of fault information such as tearing, belt deviation, slippage, and foreign objects on the belt, and real-time alarms; Inspection model for power distribution rooms: Real-time acquisition of images of each power distribution room, analysis of data from the acquisition panel instruments, information on the environment of the power distribution room, worker operations, etc., and real-time alarms; ③ Production Safety Analysis: Violation Analysis Model: Real-time acquisition of images of critical production environments; based on the characteristics of iron ore processing, analysis of violation information such as personnel accidentally entering the wrong area, misuse of tools, crossing equipment, and live operation, and real-time alarms are generated. Miner identification model: Through facial recognition, the information of miners is monitored in real time, and real-time alarms are triggered for abnormal information related to miners and work locations; Miner safety analysis model: Through intelligent vision, it monitors the wearing information of miners' safety helmets, work clothes and personal protective equipment in real time and issues real-time alarms; Fire identification and analysis model: Through intelligent vision technology, it can monitor the mining workshop, external network cable trays, and fire points in real time, analyze the fire situation in real time, and issue real-time alarms.

7. The intelligent visual analysis and management system for iron ore pre-mining and beneficiation processes as described in claim 6, characterized in that, (2) Algorithm integration service: Build an algorithm running platform with mainstream programming languages ​​such as Python, C, and Java as development prototypes to provide import and application of externally written algorithms; (3) Algorithm call service The algorithm service layer provides an algorithm invocation service, with the following specific steps: The first step is to configure the algorithm. After the backend service starts, it will generate the calling thread, algorithm calling cycle, return value processing method, etc. for each algorithm according to the algorithm list configured in the business interface, and start the engine of each algorithm. The second step is that within each algorithm engine, the timer engine will periodically start an algorithm call cycle and call the corresponding method of the algorithm service; The third step is to send the result feedback value of the algorithm service to the algorithm recycling module. The algorithm recycling module will automatically update the data bus according to the data bus point corresponding to the result value and record the sent result.

8. The intelligent visual analysis and management system for iron ore pre-mining and beneficiation processes as described in claim 1, characterized in that, The system management layer serves as the interactive layer for the system platform to provide business support for the system management and mining workers. The system provides a comprehensive real-time video interface for mining, a comprehensive image library interface for mining, a comprehensive alarm management interface for mining, and a mining interlocking control interface for mining. (1) Selecting a comprehensive real-time video interface The integrated real-time video interface provides a complete real-time video stream system for the ironmaking process, and enables smooth playback within a BS architecture. Functionally, it is divided into: a process vision system, an ironmaking vision system, and a safety vision system. In addition to providing video streams processed by intelligent vision algorithms, it also interfaces with the industrial television system in the ironmaking process and provides configurable window interface services to enable customized calling and display of the entire plant's video within this system. (2) Selecting the comprehensive image library interface Provides regular qualitative query services for all images; The application provides two drop-down lists. The first drop-down list allows you to select a workshop, and the second drop-down list displays the corresponding equipment. Selecting a device in the drop-down list will display images of normal production and malfunctions related to the selected device in the list. For common image faults, we provide standard fault analysis and handling suggestions, and offer automatic diagnostic services. It also provides data entry services, and users can also compile fault cause analysis and handling procedures based on their own work experience, so that similar faults can be handled in accordance with regulations in the future. (3) Select the integrated alarm management interface The application provides two drop-down lists. The first drop-down list selects the workshop, and the second drop-down list displays the corresponding equipment. Selecting a device in the drop-down list will display all the maintained equipment parameters under the selected device in the list. Real-time curves or historical curves of the relevant equipment parameters can be displayed. (4) Select the interlocking control management interface This interface parses image data, analyzes and generates optimized operations or necessary interlocking shutdown controls during the parsing process, provides feedback control opinions on production parameters, implements data communication services in the background, and provides emergency handling services in the event of a production accident. ④ Mining and beneficiation process: Ore quality control strategy: Analyze the physical and chemical properties of ore, analyze ore composition based on images, and provide mining process parameter adjustment strategy services; Ore particle size control strategy: Analyze the ore particle size on the conveyor belt in the mining and beneficiation process, and statistically analyze the proportion of different particle sizes of ore, and provide crushing process parameter adjustment strategy services; Ore Magnetic Control Strategy: Analyze the magnetic properties of ore on conveyor belts in mining and beneficiation processes, and provide mining process parameter adjustment strategy services; Flotation foam control strategy: Analyze information such as foam generation speed, size, and direction of movement, and then comprehensively analyze the flotation effect, and provide services for adjusting mineral processing equipment parameters and reagent dosage; Filtration effect analysis model: Analyzes the filtration dehydration rate, then analyzes the moisture content of the ore cake, and provides filtration equipment parameter adjustment strategy services; ⑤ Device status parsing class: Vehicle status analysis and dispatch: Real-time monitoring of mining vehicle model, specifications, vehicle location, speed, and load information, and analysis of vehicle behavior to provide data support for intelligent vehicle dispatching; Visual positioning and analysis control: Analyze the surrounding field of vision of mining vehicles when they move forward or backward, establish real-time visual positioning and models, and provide data support for autonomous driving of vehicles; Equipment operation analysis and control: High-definition cameras are used to collect the operating status of the screening equipment in real time, analyze the rationality of the operating status, and provide parameter adjustment strategy services for the filtration equipment. ① Production Safety Analysis: The system analyzes and alarms images related to safety violations, and can implement a virtual electronic fence function according to production management needs; the system is interconnected and organically integrated with the video surveillance system; it interlocks corresponding processing methods, and tracks the responsible person and the cause of the accident through video.

9. The intelligent visual analysis and management system for iron ore pre-mining and beneficiation processes as described in claim 8, characterized in that, The system management layer provides a system management module, which includes a unified entry management application, an account and role management application, and a system operation log application. The unified entry management application provides a third-party single sign-on link setting. After logging in through the iron ore processing and control platform, users can directly click to enter the third-party system from the entry interface. The account role management application can maintain account information, including adding, modifying, and deleting account information, and can maintain role information. Role information determines permissions. After binding an account with a role, permission restrictions on the account can be implemented. The relationship between roles and accounts is one-to-many. The system operation log application saves logs generated from both automatic and manual operations in the system. Automatic operations include call information for external and internal interfaces, including whether the operation was successful and the error information after failure. Manual operations include event records triggered by all buttons.

Images