Intelligent mine unattended system

Abstract

The invention discloses an intelligent mine unattended system, relates to the technical field of mine intelligent metering, and aims to solve the problem of weighing cheating in mineral aggregate transportation weighing, especially the scene of overweight of the whole vehicle when the vehicle is not loaded with mineral aggregate. The system comprises an anti-cheating sensing layer, a secure transmission layer, an intelligent anti-cheating analysis layer and a linkage disposal layer. The anti-cheating sensing layer acquires multi-dimensional anti-cheating data through encryption weighing, vehicle holographic monitoring, multimode identity authentication, environmental state monitoring and no-load feature acquisition units; the secure transmission layer adopts an industrial Ethernet + 5G private network double-link redundancy architecture to realize national secret encryption transmission; the intelligent anti-cheating analysis layer completes cheating study and judgment through data fusion, a multi-dimensional cheating recognition model, block chain evidence storage and a cheating behavior knowledge base; and the linkage disposal layer realizes cheating vehicle locking, early warning pushing and evidence retention. The system improves the authenticity of weighing data, reduces the economic loss of a mine, and is suitable for an unattended weighing scene of the mine.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent measurement of mines, in particular to an intelligent mine unattended system. BACKGROUND

[0002] In the production and operation of mines, the weighing of the transported ore is the core link of measuring the ore output, accounting for the cost and ensuring the enterprise benefit. With the popularization of unattended weighing technology, the problems of low efficiency and high safety risk of traditional manual weighing are alleviated, but the cheating behaviors in weighing are diversified and hidden, mainly including: 1. Signal interference cheating, shielding or tampering with the output signal of the weighing sensor through an electronic jammer, resulting in distorted weighing data; 2. Physical cheating, such as incomplete weighing of the mine car, partial load parking, intermediate unloading or adding removable counterweights; 3. Data tampering cheating, modifying the weighing record by cracking the system interface or forging the identity; 4. Vehicle information cheating, confusing the identity of the mine car by using fake plates and replacing RFID tags, and falsely reporting the weight of the ore; 5. Empty load cheating, making the whole vehicle overweight by adding temporary counterweights (such as water tanks, sandbags, metal blocks) or carrying non-ore weights when the vehicle is not loaded with ore, in order to falsely increase the net weight of the subsequent ore weighing. This cheating method is highly concealed, and the traditional system relying only on single weighing data is difficult to identify.

[0003] The existing anti-cheating means has obvious deficiencies: it only relies on a single hardware anti-cheating device, which is difficult to deal with multiple types of cheating behaviors; it lacks comprehensive monitoring of the dynamic state of the vehicle and the weighing process, and cannot identify hidden cheating such as partial load and incomplete weighing; especially for the cheating scenario of the whole vehicle being overweight when the vehicle is not loaded with ore, there is no effective historical data comparison and empty load characteristic analysis mechanism, resulting in frequent occurrence of this type of cheating behavior; the security protection of the weighing data storage and transmission link is weak, and it is easy to be tampered with and difficult to trace. These problems make it difficult to guarantee the measurement reliability of the unattended weighing system, causing huge economic losses to the mine enterprises. Therefore, it is an urgent technical problem to develop an anti-cheating intelligent mine unattended system integrating multi-dimensional monitoring, intelligent analysis and encrypted tracing. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides an intelligent mine unattended system.

[0005] To achieve the above purpose, the present application provides the following technical solution: an intelligent mine unattended system, comprising an anti-cheating perception layer, a secure transmission layer, an intelligent anti-cheating analysis layer and a linkage disposal layer.

[0006] The anti-cheating perception layer includes an encrypted weighing unit, a vehicle holographic monitoring unit, a multi-mode identity authentication unit, an environment state monitoring unit, and an empty load feature acquisition unit; the encrypted weighing unit acquires encrypted weighing data and interference signal data, the vehicle holographic monitoring unit acquires mine car profile, position, and appearance data, the multi-mode identity authentication unit acquires mine car and driver identity data, the environment state monitoring unit acquires weighing area environment parameters, and the empty load feature acquisition unit acquires mine car chassis image and infrared temperature distribution data.

[0007] The secure transmission layer adopts a dual-link redundancy architecture of "industrial Ethernet + 5G private network", and realizes end-to-end encrypted transmission of data through a national encryption algorithm.

[0008] The intelligent anti-cheating analysis layer includes a data fusion module, a multi-dimensional cheating recognition model, a blockchain storage module, and a cheating behavior knowledge base; the data fusion module is associated with each perception data, the multi-dimensional cheating recognition model includes an empty load weight increase recognition sub-module, the blockchain storage module stores vehicle empty load history data and feature archives, and the cheating behavior knowledge base stores empty load weight increase cheating cases.

[0009] The linkage treatment layer includes an automatic control unit, an audible and visual alarm unit, a remote monitoring center, and a cheating behavior tracing unit; the automatic control unit locks the cheating vehicle and retains the empty load weight increase cheating evidence, the audible and visual alarm unit sends early warning signals and special prompt words, the remote monitoring center pushes the empty load weight increase cheating detailed information, and the cheating behavior tracing unit generates a report containing an empty load weight increase special field.

[0010] Preferably, the encrypted weighing unit includes a digital signature intelligent weighing sensor and an interference signal detection module, the intelligent weighing sensor is built-in with a national encryption SM4 chip, and the interference signal detection module can identify 100-2400MHz frequency band cheating signals.

[0011] Preferably, the vehicle holographic monitoring unit includes a laser profile scanner, a high-definition industrial camera, and a millimeter wave radar, which are used to determine whether the mine car is completely weighed, whether the loading volume and weight are matched, and the like; the empty load feature acquisition unit includes a bottom high-definition camera and an infrared thermal imager, which are used to identify whether the mine car chassis is temporarily equipped with additional weights and whether the temperature difference of the car body is large.

[0012] Preferably, the multi-dimensional cheating recognition model adopts a deep learning and rule engine fusion algorithm, and includes a signal cheating recognition sub-model, a vehicle state cheating recognition sub-model, an identity cheating recognition sub-model, and an empty load weight increase recognition sub-module.

[0013] The signal cheating recognition sub-model is an improved LSTM network, an attention mechanism is introduced, and the input is a continuous signal sequence of a weighing sensor, and cheating recognition is realized by extracting signal time domain and frequency domain features;

[0014] The vehicle state cheating recognition sub-model is a "rule engine + CNN" hybrid architecture, the rule engine presets incomplete weighing, partial load, and volume-weight mismatch determination rules, and the CNN extracts vehicle loading form features;

[0015] The identity cheating recognition sub-model is a multi-modal feature fusion algorithm, RFID tag information, license plate recognition results, Beidou positioning trajectories and driver face feature vectors are input, and identity consistency determination is realized through a fully connected neural network;

[0016] The empty load weight increase recognition sub-module adopts a "historical data comparison + visual feature recognition" + multi-physical quantity detection" multi-dimensional determination mechanism, the historical data comparison calls the average value, trend and prepared mass range of the vehicle empty load weight in the past 30 days to determine weight abnormalities, the visual feature recognition identifies the attachments through the YOLOv8 model, the infrared temperature difference and the temporary laser profile comparison, and the multi-physical quantity detection determines the counterweight through vibration spectrum, electromagnetic signal and ultrasonic gap detection;

[0017] The four sub-models are fused to output the final determination result through a weighted voting method.

[0018] Preferably, the blockchain storage module adopts a consortium chain architecture, and the block contains weighing data frames, cheating recognition results, operation logs, vehicle empty load historical data and empty load feature archives.

[0019] Compared with the prior art, the intelligent mine unattended system has the following beneficial effects:

[0020] 1. The multi-source sensing unit is used to monitor multi-dimensional cheating behaviors such as signal interference, vehicle state and identity information, especially the empty load feature acquisition unit and the empty load weight increase recognition sub-module are added, the cheating recognition problem of the whole vehicle weight being too heavy when the vehicle is not loaded with ore is solved, and more than 95% of known cheating types can be recognized;

[0021] 2. The cheating recognition model is obtained by fusing deep learning and a rule engine, the encrypted weighing data and the vehicle holographic monitoring data are combined, the cheating recognition accuracy is greater than or equal to 99%, the empty load weight increase cheating recognition accuracy is greater than or equal to 99.5%, and the false positive rate is less than or equal to 0.5%;

[0022] 3. National secret encryption algorithms and blockchain storage technologies are used in all links from collection, transmission to storage, the vehicle empty load historical data and feature archives are specially stored, the weighing data is ensured to be tamper-proof and traceable, and the data tampering cheating risk is eliminated;

[0023] 4. The linkage treatment layer realizes automatic locking and remote early warning of cheating behavior, retains special evidence for overload cheating and pushes detailed information, response time is less than or equal to 3s, avoids cheating vehicles from driving away and is difficult to trace, and reduces mine economic losses.

[0024] The above description is only a summary of the technical solutions of the present application. In order to more clearly understand the technical means of the present application, and to implement the content of the description, the following will be described in detail with the preferred embodiments of the present application and the accompanying drawings. The specific embodiments of the present application are given in detail by the following examples and their accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0025] The accompanying drawings described herein are used to provide further understanding of the present application, and form a part of the present application. The schematic embodiments of the present application and their descriptions are used to explain the present application, and do not constitute an improper limitation on the present application. In the drawings:

[0026] Fig. 1 It is a structural schematic diagram of the intelligent mine unattended system;

[0027] Fig. 2 It is a device layout schematic diagram of the anti-cheating sensing layer in the present application;

[0028] Fig. 3 It is a workflow diagram of the intelligent anti-cheating analysis layer of the present application. DETAILED DESCRIPTION

[0029] Please combine Figs. 1 to 3 the drawings, the present application provides an intelligent mine unattended system, which comprises an anti-cheating sensing layer, a safe transmission layer, an intelligent anti-cheating analysis layer and a linkage treatment layer, each layer cooperates to realize the weighing anti-cheating function, especially a special anti-cheating mechanism is designed for the scene that the whole vehicle is overweight when the vehicle is not loaded with mine materials:

[0030] The anti-cheating sensing layer is used for collecting multi-dimensional data related to cheating behavior in the weighing process, including weighing original data, vehicle state data, environmental interference data and identity authentication data, and an empty load feature acquisition unit is additionally arranged for overload cheating; the anti-cheating sensing layer comprises:

[0031] The encrypted weighing unit adopts an intelligent weighing sensor with digital signature (the precision level is not less than 0.05%), the sensor is internally provided with a unique hardware ID and a national secret SM4 encryption chip, each frame of weighing data is attached with a chip signature and a time stamp, so as to prevent the signal from being tampered; the matched weighing instrument is provided with an interference signal detection module, which can identify the cheating signal in the frequency band of 100-2400MHz, and record the interference occurrence time and intensity;

[0032] Vehicle holographic monitoring unit: composed of 3 sets of laser profile scanners (scanning frequency ≥ 50 Hz), 4 high-definition industrial cameras (resolution ≥ 4K) and millimeter wave radars; the laser scanner collects the three-dimensional profile and load height of the mine car to determine whether there is abnormal loading such as overloading; the high-definition camera takes pictures of the front, rear, left and right sides of the mine car to identify the license plate, logo and body features, and realize the uniqueness verification of the vehicle appearance; the millimeter wave radar monitors the parking position of the mine car and outputs the distance data between the wheels and the edge of the scale platform to determine whether it is completely on the scale;

[0033] Multi-mode identity authentication unit: integrated with ultra-high frequency RFID card reader (read-write distance 0.5-10m, anti-metal interference), Beidou positioning module and driver face recognition human-computer interaction terminal; the RFID tag adopts a passive anti-tear design and stores unique information such as mine car number, rated load, belonging unit, vehicle kerb mass (empty standard weight ± 5%) etc.; the Beidou positioning module collects real-time mine car position coordinates, compares with the preset weighing area electronic fence, records the vehicle track to and from the mine and unloading point, and determines whether it is in the empty return state; the driver needs to verify his identity through face recognition to ensure the legality of the operation permission;

[0034] Environmental state monitoring unit: deploy vibration sensors, temperature and humidity sensors and dust sensors to collect scale platform vibration frequency, environmental temperature and humidity and dust concentration to distinguish normal weighing vibration from cheating knocking vibration and monitor the influence of environmental factors on equipment accuracy;

[0035] Empty feature acquisition unit: composed of 2 bottom high-definition cameras (installed below the scale platform on both sides, resolution ≥ 2K), infrared thermal imager, electromagnetic induction sensor and ultrasonic sensor; the bottom camera takes pictures of the mine car chassis and carriage bottom to identify whether the weight increasing devices such as detachable counterweight and temporary water tank are installed; the infrared thermal imager collects the temperature distribution of the mine car body to distinguish between the vehicle inherent components and the temporary installed abnormal temperature counterweight (such as water tank filled with cold water with a temperature difference ≥ 5℃ from the vehicle body); the electromagnetic induction sensor (detection range 0.1-1m, sensitivity ≥ 10μT) is deployed at the edge of the scale platform to detect the chassis metal counterweight; the ultrasonic sensor (measurement range 0.2-5m, accuracy ± 0.5%) is installed on the overhead bracket of the scale platform to detect the clearance between the carriage bottom and the chassis; record the profile features, weight data and multi-physical quantity parameters of the vehicle when empty, and establish the vehicle empty feature file.

[0036] The secure transmission layer is used to achieve encrypted transmission of data and command interaction of the anti-cheating perception layer. It adopts a dual-link redundant architecture of "industrial Ethernet + 5G private network". The industrial Ethernet transmits encrypted weighing data and control commands, while the 5G private network transmits high-definition images and laser scanning data. Data transmission uses the national cryptographic SM2 / SM3 algorithm for end-to-end encryption. The transmission protocol supports message integrity verification and replay attack protection. A network isolation firewall is set up to prohibit unauthorized external access to the internal network of the system.

[0037] The intelligent anti-cheating analysis layer, as the core of the system's anti-cheating mechanism, is used to integrate and analyze the collected data and judge cheating behavior. It adds an empty-load weight gain identification sub-module for scenarios where the vehicle weight is too heavy when it is not loaded with ore. It includes a data fusion module, a multi-dimensional cheating identification model, a blockchain evidence storage module, and a cheating behavior knowledge base.

[0038] Data fusion module: Connects to data from various sensing units, performs time synchronization (synchronization accuracy ≤1ms), format conversion and correlation matching, and generates a unified data frame containing "weighing data - vehicle characteristics - identity information - environmental parameters - empty load characteristics";

[0039] Multi-dimensional cheating detection model: This model employs a fusion algorithm combining deep learning and a rule engine to achieve accurate identification of cheating behavior through multi-model collaboration. The specific design of each sub-model is as follows:

[0040] Signal cheating detection sub-model: An improved LSTM (Long Short-Term Memory) algorithm is used. The input is a continuous 500ms signal sequence (sampling frequency 1kHz) collected by a weighing sensor. An attention mechanism is introduced to enhance the weight allocation of abnormal signal segments. The model has three hidden layers with 256, 128, and 64 neurons per layer, respectively, and ReLU activation function. The output layer is a binary classification (normal / cheating) layer using the Sigmoid activation function. During training, cross-entropy loss function and Adam optimizer are used, with dynamic adjustment of the learning rate (initially 0.001, decreasing by 10% every 10 rounds). Dropout (dropout rate=0.3) is used to prevent overfitting. The model extracts frequency domain features (such as harmonic components and spectral distortion) and time domain features (such as peak fluctuations and mean abrupt changes) to identify cheating behaviors such as electronic interference and signal tampering, achieving an accuracy of ≥99%.

[0041] Vehicle Status Cheating Detection Sub-model: Adopts a hybrid architecture of "rule engine + CNN (convolutional neural network)"; the rule engine has three preset judgment rules: ① Millimeter-wave radar data judgment: when the distance between any wheel and the edge of the weighing platform is ≤10cm, it is judged as incomplete weighing; ② Weight distribution rule: calculates the off-center load coefficient (weight on one side / total weight) based on the weight of each support point of the weighing sensor; when the off-center load coefficient is >0.6 or <0.4, it is judged as off-center load cheating; ③ Volume-weight matching rule: calculates the cargo volume of the mining truck based on laser contour scanning data (using a 3D point cloud volume calculation algorithm), combined with the density of the ore (… The theoretical weight is calculated based on a preset value of 2.8t / m³ ± 0.2t / m³. An anomaly is triggered when the absolute value of the deviation between the actual weighed weight and the theoretical weight exceeds 8%. The input to the CNN part is a top view of the mine car generated by laser contour scanning (resolution 512×512). The loading morphology features are extracted through 5 convolutional layers (3×3 kernel size, stride 1) and 2 pooling layers (2×2 pooling kernel size) to determine whether there is any hidden physical cheating such as adding load weights or unloading midway. The output of the CNN model and the result of the rule engine are output through a logical "OR" operation to output the final judgment result. The overall recognition accuracy is ≥98.5%.

[0042] The identity fraud detection sub-model employs a multimodal feature fusion algorithm. It takes RFID tag information (mine truck number, rated load, curb weight), license plate recognition results, BeiDou positioning trajectory (whether the trajectory in the past 10 minutes is within the preset transportation route), and the driver's facial recognition feature vector (128 dimensions) as input. These features are then concatenated and input into a fully connected neural network. The fully connected layer consists of two layers (64 and 32 neurons respectively), with LeakyReLU activation. The output layer is a three-class classification (identity matching / partial inconsistency / complete inconsistency), using the Softmax activation function. Model training uses the Focal Loss loss function to address class imbalance. Contrastive learning enhances the discriminative power of different identity features. When any two modalities do not match (e.g., the RFID mine truck number and license plate recognition result are inconsistent), the model determines it as identity fraud, with an accuracy rate ≥99.2%.

[0043] The empty-load weight gain identification submodule employs a multi-dimensional judgment mechanism combining historical data comparison, visual feature recognition, and multi-physical quantity detection to achieve comprehensive identification of excessive vehicle weight when the vehicle is not loaded with ore: ① Historical data comparison mechanism: Retrieves empty-load weighing data (at least 5 valid data points) of the vehicle from the blockchain storage module within the past 30 days, calculates the mean and standard deviation of the empty-load weight, and triggers a weight anomaly warning when the current empty-load weight exceeds "mean ± 2 times standard deviation" or exceeds the curb weight stored in the RFID tag within ± 5% of the range; simultaneously, it introduces vehicle empty-load weight trend analysis, and if the empty-load weight shows an increasing trend for 3 consecutive times with a single increase ≥ 0.5t, a warning is triggered even if it does not exceed the threshold; ② Visual feature recognition mechanism: Inputs the mining truck chassis image (2K resolution) captured by the bottom camera into the YOLOv8 object detection model, and incorporates a dataset of common mining counterweight devices (including 12 types of counterweight samples such as water tanks, sandbags, metal blocks, and concrete blocks) during training. The model achieves an accuracy rate of ≥98% in identifying foreign objects on the chassis. Simultaneously, a top laser scanner collects the 3D contour of the vehicle in its unloaded state, comparing it with the standard unloaded contour stored in the blockchain. If the contour volume difference is ≥0.3m³, it is determined that new attachments exist. ③ Multi-physical quantity detection mechanism: Vibration sensors collect the vibration spectrum of the vehicle when it drives onto the weighing platform unloaded, comparing it with a standard unloaded vibration spectrum library. If the characteristic frequency difference is ≥15%, it is determined that there may be metal counterweights. Electromagnetic induction sensors detect whether there are high-intensity magnetic field signals (corresponding to metal counterweights) in the chassis area; an anomaly is triggered when the signal strength is ≥50μT. Ultrasonic sensors detect the gap between the bottom of the cargo box and the chassis; if the difference compared with the standard gap value is ≥10cm, it is determined that there may be interlayer counterweights. The results of multiple mechanisms are output through a logical "OR" operation. When any mechanism determines an anomaly, it is identified as cheating behavior where the vehicle's overall weight is too high when it is not loaded with ore. The overall identification accuracy is ≥99.5%.

[0044] Model fusion strategy: The weighted voting method is used to fuse the output results of the four sub-models. The weights of the signal cheating identification sub-model, vehicle status cheating identification sub-model, identity cheating identification sub-model, and empty load weight increase identification sub-module are set to 0.3, 0.25, 0.2, and 0.25, respectively. When the fusion result is ≥0.5, it is determined that cheating behavior exists, ensuring full coverage identification in multiple cheating scenarios, especially strengthening the identification weight of empty load weight increase cheating.

[0045] Blockchain Evidence Storage Module: Adopting a consortium blockchain architecture, it generates blocks from the encrypted data frames, cheating detection results, and operation logs of each weighing transaction. The blocks contain the hash value and timestamp of the previous block. At the same time, it also stores the historical data of the empty weight and empty characteristic files of each vehicle, so as to realize the immutability and full traceability of weighing data.

[0046] Cheating behavior knowledge base: Stores historical cheating cases, cheating feature parameters and handling strategies. It focuses on cheating cases where the vehicle weight is too heavy when the vehicle is not loaded with ore (including counterweight type, abnormal weight range, visual features, etc.). It continuously optimizes the accuracy of the cheating identification model through incremental learning.

[0047] The linkage and processing layer is used to respond to and handle identified cheating behaviors in real time; it includes an automatic control unit, an audible and visual alarm unit, a remote monitoring center, and a cheating behavior tracing unit.

[0048] Automatic control unit: When cheating is detected, the gate is immediately locked (to prevent the mine car from leaving), the weighing instrument stops uploading data, and the subsequent transportation scheduling authority of the cheating vehicle is cut off; for empty load weight increase cheating, the recording and saving functions of the bottom camera and infrared thermal imager are additionally triggered to preserve the cheating evidence.

[0049] Audible and visual alarm unit: emits an audible and visual alarm signal of 110dB or higher, and simultaneously displays the type of cheating and handling prompts on the on-site display screen; for cheating by adding weight without load, it specifically displays the prompt "Suspected weight addition without load, please check vehicle counterweight";

[0050] Remote monitoring center: pushes cheating warning information to the terminal of the on-duty personnel in real time (including images of cheating vehicles and screenshots of abnormal data). For cheating of adding weight without load, it simultaneously pushes images of foreign objects on the chassis taken by the bottom camera and infrared temperature distribution map, allowing the on-duty personnel to confirm the on-site situation through remote video and issue manual handling instructions.

[0051] Cheating behavior tracing unit: Automatically generates cheating behavior reports, including cheating time, vehicle information, cheating type, data evidence chain and handling results. For empty load weight increase cheating, it records in detail information such as abnormal weight values, historical average comparison, and visually identified counterweight types. It supports querying historical cheating records by time period, vehicle number and other dimensions.

[0052] in:

[0053] Anti-cheating perception layer configuration: The encrypted weighing unit uses a Zemic H8C-C3-150t intelligent weighing sensor (accuracy 0.05%), paired with an XK3190-DS3 anti-cheating instrument (interference detection frequency band 100-2400MHz); the vehicle holographic monitoring unit uses a SICK LMS511 laser scanner (scanning angle 190°, distance resolution 1mm), a Hikvision DS-2CD7A87F / E-LZ4K camera, and a Huawei MDR-80 millimeter-wave radar (detection distance 0.5-80m); the multi-mode identity authentication unit uses a Yuanwanggu UHF RFID reader (model YXU2861), a Beidou BD-2 positioning module (positioning accuracy ≤10m), and a Hikvision DS-K5671 face terminal (recognition accuracy ≥99.7%), with RFID tags storing the curb weight of the mining truck (e.g., 25t±1.25t); the environmental status monitoring unit uses a PCB... The system includes a 356A15 vibration sensor (range ±50g), an SHT30 temperature and humidity sensor, and a GCG1000 dust sensor. The no-load feature acquisition unit uses a Hikvision DS-2CD3T26WD-I5 bottom camera (2K resolution, 25fps), an FLIRA655sc infrared thermal imager (temperature range -20℃-150℃, resolution 640×512), a Honeywell CSN-A1 electromagnetic induction sensor (detection range 0.1-1m), and a SICK UM30-213110 ultrasonic sensor (measurement range 0.2-5m).

[0054] Security transmission layer configuration: The industrial Ethernet uses Huawei S5735-S series gigabit switches (supporting VLAN isolation), and the 5G private network uses Huawei 5G CPE Pro 3 (supporting SA mode); data encryption uses the national cryptographic SM2 / SM3 algorithm, and the transmission protocol uses MQTT-SN (supporting message encryption and verification); the firewall is the Huawei USG6000 series next-generation firewall.

[0055] The intelligent anti-cheating analysis layer is configured as follows: The data fusion module is developed using Python and employs the PySpark framework for distributed data processing; the multi-dimensional cheating identification model is built on the TensorFlow 2.0 framework, with a training dataset containing 100,000 normal weighing data points and 50,000 cheating simulation data points (covering 20 common cheating types, including 3,000 empty-load weight gain cheating data points), and the model inference speed is ≤200ms / data point; the signal cheating identification sub-model has undergone 50 rounds of training iterations, with a validation set accuracy consistently above 99.2%; the vehicle status cheating identification sub-model's rule engine has been debugged and optimized using 1,000 sets of actual test data, and the CNN part has undergone 30 rounds of training iterations, achieving a comprehensive accuracy of 98.8%; the identity cheating identification sub-model is trained using 50,000 sets of multimodal identity data, and the validation set accuracy is consistently above 99.2%. The evidence set accuracy reached 99.5%; the YOLOv8 model training of the empty-load weight gain identification submodule used a dataset containing 5,000 images of mine counterweight devices, iterated for 40 rounds, and achieved a foreign object identification accuracy of 98.5%; the historical data comparison mechanism was verified through 100,000 historical empty vehicle data, and the weight anomaly judgment accuracy reached 99.8%; the multi-physical quantity detection mechanism was calibrated through 20,000 sets of measured data, and the vibration spectrum and electromagnetic signal identification accuracy reached 98.2% and 99.1% respectively, with a comprehensive identification accuracy of 99.5%; the blockchain evidence storage module adopted the Hyperledger Fabric consortium chain, with nodes including the mine management center, weighing stations and third-party audit nodes, and a block generation interval of ≤10s, and a dedicated on-chain channel for "vehicle empty-load files" was opened; the cheating behavior knowledge base was stored in a MySQL database and supported daily automatic incremental updates.

[0056] Linkage and handling layer configuration: The automatic control unit uses a Siemens S7-1200 PLC to control the barrier gate and weighing instrument, triggering a video recording and saving command for empty-load weight increase cheating; the audible and visual alarm unit uses an industrial alarm with a decibel level of ≥120dB and a 55-inch outdoor LED display screen, supporting customized cheating prompts; the remote monitoring center deploys a Huawei IVS3800 video cloud platform, supporting 4K video real-time preview and alarm information push, and simultaneously displaying chassis images and infrared temperature maps; the cheating behavior tracing unit develops a web query system, supporting Excel format report export, including a special field for empty-load weight increase cheating.

[0057] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Those skilled in the art can readily implement the present invention based on the accompanying drawings and the above description. However, any modifications, alterations, or variations made by those skilled in the art without departing from the scope of the present invention, utilizing the disclosed technical content, are equivalent embodiments of the present invention. Furthermore, any modifications, alterations, or variations made to the above embodiments based on the essential technology of the present invention are still within the protection scope of the present invention.

Claims

1. An intelligent unmanned mine system, characterized in that, It includes an anti-cheating perception layer, a secure transmission layer, an intelligent anti-cheating analysis layer, and a coordinated response layer; The anti-cheating perception layer includes an encrypted weighing unit, a vehicle holographic monitoring unit, a multi-mode identity authentication unit, an environmental status monitoring unit, and an empty-load feature acquisition unit. The encrypted weighing unit collects encrypted weighing data and interference signal data; the vehicle holographic monitoring unit collects the outline, position and appearance data of the mining truck; the multi-mode identity authentication unit collects the identity data of the mining truck and the driver; the environmental status monitoring unit collects the environmental parameters of the weighing area; and the empty-load feature acquisition unit collects the image of the mining truck chassis and infrared temperature distribution data. The secure transmission layer adopts a dual-link redundancy architecture of "industrial Ethernet + 5G private network" and uses national cryptographic encryption algorithms to achieve end-to-end encrypted data transmission. The intelligent anti-cheating analysis layer includes a data fusion module, a multi-dimensional cheating identification model, a blockchain evidence storage module, and a cheating behavior knowledge base. The data fusion module associates and matches various sensing data, the multi-dimensional cheating identification model includes an empty load weight increase identification sub-module, the blockchain evidence storage module stores vehicle empty load historical data and feature files, and the cheating behavior knowledge base stores empty load weight increase cheating cases. The linkage processing layer includes an automatic control unit, an audible and visual alarm unit, a remote monitoring center, and a cheating behavior tracing unit. The automatic control unit locks the cheating vehicle and retains evidence of empty-load weight gain cheating. The audible and visual alarm unit issues a warning signal and a special prompt. The remote monitoring center pushes detailed information on empty-load weight gain cheating. The cheating behavior tracing unit generates a report containing a special field for empty-load weight gain.

2. The intelligent unmanned mine system according to claim 1, characterized in that: The encrypted weighing unit includes an intelligent weighing sensor with digital signature and an interference signal detection module. The intelligent weighing sensor has a built-in national standard SM4 encryption chip, and the interference signal detection module can identify cheating signals in the 100-2400MHz frequency band.

3. The intelligent unmanned mine system according to claim 1, characterized in that: The vehicle holographic monitoring unit includes a laser contour scanner, a high-definition industrial camera, and a millimeter-wave radar, used to determine whether the mining truck is fully weighed and whether the loading volume and weight match; the empty feature acquisition unit includes a bottom high-definition camera and an infrared thermal imager, used to identify whether temporary counterweights have been added to the mining truck chassis and the temperature difference of the vehicle body.

4. The intelligent unmanned mine system according to claim 1, characterized in that: The multi-dimensional cheating identification model adopts a deep learning and rule engine fusion algorithm, including a signal cheating identification sub-model, a vehicle status cheating identification sub-model, an identity cheating identification sub-model, and an empty load weight increase identification sub-module. The signal cheating identification sub-model is an improved LSTM network that introduces an attention mechanism. The input is a continuous signal sequence from a weighing sensor, and cheating identification is achieved by extracting the time-domain and frequency-domain features of the signal. The vehicle status cheating identification sub-model is a hybrid architecture of "rule engine + CNN". The rule engine presets judgment rules for incomplete weighing, uneven loading, and volume-weight mismatch, while the CNN extracts the loading morphology features of the mining truck. The identity fraud identification sub-model is a multimodal feature fusion algorithm. It takes RFID tag information, license plate recognition results, Beidou positioning trajectory and driver facial feature vector as input, and realizes identity consistency determination through a fully connected neural network. The empty weight gain identification submodule adopts a multi-dimensional judgment mechanism of "historical data comparison + visual feature recognition + multi-physical quantity detection". Historical data comparison retrieves the average empty weight, trend and curb weight range of the vehicle in the past 30 days to determine weight abnormalities. Visual feature recognition identifies attachments by comparing the YOLOv8 model with the laser profile. Multi-physical quantity detection determines the counterweight by vibration spectrum, electromagnetic signal and ultrasonic gap detection. The four sub-models are merged using a weighted voting method to output the final judgment result.

5. The intelligent unmanned mine system according to claim 1, characterized in that: The blockchain evidence storage module adopts a consortium blockchain architecture, and the blocks contain weighing data frames, cheating identification results, operation logs, vehicle empty-load history data, and empty-load characteristic files.

Images