Multi-scene fusion fuel management centralized video monitoring and warning method and related device

Abstract

The invention discloses a multi-scene fusion fuel management centralized video monitoring and warning method and a related device, and belongs to the technical field of fire coal operation management. The method comprises the following steps: collecting multi-dimensional environment data and operation behavior data in an operation process; carrying out fusion analysis on the collected multi-dimensional environment data and operation behavior data, and identifying an abnormal operation behavior; and according to the identified abnormal operation behavior, executing a corresponding grading early warning and equipment control instruction. Based on a block chain technology, full-process operation data and events from data acquisition to alarm response are subjected to evidence storage, and a credible traceability evidence chain is formed. According to the invention, a centralized monitoring system with multi-dimensional perception, full-chain analysis and intelligent linkage is constructed, and a closed-loop supervision capability covering the whole process of collection-transportation-sample preparation-assay is formed.

Description

Technical Field

[0001] This invention belongs to the field of coal-fired operation management technology, and relates to a multi-scenario integrated centralized video monitoring and alarm method and related device for fuel management. Background Technology

[0002] In the energy industry, fuel quality management (such as coal) is a core element in ensuring production efficiency and cost control. The accuracy of fuel collection, sampling, and testing is paramount, directly impacting trade settlement and combustion efficiency. However, in practice, these processes involve human intervention and lack effective oversight, posing a risk of fraud. To address this challenge, existing technologies widely employ video surveillance systems and rudimentary intelligent analysis techniques, but their application in complex industrial scenarios still suffers from systemic shortcomings.

[0003] First, traditional monitoring systems primarily rely on visible light video surveillance, which faces significant challenges in core fuel management scenarios—such as sample preparation workshops. These environments typically feature high dust concentrations, strong mechanical vibrations, and dense metal equipment. Existing single-visible-light monitoring solutions perform poorly under these conditions: high dust levels cause light scattering, resulting in blurred images and loss of target details; mechanical vibrations cause continuous video jitter, leading to missing key operation frames and hindering stable temporal analysis; and dense metal equipment interferes with wireless signal transmission, affecting the reliability of sensor data. Second, at the behavioral recognition and analysis level, existing technologies are mostly based on single-frame analysis of two-dimensional images or simple temporal threshold judgments. These methods are severely inadequate in analyzing behaviors where operators deliberately evade monitoring (such as turning their backs to the camera, sideways obstruction, or using tools for cover). Monocular vision-based skeletal joint detection algorithms are susceptible to changes in viewing angle and have limited accuracy in analyzing micro-postures such as shoulder joint rotation and torso twisting, making it difficult to distinguish between compliant operations and disguised violations. Meanwhile, existing algorithms lack a deep semantic understanding of continuous operational processes, failing to transform discrete operational actions (such as coal shoveling, sorting, and bottling) into quantifiable node events conforming to standard operating procedures. Currently, various monitoring subsystems in fuel management (such as sampling point monitoring, transportation monitoring, sample preparation workshop monitoring, and laboratory data) typically adopt an independent deployment and divide-and-conquer approach. This results in inconsistent data standards across systems and a lack of effective spatiotemporal alignment and correlation analysis mechanisms. Video stream data, equipment vibration data, tool trajectory data, and final test results data are fragmented, making it impossible to construct a complete causal chain of "physical operation - equipment status - sample quality." When an anomaly occurs in one link, the system struggles to trace the root cause of the problem upstream and cannot conduct cross-link risk linkage early warning. Furthermore, the architecture of existing systems is often rigid, and insufficient edge computing capabilities lead to lagging intelligent analysis, failing to achieve sub-second real-time responses; while cloud-based models have long update cycles, making it difficult to quickly adapt to new fraudulent methods. Data silos between different plant areas also hinder knowledge sharing and collaborative defense against violations, leaving room for organized and systematic fraudulent activities. Furthermore, the monitoring data is disconnected from business logic, and the stored evidence is mostly raw video or simple event tags, lacking the associated encapsulation of multi-dimensional information such as operational context, device status, and environmental parameters. This results in a fragmented and poorly interpretable chain of evidence. When disputes arise or audits are required, it is difficult to provide clear, complete, tamper-proof, and highly persuasive technical evidence.

[0004] In summary, the existing fuel management monitoring technologies are limited by the singularity of their sensing means, the shallowness of behavior analysis, the fragmentation of system architecture, and the imperfection of the traceability system, and it is difficult to cope with the increasingly concealed and intelligent fraud risks in complex industrial scenarios. Therefore, there is an urgent need in this field for an operation monitoring method that can integrate multi-scene information, perform in-depth intelligent analysis, achieve real-time linkage response, and build a full-process credible evidence chain. Summary of the Invention

[0005] The purpose of the present invention is to provide a multi-scene fusion fuel management centralized video monitoring and warning method and related devices to solve the technical problems of poor recognition accuracy of abnormal operations and lack of closed-loop supervision ability in the existing technology.

[0006] To achieve the above object, the present invention adopts the following technical solutions: In the first aspect, the present invention provides a multi-scene fusion fuel management centralized video monitoring and warning method, including the following steps: Collect multi-dimensional environmental data and operation behavior data during the operation process; Perform fusion analysis on the collected multi-dimensional environmental data and operation behavior data to identify abnormal operation behaviors; Execute corresponding hierarchical warning and equipment control instructions according to the identified abnormal operation behaviors.

[0007] Further, the step of collecting multi-dimensional environmental data and operation behavior data during the operation process specifically includes: Collect image data of the target area through a visible light camera; Collect temperature distribution data of the target area through an infrared thermal imager; Collect point cloud data through a millimeter wave radar and construct a three-dimensional point cloud space field; Collect the running vibration spectrum data of key sample preparation equipment through a vibration sensor; Collect centimeter-level accuracy displacement trajectory data of operation tools and personnel through UWB positioning.

[0008] Further, the step of performing fusion analysis on the collected multi-dimensional environmental data and operation behavior data to identify abnormal operation behaviors specifically includes: Preprocess the collected multi-dimensional environmental data and operation behavior data; Based on the preprocessed operation behavior data, through binocular vision three-dimensional reconstruction technology, analyze the joint rotation angles and micro-gesture features of the operator to obtain the first feature data; Based on the preprocessed multi-dimensional environmental data, use a temporal convolutional network to analyze the continuous temporal features of the operation rhythm and tool movement trajectory to obtain the second feature data; By combining the first and second feature data, a spatiotemporal correlation map of personnel, equipment, and materials is constructed. Correlation analysis and causal inference are performed on multi-source abnormal features to identify abnormal operational behaviors.

[0009] Furthermore, the step of preprocessing the collected multidimensional environmental data and operational behavior data specifically includes: When the environmental dust concentration is detected to exceed the preset threshold, the dynamic spectral compensation and non-uniform illumination enhancement algorithm is activated for the image data. To address the video jitter problem caused by mechanical vibration, an electronic image stabilization mechanism based on six-axis inertial sensing is used to perform electronic image stabilization processing on video stream data. Spatiotemporal registration was performed on millimeter-wave radar point cloud data and data collected by infrared thermal imagers to construct a fused three-dimensional motion trajectory model.

[0010] Furthermore, the step of executing corresponding graded early warning and equipment control commands based on the identified abnormal operating behavior specifically includes: When a transient anomaly is detected, a Level 1 warning is activated, and visual corrective guidance is projected to on-site operators using augmented reality devices; When multiple pieces of evidence confirm the existence of a risk of violation, a Level 2 response is initiated, and the key operating parameters of the sample preparation equipment are locked through the industrial control protocol. When a systemic fraud risk is detected, a level 3 alarm is activated, blockchain evidence is stored, and the alarm event is pushed to the higher-level regulatory platform. At the same time, a hard blocking operation is performed on the device.

[0011] Furthermore, the method also includes: based on blockchain technology, storing and verifying the operational data and events throughout the entire process from data collection to alarm response, forming a trusted traceability evidence chain.

[0012] Furthermore, the step of storing and verifying operational data and events throughout the entire process from data collection to alarm response based on blockchain technology to form a trusted and traceable evidence chain specifically includes: Generate a unique digital fingerprint for each fuel sample; The operator's biometric data, equipment operating parameters, environmental data, and abnormal event tags are associated and encapsulated with the digital fingerprint to form an operational evidence chain; Establish a quantitative mapping model between sample quality indicators and operational compliance; When abnormal results occur during the testing process, the abnormal results are traced back to the specific violation process based on the quantitative mapping model and the operation evidence chain stored on the blockchain.

[0013] Secondly, the present invention provides a multi-scenario integrated centralized video monitoring and alarm system for fuel management, comprising: The data acquisition module is used to collect multi-dimensional environmental data and operational behavior data during the operation process; The anomaly detection module is used to fuse and analyze the collected multidimensional environmental data and operational behavior data to identify abnormal operational behaviors. The alarm module is used to execute corresponding graded early warnings and equipment control commands based on the identified abnormal operating behaviors.

[0014] Thirdly, the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the multi-scenario fusion fuel management centralized video monitoring and alarm method described above.

[0015] Fourthly, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the multi-scenario fusion fuel management centralized video monitoring and alarm method described above.

[0016] Compared with the prior art, the present invention has the following beneficial effects: This invention discloses a multi-scenario integrated centralized video monitoring and alarm method and related device for fuel management. By collecting multi-dimensional environmental data and operational behavior data during operations, it can comprehensively and accurately acquire various key information in fuel management scenarios, providing a rich and reliable data foundation for subsequent analysis. Secondly, the collected data is fused and analyzed to identify abnormal operational behaviors, overcoming the limitations of single data source analysis and greatly improving the accuracy and timeliness of abnormal behavior identification. This effectively avoids fuel management safety accidents caused by human negligence or misoperation, ensuring operational safety. Based on the identified abnormal operational behaviors, corresponding graded early warning and equipment control commands are executed. The graded early warning mechanism enables staff to take rapid and targeted countermeasures according to the severity of the anomaly; thereby improving the stability and reliability of the entire fuel management system, reducing operation and maintenance costs, and improving management efficiency, providing an efficient, intelligent, and safe solution for fuel management in multiple scenarios.

[0017] Furthermore, this invention integrates the complementary advantages of visible light, infrared thermal imaging, and millimeter-wave radar to construct a three-dimensional perception system that penetrates dust, vibration, and electromagnetic interference. Through dynamic spectral compensation and non-uniform illumination enhancement algorithms, the light scattering effect in high-dust environments is eliminated, ensuring clear discernibility of operational details. A millimeter-wave radar array constructs a three-dimensional point cloud spatial field, penetrating visual blind spots to capture the trajectory of micro-movements of the human body, forming a spatiotemporal data redundancy verification with the perceived temperature distribution from infrared thermal imaging. UWB centimeter-level positioning technology tracks the displacement path of tools in real time, and combined with the vibration spectrum characteristics of the equipment, establishes a digital twin comparison model of tool usage standards, effectively identifying covert operations such as illegal opening of sample containers and abnormal vibration of the divider.

[0018] Furthermore, unlike traditional two-dimensional pose estimation methods, this system constructs a four-dimensional spatiotemporal correlation analysis model (three-dimensional space + time dimension), accurately analyzing joint rotation angles and operation trajectories through binocular vision 3D reconstruction technology. An improved skeletal joint detection algorithm incorporates inverse dynamics principles, overcoming viewpoint occlusion limitations and accurately identifying avoidance postures such as side-stepping and tool occlusion. A temporal convolutional network establishes a dynamic baseline model, quantifying temporal features such as operation rhythm and tool interaction frequency into a compliance index, achieving millisecond-level identification of fraud patterns such as abnormal stagnation and accelerated score reduction. A knowledge graph-driven case library continuously optimizes feature weights through transfer learning, enabling the system to adaptively identify novel fraud methods.

[0019] Furthermore, lightweight spatiotemporal feature extraction models are deployed at edge nodes to achieve sub-second initial screening of abnormal behavior; a federated learning platform is built in the cloud to aggregate data from multiple plants to train high-precision identification models, achieving knowledge evolution through encrypted parameter sharing. This mechanism, while ensuring data privacy, reduces the response time for identifying new fraud patterns from weekly to hourly. The tiered early warning system is deeply coupled with the equipment control system. When critical violations are detected, hard blocking measures such as locking equipment parameters and sealing sample containers can be automatically triggered, forming a complete control chain of "identification-response-handling".

[0020] Furthermore, by applying lightweight blockchain technology, diverse data such as operator biometrics, equipment operating conditions, and environmental parameters are encapsulated into immutable digital fingerprints. The test results of each coal sample are automatically linked to the spatiotemporal trajectory data of the sampling and sample preparation processes. A partial least squares regression model is used to establish a quantitative mapping relationship between operational compliance and sample quality, enabling two-way traceability of abnormal detection results. When violations occur, the augmented reality assistance system projects three-dimensional compliance action guidance through AR glasses, forming a closed-loop on-site intervention system of "violation warning - standard demonstration - process correction." Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a schematic diagram of the system of the present invention; Figure 3 This is a system flowchart of the multi-scenario integrated fuel management centralized video monitoring and AI alarm platform according to an embodiment of the present invention; Figure 4 This is a flowchart illustrating the implementation process of the multi-scenario integrated fuel management centralized video monitoring and AI alarm platform according to an embodiment of the present invention. Detailed Implementation

[0023] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0024] The following detailed description is exemplary and intended to provide further detailed explanation of the invention. Unless otherwise specified, all technical terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terminology used in this invention is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention.

[0025] See Figure 1 This invention addresses the need for fraud risk prevention in the coal sampling, preparation, and analysis process. The technical solution constructs a centralized monitoring system with multi-dimensional perception, full-chain analysis, and intelligent linkage, forming a closed-loop regulatory capability covering the entire process of "collection-transportation-sample preparation-analysis." This solution overcomes the technical bottlenecks of traditional single-scenario monitoring by achieving accurate identification and real-time blocking of concealed fraudulent activities through multi-modal data fusion, spatiotemporal correlation analysis, and adaptive decision-making. A multi-scenario integrated centralized video monitoring and alarm method for fuel management is disclosed, including the following steps: S1 collects multi-dimensional environmental data and operational behavior data during the operation process; Image data of the target area is acquired using a visible light camera; Temperature distribution data of the target area is collected using an infrared thermal imager; Point cloud data is collected using millimeter-wave radar to construct a three-dimensional point cloud spatial field; Vibration spectrum data of key sample preparation equipment were collected using vibration sensors. UWB positioning is used to collect centimeter-level precision displacement trajectory data of operating tools and personnel.

[0026] This embodiment establishes a heterogeneous sensing network integrating visible light, infrared thermal imaging, millimeter-wave radar, and vibration sensing, forming a holographic mapping capability for physical space. In the high-dust area of ​​the sample preparation workshop, dynamic spectral compensation technology is used to eliminate the refraction interference of suspended particles, and a non-uniform illumination enhancement algorithm is used to improve the clarity of the target outline. The millimeter-wave radar array constructs a three-dimensional point cloud spatial field, penetrating dust to capture the trajectory of human movement, forming a spatiotemporal complement with the perceived temperature distribution data from infrared thermal imaging. To address the video jitter problem caused by mechanical vibration, a six-axis inertial sensor-assisted electronic image stabilization mechanism is designed, achieving sub-pixel-level image stabilization through inverse motion vector compensation. The UWB positioning system tracks the displacement trajectory of tools with centimeter-level accuracy, establishing a digital twin comparison model between the tool usage path and standard operating procedures.

[0027] S2, integrates and analyzes the collected multidimensional environmental data and operational behavior data to identify abnormal operational behaviors; S201, Preprocess the collected multidimensional environmental data and operational behavior data; 1) When the environmental dust concentration is detected to exceed the preset threshold, the dynamic spectral compensation and non-uniform illumination enhancement algorithm is activated for the image data; 2) To address the video jitter problem caused by mechanical vibration, an electronic image stabilization mechanism based on six-axis inertial sensing is used to perform electronic image stabilization processing on the video stream data; 3) Spatiotemporal registration of millimeter-wave radar point cloud data and data collected by infrared thermal imager is performed to construct a fused three-dimensional motion trajectory model.

[0028] S202, based on the preprocessed operational behavior data, the joint rotation angle and micro-posture features of the operator are analyzed through binocular vision 3D reconstruction technology to obtain the first feature data; S203, based on the preprocessed multidimensional environmental data, uses a temporal convolutional network to analyze the continuous temporal characteristics of the operation rhythm and tool movement trajectory, and obtains the second feature data; S204. Combining the first feature data and the second feature data, a spatiotemporal correlation map of personnel, equipment and materials is constructed. Correlation analysis and causal inference are performed on multi-source abnormal features to identify abnormal operation behaviors.

[0029] Specifically, the core algorithm layer of this embodiment adopts a hierarchical analysis architecture, with the bottom layer deploying a real-time posture estimation model based on improved skeletal joint detection. A human kinematic model is constructed using binocular vision 3D reconstruction technology, and the rotation angles of the shoulder and elbow joints are analyzed using inverse dynamics algorithms to accurately identify avoidance postures such as side occlusion and deliberate back-facing. The temporal analysis module introduces a temporal convolutional network to capture temporal features such as abnormal operation rhythms and tool timeouts, establishing a dynamic baseline library of normal operation modes. In the advanced semantic understanding layer, a spatiotemporal correlation graph of "personnel-equipment-materials" is constructed, transforming discrete sensor data into a chain of business events with causal relationships. For example, when the vibration spectrum of the divider deviates from the baseline mode, the system automatically associates the operator's posture data with the tool's movement trajectory, and uses an attention mechanism to weightedly fuse multi-dimensional abnormal features to identify complex fraud patterns such as deliberately accelerated divider reduction and sample permutation. A dual-channel redundant processing architecture is designed to address the dynamic interference characteristics of industrial scenarios. Under normal operating conditions, visible light video stream serves as the main analysis channel, while infrared and radar data serve as verification sources. When dust concentration exceeds the standard or lighting conditions deteriorate, the system automatically switches to millimeter-wave radar-dominated mode, generating a human skeleton topology through point cloud density clustering. An adversarial training framework is established, introducing adversarial examples such as tool reflection and pose camouflage during the model training phase to enhance the algorithm's robustness under complex interference. An edge-cloud collaborative federated learning mechanism is designed, where local models in each plant share learning results through encrypted gradient parameters, ensuring data privacy while continuously improving model generalization capabilities and effectively responding to the rapid evolution of new fraud methods.

[0030] S3, based on the identified abnormal operating behavior, execute corresponding graded early warning and equipment control commands.

[0031] When a transient anomaly is detected, a Level 1 warning is activated, and visual corrective guidance is projected to on-site operators using augmented reality devices; When multiple pieces of evidence confirm the existence of a risk of violation, a Level 2 response is initiated, and the key operating parameters of the sample preparation equipment are locked through the industrial control protocol. When a systemic fraud risk is detected, a level 3 alarm is activated, blockchain evidence is stored, and the alarm event is pushed to the higher-level regulatory platform. At the same time, a hard blocking operation is performed on the device.

[0032] Specifically, this embodiment constructs a three-tiered risk warning system: Level 1 warning targets transient anomalies, guiding on-site correction through visual cues projected by AR glasses; Level 2 response links multi-dimensional confirmatory evidence, triggering equipment control commands to lock key parameters of the sample preparation machine; Level 3 alarm addresses systemic risks, automatically initiating blockchain evidence storage and pushing it to the group's monitoring platform. The response mechanism is deeply integrated with the physical control system. When an illegal opening of the sample container is detected, the PLC linkage device immediately freezes the electromagnetic lock of the hatch; if an abnormal increase in the reduction speed is detected, the system automatically writes read-only registers to protect the equipment's operating parameters. An augmented reality assistance system is designed to overlay 3D reconstructed compliance action guidance within the operator's field of vision when a violation occurs, providing real-time alerts for action deviations through visually salient markings.

[0033] S4, based on blockchain technology, stores and proves all operational data and events throughout the entire process from data collection to alarm response, forming a trusted traceability evidence chain.

[0034] Generate a unique digital fingerprint for each fuel sample; The operator's biometric data, equipment operating parameters, environmental data, and abnormal event tags are associated and encapsulated with the digital fingerprint to form an operational evidence chain; Establish a quantitative mapping model between sample quality indicators and operational compliance; When abnormal results occur during the testing process, the abnormal results are traced back to the specific violation process based on the quantitative mapping model and the operation evidence chain stored on the blockchain.

[0035] This embodiment constructs a full-link evidence storage mechanism encompassing "data acquisition - feature extraction - event determination - handling feedback," employing lightweight blockchain technology to ensure the immutability of operation logs. Each coal sample generates a unique digital fingerprint, linked to multi-dimensional information such as the biometrics of the collection personnel, equipment operating data, and environmental parameters. Mass spectrometry analysis data from the testing phase is automatically cross-validated with the sample preparation process recorded in video. A partial least squares regression model is used to establish a quantitative relationship between physical operations and sample quality, enabling reverse tracing of abnormal test results. A knowledge graph-driven case library is established to deconstruct patterns and extract features from historical violations, forming a transferable set of fraud pattern recognition rules.

[0036] See Figure 2This invention discloses a multi-scenario integrated centralized video monitoring and alarm system for fuel management, comprising a data acquisition module, an anomaly identification module, and an alarm module. The data acquisition module collects multi-dimensional environmental data and operational behavior data during operations; the anomaly identification module performs fusion analysis on the collected multi-dimensional environmental data and operational behavior data to identify abnormal operational behaviors; and the alarm module executes corresponding tiered warnings and equipment control commands based on the identified abnormal operational behaviors. This invention penetrates industrial environmental interference through multi-modal perception and decodes operational intentions through spatiotemporal correlation analysis, constructing a complete technological ecosystem of "multi-dimensional perception - intelligent judgment - real-time control - continuous evolution." Compared to traditional monitoring systems, it achieves qualitative breakthroughs in anomaly identification sensitivity, complex behavior resolution, and system response real-time performance, establishing a new paradigm of intelligent supervision for fuel quality management in the energy industry and driving the industry's digital transformation into a new stage of explainability, traceability, and evolution.

[0037] Example: See Figure 3 and Figure 4 This embodiment discloses an implementation method for the centralized monitoring platform for fuel management of China Huaneng Group, and describes a multi-scenario integrated centralized video monitoring and AI alarm platform for fuel management, as detailed below: I. System Architecture and Hardware Deployment In the entire process of fuel management sampling, preparation and testing at China Huaneng Group, the platform has built a three-dimensional monitoring network that integrates "end-edge-cloud".

[0038] 1. Multimodal sensing layer: Sampling process: The unmanned sampling vehicle is equipped with a Hikvision DS-2CD7A26G0 / P-IZHSY dual-spectrum gimbal, which integrates a FLIR A65 infrared thermal imaging module to collect coal seam images and temperature distribution data in real time; In the transportation phase: Coal trucks are equipped with TI IWR6843 millimeter-wave radar (60GHz) and UWB positioning tags (Decawave DW3000) to construct a 3D point cloud monitoring field for the truck bed; in the sample preparation workshop, a 4.5m×3.2m sample preparation table is equipped with a Baslerace2 industrial camera (5 megapixels) and a FLIR AX8 thermal imager diagonally, with vibration sensors (sensitivity 10mV / g) and millimeter-wave radar arrays embedded in the side walls, and operating tools have built-in UWB tags (accuracy ±3cm); in the coal sample testing laboratory, a ThermoScientific iCAP 7400 mass spectrometer with an integrated data fingerprint module is used to directly connect the raw spectral data to the blockchain evidence storage system, and an NVIDIA Jetson AGX is deployed in the sample preparation workshop. Using Orin edge nodes, running a YOLOv7+HRNet hybrid model, the following functions are achieved: real-time parsing of the coordinates of 17 skeletal joints of the human body (accuracy ±5 pixels); calculation of the Hausdorff distance between the tool displacement trajectory and the standard path (update rate 30Hz); and vibration spectrum FFT analysis (frequency resolution 0.1Hz).

[0039] II. Implementation Path of Key Technologies 1. Multimodal data fusion: When the dust concentration in the sample preparation workshop is >50mg / m³, a cross-modal compensation mechanism is activated: dust noise is eliminated by non-local mean filtering, and the adaptive Retinex algorithm is used to improve image contrast (PSNR value is improved by 8dB); millimeter-wave radar point cloud and infrared thermal imaging data are spatiotemporally registered to construct a three-dimensional model of human motion trajectory (positioning error <2cm); a six-axis inertial sensor (MPU-6050) is used to assist electronic image stabilization to eliminate image jitter caused by mechanical vibration (compensation accuracy 0.1°).

[0040] 2. Intelligent Analysis Engine: Spatiotemporal Relationship Graph Construction: Based on the Neo4j graph database, establish the entity relationship of "personnel-tools-equipment". When the following association anomalies are detected, an alarm is triggered: the operator's shoulder joint rotation angle is >45° and lasts for 3 seconds (calculated by inverse kinematics algorithm); the standard deviation of the sampling shovel movement speed exceeds the limit (UWB trajectory analysis); the vibration main frequency offset of the divider is >2Hz (FFT spectrum analysis).

[0041] 3. LSTM-Attention Fusion Model: Weighted fusion of abnormal features across different stages. For example, when there is a temporal correlation between sudden changes in coal quality at the sampling point and abnormal vibration during the sample preparation stage, the model automatically increases the risk level.

[0042] III. Tiered Response Mechanism: 1. Level 1 warning: The AR glasses (Microsoft HoloLens 2) project visual guidance to correct the tool holding angle deviation (such as the sample shovel tilt angle >10°). 2. Secondary response: Lock the operation permissions of the sample making machine through the OPC UA protocol. For example, when the back-to-monitor timeout exceeds 5 seconds and the tool trajectory is abnormal, the PLC immediately freezes the speed adjustment function of the splitter. 3. Three-level processing: The blockchain evidence storage system (Hyperledger Fabric) automatically records the multimodal evidence chain (including millimeter-wave point cloud, thermal imaging data, and vibration spectrum) and pushes it to the group's monitoring platform within 3 seconds.

[0043] IV. Typical Application Scenarios Anti-sampling measures: At the Zhuneng Coal Mine sampling point, the system detected a 12-meter offset in the sampling coordinates (threshold ±5 meters). Infrared thermal imaging showed abnormal coal seam temperature distribution (temperature difference >3℃). The edge node triggered a drone verification within 500ms, confirming that the sampling point had been forged. The blockchain recorded the evidence of the violation and froze the sample, preventing an economic loss of 800,000 yuan.

[0044] Sample preparation tamper-proofing: An operator deliberately turned their back to the monitoring system (for 6 seconds) during sample reduction. The system triggered a level 3 response based on the following multi-dimensional judgments: HRNet detected a shoulder joint rotation angle of 58° (threshold 45°); UWB trajectory showed a standard deviation of 2.8 cm / s for the sample shovel's movement speed (benchmark 1.2 cm / s); the vibration spectrum's dominant frequency suddenly increased from 35 Hz to 40 Hz. The PLC immediately locked the sample preparation machine, AR glasses projected a 3D reconstruction of the violation, and blockchain-stored evidence data provided a complete chain of evidence for subsequent accountability.

[0045] V. Implementation Results and Innovations Enhanced risk control capabilities: The accuracy rate for identifying covert fraud reached 98.5% (compared to ≤75% for traditional solutions), with response time reduced to 300ms; a total of 17 violations were blocked, reducing economic losses by 21 million yuan. A pioneering multimodal spatiotemporal correlation graph technology was developed to achieve causal reasoning based on "physical operation - equipment status - sample quality"; a federated learning-driven adaptive model was developed, improving the identification rate of new types of fraud by 63% within 6 months; the lightweight blockchain evidence storage system supports 2000 transactions per second (TPS), with a stored data volume of 45TB.

[0046] This platform achieves a leapfrog upgrade in coal management from "manual sampling" to "full-chain intelligent control" through multi-scenario perception fusion and intelligent decision-making closed loop, providing a benchmark example for the digital transformation of the energy industry. At the same time, through a dual-channel redundant architecture and intelligent learning mechanism, it maintains a detection accuracy rate of >90% even under dust concentration >100mg / m³ or extreme light conditions. The model autonomously iterates and optimizes its ability to identify new types of fraud every month. In the case of fraud in coal sample collection and analysis at China Huaneng Group, it reshapes the fuel management process with intelligent means, effectively curbs fraud risks, promotes the transformation of energy industry supervision from manual sampling to full-chain intelligent control, and provides an innovative practice example for the digital transformation of the industry.

[0047] In one embodiment of the present invention, a computer device is provided, comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in the computer storage medium to achieve a corresponding method flow or corresponding function. The processor described in this embodiment of the present invention can be used in the operation of a multi-scenario integrated fuel management centralized video monitoring and alarm method.

[0048] This invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, this storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the multi-scenario fusion fuel management centralized video monitoring and alarm method in the above embodiments.

[0049] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0050] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0051] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0052] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0053] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A multi-scenario integrated centralized video monitoring and alarm method for fuel management, characterized in that, Includes the following steps: Collect multidimensional environmental and operational data during the operation process; The collected multidimensional environmental data and operational behavior data are fused and analyzed to identify abnormal operational behaviors; Based on the identified abnormal operational behavior, execute corresponding graded early warning and equipment control commands.

2. The multi-scenario integrated centralized video monitoring and alarm method for fuel management according to claim 1, characterized in that, The steps involved in collecting multidimensional environmental data and operational behavior data during the data acquisition process specifically include: Image data of the target area is acquired using a visible light camera; Temperature distribution data of the target area is collected using an infrared thermal imager; Point cloud data is collected using millimeter-wave radar to construct a three-dimensional point cloud spatial field; Vibration spectrum data of key sample preparation equipment were collected using vibration sensors. UWB positioning is used to collect centimeter-level precision displacement trajectory data of operating tools and personnel.

3. The multi-scenario integrated centralized video monitoring and alarm method for fuel management according to claim 1, characterized in that, The steps for fusing and analyzing the collected multidimensional environmental data and operational behavior data to identify abnormal operational behaviors specifically include: Preprocess the collected multidimensional environmental data and operational behavior data; Based on the preprocessed operational behavior data, the joint rotation angles and micro-posture features of the operator are analyzed using binocular vision 3D reconstruction technology to obtain the first feature data; Based on the preprocessed multidimensional environmental data, a temporal convolutional network is used to analyze the continuous temporal characteristics of the operation rhythm and tool movement trajectory to obtain the second feature data. By combining the first and second feature data, a spatiotemporal correlation map of personnel, equipment, and materials is constructed. Correlation analysis and causal inference are performed on multi-source abnormal features to identify abnormal operational behaviors.

4. The multi-scenario integrated centralized video monitoring and alarm method for fuel management according to claim 3, characterized in that, The steps for preprocessing the collected multidimensional environmental data and operational behavior data specifically include: When the environmental dust concentration is detected to exceed the preset threshold, the dynamic spectral compensation and non-uniform illumination enhancement algorithm is activated for the image data. To address the video jitter problem caused by mechanical vibration, an electronic image stabilization mechanism based on six-axis inertial sensing is used to perform electronic image stabilization processing on video stream data. Spatiotemporal registration was performed on millimeter-wave radar point cloud data and data collected by infrared thermal imagers to construct a fused three-dimensional motion trajectory model.

5. The multi-scenario integrated centralized video monitoring and alarm method for fuel management according to claim 1, characterized in that, The step of executing corresponding graded early warning and equipment control commands based on the identified abnormal operation behavior specifically includes: When a transient anomaly is detected, a Level 1 warning is activated, and visual corrective guidance is projected to on-site operators using augmented reality devices; When multiple pieces of evidence confirm the existence of a risk of violation, a Level 2 response is initiated, and the key operating parameters of the sample preparation equipment are locked through the industrial control protocol. When a systemic fraud risk is detected, a level 3 alarm is activated, blockchain evidence is stored, and the alarm event is pushed to the higher-level regulatory platform. At the same time, a hard blocking operation is performed on the device.

6. The multi-scenario integrated centralized video monitoring and alarm method for fuel management according to claim 1, characterized in that, Also includes: Based on blockchain technology, operational data and events throughout the entire process from data collection to alarm response are stored and evidenced to form a trusted and traceable evidence chain.

7. The multi-scenario integrated centralized video monitoring and alarm method for fuel management according to claim 6, characterized in that, The steps of storing and verifying operational data and events throughout the entire process from data collection to alarm response, based on blockchain technology, to form a trusted and traceable chain of evidence, specifically include: Generate a unique digital fingerprint for each fuel sample; The operator's biometric data, equipment operating parameters, environmental data, and abnormal event tags are associated and encapsulated with the digital fingerprint to form an operational evidence chain; Establish a quantitative mapping model between sample quality indicators and operational compliance; When abnormal results occur during the testing process, the abnormal results are traced back to the specific violation process based on the quantitative mapping model and the operation evidence chain stored on the blockchain.

8. A multi-scenario integrated centralized video monitoring and alarm system for fuel management, characterized in that, include: The data acquisition module is used to collect multi-dimensional environmental data and operational behavior data during the operation process; The anomaly detection module is used to fuse and analyze the collected multidimensional environmental data and operational behavior data to identify abnormal operational behaviors. The alarm module is used to execute corresponding graded early warnings and equipment control commands based on the identified abnormal operating behaviors.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the multi-scenario fusion fuel management centralized video monitoring and alarm method as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the multi-scenario fusion fuel management centralized video monitoring and alarm method as described in any one of claims 1-7.

Images