Mine intelligent transportation method and system

Abstract

This invention relates to the field of intelligent transportation technology and discloses an intelligent transportation method and system for mines. The method includes: performing time-series filtering and gradient denoising on temperature data of mine transportation nodes collected by a thermal imaging camera to obtain a temperature feature matrix; identifying transportation anomaly information based on the temperature feature matrix; performing temperature pattern matching based on the transportation anomaly information to generate a transportation operation instruction set; using the transportation operation instruction set as constraints to plan temperature-safe paths in mine tunnels to obtain a transportation execution plan; after real-time location of mine workers using a thermal imaging camera, converting the transportation execution plan into standardized voice commands and pushing them to corresponding workers via a walkie-talkie network, and acquiring real-time scheduling feedback data including command reception confirmation and execution progress. This invention improves the intelligence level and safety of mine transportation by accurately locating workers through thermal imaging and combining it with a walkie-talkie network for zoned push notifications.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation technology, and in particular to an intelligent transportation method and system for mines. Background Technology

[0002] Mine transportation, a crucial link in coal mine production, faces challenges such as a complex underground environment, variable temperature conditions, and frequent personnel and equipment movement. During mine transportation, abnormal situations such as equipment overheating, personnel overcrowding, and cargo accumulation frequently occur. These abnormalities are often accompanied by significant temperature variations, such as abnormal temperature rise due to prolonged operation of transportation equipment, temperature concentration in densely populated work areas, and localized temperature distribution anomalies caused by improper cargo stacking. Traditional mine transportation management relies primarily on manual inspections and experience-based judgment, making it difficult to achieve real-time monitoring and accurate identification of temperature anomalies, and thus hindering the timely detection of potential safety hazards.

[0003] Existing mine transportation dispatching systems generally suffer from low levels of intelligence, slow response speed, and poor dispatching accuracy. On the one hand, traditional monitoring methods mainly rely on video surveillance or sensor point monitoring, lacking comprehensive perception of temperature distribution and failing to accurately identify abnormal states in complex environments. On the other hand, existing dispatching instruction generation often relies on manual decision-making or simple rule engines, lacking intelligent abnormal pattern recognition and instruction matching capabilities, making it difficult to automatically generate accurate transportation operation instructions based on real-time temperature changes. Summary of the Invention

[0004] This invention provides an intelligent transportation method and system for mines. The invention uses thermal imaging to accurately locate workers and combines it with a walkie-talkie network to achieve zoned push notifications, thereby improving the intelligence level and safety of mine transportation.

[0005] In a first aspect, the present invention provides an intelligent transportation method for mines, the intelligent transportation method for mines comprising:

[0006] Temporal filtering and gradient denoising were performed on the temperature data of mine transportation nodes acquired by thermal imaging cameras to obtain the temperature feature matrix;

[0007] Based on the temperature feature matrix, transportation anomaly information is identified;

[0008] Based on the transportation anomaly information, temperature pattern matching is performed to generate a transportation operation instruction set;

[0009] The transportation operation instruction set is used as a constraint to plan the temperature safety path of the mine passage, resulting in a transportation execution plan.

[0010] After locating the specific positions of mine workers in real time using thermal imaging cameras, the transportation execution plan is converted into standardized voice commands and pushed to the corresponding workers via walkie-talkie network partitions, and scheduling feedback data including command reception confirmation and execution progress is obtained in real time.

[0011] Optionally, in a first implementation of the first aspect of the present invention, the step of performing time-series filtering and gradient denoising on the temperature data of mine transportation nodes acquired by the thermal imaging camera to obtain a temperature feature matrix includes:

[0012] Temperature data of mine transportation nodes are collected by thermal imaging cameras, and the temperature data of the mine transportation nodes are normalized according to the mine environmental reference temperature to obtain standardized temperature data.

[0013] A sliding time window algorithm is used to calculate the temperature weighted average value between consecutive frames in the standardized temperature data, and the standardized temperature data is then subjected to time-series smoothing based on the temperature weighted average value to obtain time-series smoothed temperature data.

[0014] Calculate the temperature difference between each pixel and its neighboring pixels in the time-smoothed temperature data, and perform spatial denoising on the time-smoothed temperature data based on the temperature difference to obtain spatially denoised temperature data;

[0015] Based on the spatially denoised temperature data, the temperature distribution value of the current frame, the temperature change rate between previous and next frames, and the temperature gradient value between pixels are calculated respectively, and a temperature feature matrix including the temperature distribution matrix, the temperature change rate matrix, and the temperature gradient matrix is ​​constructed.

[0016] Optionally, in a second implementation of the first aspect of the present invention, the step of identifying transportation anomaly information based on the temperature feature matrix includes:

[0017] A time-series analysis is performed on the temperature change rate matrix in the temperature feature matrix to determine the location of abnormal heating in the transportation equipment, and equipment abnormality mask data is generated based on the location of abnormal heating in the transportation equipment.

[0018] Based on the temperature distribution matrix in the temperature feature matrix, temperature pixels within the target temperature range are selected, and pixel distribution analysis of densely populated areas is performed based on the temperature pixels to obtain personnel anomaly mask data.

[0019] The temperature distribution uniformity within the cargo storage area is calculated based on the temperature gradient matrix in the temperature feature matrix, and cargo anomaly mask data is determined based on the temperature distribution uniformity.

[0020] The equipment anomaly mask data, personnel anomaly mask data, and cargo anomaly mask data are spatially mapped and anomaly severity is calculated to generate transportation anomaly information containing anomaly type, location coordinates, and severity level.

[0021] Optionally, in a third implementation of the first aspect of the present invention, the step of calculating the temperature distribution uniformity within the cargo storage area based on the temperature gradient matrix in the temperature feature matrix, and determining cargo anomaly mask data based on the temperature distribution uniformity, includes:

[0022] The regions with temperature gradients less than a preset gradient range are extracted from the temperature gradient matrix of the temperature feature matrix and designated as cargo storage areas.

[0023] The temperature distribution uniformity is calculated based on the cargo storage area, and candidate areas of abnormal temperature distribution are identified based on the temperature distribution uniformity.

[0024] By using the candidate regions of abnormal temperature distribution and adopting a temperature change tracking mode, abnormal areas of cargo accumulation are identified, and the abnormal areas of cargo accumulation are converted into binary code form to construct cargo abnormality mask data corresponding to the temperature gradient matrix.

[0025] Optionally, in a fourth implementation of the first aspect of the present invention, the step of performing temperature pattern matching based on the transportation anomaly information to generate a transportation operation instruction set includes:

[0026] Extract the anomaly type, location coordinates, and severity level from the transportation anomaly information, and encode the anomaly type as equipment overheating identifier, personnel gathering identifier, and cargo stacking identifier to obtain an anomaly feature encoding vector;

[0027] Based on the abnormal feature encoding vector, a preset transportation instruction matching dictionary is queried to obtain a set of candidate instruction templates;

[0028] The cosine similarity between the abnormal feature encoding vector and the feature vector of each instruction template in the candidate instruction template set is calculated. When the similarity value exceeds the similarity threshold, the corresponding instruction template is activated and a transportation operation instruction set is generated parameterized according to the location coordinates and the severity level.

[0029] Optionally, in a fifth implementation of the first aspect of the present invention, the step of querying a preset transportation instruction matching dictionary based on the abnormal feature encoding vector to obtain a candidate instruction template set includes:

[0030] Parse the anomaly type identifier in the anomaly feature encoding vector. When the anomaly type identifier is an equipment overheat identifier, query the equipment class instruction branch. When the anomaly type identifier is a personnel gathering identifier, query the personnel class instruction branch. When the anomaly type identifier is a cargo stacking identifier, query the cargo class instruction branch to obtain the transportation instruction matching dictionary.

[0031] The location coordinates are read from the transportation instruction matching dictionary, matched with the mine tunnel topology database to determine the connecting channels and adjacent facilities in the abnormal area, and the location-matching instruction templates are filtered out.

[0032] Based on the abnormal feature encoding vector, the location adaptation instruction template is filtered according to the emergency response intensity to obtain the level matching instruction template;

[0033] The level matching instruction template is cross-validated with the timeliness requirements for exception handling to form a set of candidate instruction templates.

[0034] Optionally, in a sixth implementation of the first aspect of the present invention, the step of using the transportation operation instruction set as a constraint condition for mine tunnel temperature safety path planning to obtain a transportation execution scheme includes:

[0035] The execution area and time requirements of the transportation operation instruction set are analyzed, the execution area is marked as a prohibited area in the route planning, and the time requirements are converted into time window parameters for route planning to obtain the constraints.

[0036] Based on the real-time monitoring of mine tunnel temperature distribution data by thermal imaging cameras, and based on the mine tunnel temperature distribution data, feature extraction is performed to construct a temperature weighted path network.

[0037] The constraints and the temperature-weighted path network are input into a mixed-integer programming solver for multi-objective optimization to obtain the multi-objective optimized path solution.

[0038] Based on the multi-objective optimized path solution, specific vehicle scheduling times, personnel assignments, and cargo transfer routes are allocated. Combined with the mine work shift schedule and equipment operating status, a detailed time plan and resource allocation table are generated to form a transportation execution plan.

[0039] Optionally, in a seventh implementation of the first aspect of the present invention, the step of inputting the constraint conditions and the temperature-weighted path network into a mixed-integer programming solver for multi-objective optimization to obtain a multi-objective optimized path solution includes:

[0040] Based on the constraints, the restricted area is represented as a binary variable constraint, the time window parameter is represented as an inequality constraint, and the temperature weighted path network is converted into a 0-1 integer variable matrix for path selection, thus obtaining a mixed integer programming solver.

[0041] The objective functions are: transportation time objective function to calculate the ratio of total route length to speed; temperature safety objective function to calculate the cumulative value of route temperature risk; route congestion objective function to count the frequency of route use during the same period; and energy cost objective function to accumulate the power consumption of vehicles and equipment, thus obtaining a comprehensive objective function.

[0042] The root node of the branch and bound algorithm is initialized based on the mixed integer programming solver and the upper bound of the search is set to positive infinity and the lower bound is set to the current optimal solution. Branch operations are performed on integer variables one by one to generate child nodes. The linear relaxation solution of each child node is calculated as the lower bound estimate of the corresponding child node to obtain the branch search tree.

[0043] Traverse the active nodes that have not been pruned in the branch search tree. When the lower bound of an active node exceeds the current optimal upper bound, prune the corresponding branch. When an active node obtains an integer feasible solution and the objective function value of the integrated objective function is greater than the current optimal solution, update the optimal solution. Repeat the branch search until all nodes have been processed to obtain the multi-objective optimized path solution.

[0044] Optionally, in the eighth implementation of the first aspect of the present invention, after locating the specific position of the mine workers in real time using a thermal imaging camera, the transportation execution plan is converted into standardized voice commands and pushed to the corresponding workers via walkie-talkie network partitions, and real-time scheduling feedback data including command reception confirmation and execution progress is obtained, including:

[0045] Based on thermal imaging cameras, the specific location of miners can be determined by detecting human body temperature characteristics and recognizing human body contours.

[0046] The vehicle scheduling time, personnel division of labor and cargo transfer route in the transportation execution plan are converted into voice command text according to the standard voice specification for mine operations, and standardized voice commands are generated based on the voice command text.

[0047] Based on the specific location of the mine workers, the walkie-talkie network is queried, the walkie-talkie channel and device address corresponding to the area where each worker is located are matched, and the standardized voice commands are pushed to the target workers in the partition.

[0048] Monitor the voice confirmation signals and execution status reports returned by each walkie-talkie terminal, and obtain scheduling feedback data including instruction reception confirmation and execution progress in real time.

[0049] Secondly, the present invention provides an intelligent mine transportation system, the intelligent mine transportation system comprising:

[0050] The denoising module is used to perform time-series filtering and gradient denoising on the temperature data of mine transportation nodes acquired by the thermal imaging camera to obtain the temperature feature matrix.

[0051] The identification module is used to identify transportation anomaly information based on the temperature feature matrix;

[0052] The matching module is used to perform temperature pattern matching based on the transportation anomaly information and generate a transportation operation instruction set;

[0053] The planning module is used to plan the temperature safety path of the mine passage as a constraint condition using the transportation operation instruction set, so as to obtain the transportation execution plan;

[0054] The push module is used to locate the specific position of the mine workers in real time using a thermal imaging camera, convert the transportation execution plan into standardized voice commands, and push them to the corresponding workers through the walkie-talkie network. It also obtains scheduling feedback data in real time, including command reception confirmation and execution progress.

[0055] The technical solution provided by this invention achieves comprehensive temperature monitoring of the mine transportation area by collecting temperature data and constructing a three-dimensional temperature feature matrix using a thermal imaging camera, overcoming the limitations of traditional point-based monitoring. It employs temperature time-series masking technology and specialized algorithms to accurately identify three types of anomalies: equipment overheating, personnel gathering, and cargo accumulation, establishing an intelligent temperature anomaly identification mechanism. Based on anomaly feature encoding and instruction matching dictionaries, cosine similarity calculation is used to achieve intelligent mapping from temperature patterns to transportation instructions, significantly improving the automation level of instruction generation. Temperature constraints are integrated into path planning, and a mixed-integer programming algorithm is used to simultaneously optimize multiple objective functions, achieving dynamic path planning that considers temperature safety. By accurately locating workers through thermal imaging and combining it with a walkie-talkie network for zoned push notifications, a precise scheduling mechanism based on real-time location is established. The entire system forms a complete technical closed loop from temperature monitoring, anomaly identification, instruction generation, path planning to scheduling execution, significantly improving the intelligence level and safety of mine transportation compared to traditional manual scheduling methods. Attached Figure Description

[0056] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1This is a schematic diagram of one embodiment of the intelligent mine transportation method according to the present invention;

[0058] Figure 2 This is a schematic diagram of one embodiment of the intelligent transportation system in mines according to the present invention. Detailed Implementation

[0059] This invention provides an intelligent transportation method and system for mines. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0060] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 One embodiment of the intelligent mine transportation method of the present invention includes:

[0061] Step S101: Perform time-series filtering and gradient denoising on the temperature data of the mine transportation nodes collected by the thermal imaging camera to obtain the temperature feature matrix;

[0062] It is understood that the executing entity of this invention can be an intelligent mine transportation system, a terminal, or a server; no specific limitation is made here. This embodiment of the invention will be described using a server as an example.

[0063] Specifically, thermal imaging cameras deployed at key transportation nodes in the mine acquire temperature images of transportation equipment areas, personnel passage areas, and cargo stacking areas, obtaining raw thermal imaging data in continuous time frames. Due to various interference sources within the mine, such as equipment heat radiation, personnel mobility, and external air convection, the acquired temperature data is normalized based on the mine's baseline temperature. This transforms all pixel temperature values ​​into a standardized temperature range related to the baseline temperature, resulting in standardized temperature data. A sliding time window algorithm is used to calculate the weighted average temperature across consecutive frames in the standardized temperature data. Within each sliding window, the weighted average temperature of corresponding pixels across consecutive frames is calculated, where the weight parameter is set with a decreasing function based on the frame's temporal position within the window to enhance the temperature contribution of recent frames. The temperature value of the current time frame is smoothed based on the weighted average, completing the temporal smoothing of the temperature data and obtaining temporally smoothed temperature data. Building upon the temporal processing, spatial denoising is performed to identify and eliminate spatial noise caused by single-frame image acquisition errors or local extreme values. This operation is achieved by calculating the temperature difference between each pixel and its eight neighboring pixels. If the difference exceeds a preset abnormal fluctuation threshold (e.g., 3℃), the pixel is marked as an anomaly and repaired using bilinear interpolation or local median fill, resulting in a spatially denoised temperature data image. Based on the denoised temperature data, the temperature distribution value of each pixel in the current image frame is calculated sequentially to describe the thermal field state of the mine at a certain time point. Simultaneously, the temperature change rate between this frame and its preceding and following frames is extracted to capture the dynamic characteristics of temperature changes over time. Furthermore, the temperature gradient value between each pixel and its neighboring pixels is calculated to express the local distribution characteristics of the spatial temperature difference. These three types of information constitute the temperature distribution matrix, the temperature change rate matrix, and the temperature gradient matrix, respectively, and are combined into a three-dimensional temperature feature matrix.

[0064] Step S102: Identify transportation anomaly information based on temperature feature matrix;

[0065] Specifically, a time-series analysis is performed on the temperature change rate matrix within the temperature feature matrix. A time-series analysis algorithm is used to extract the temperature change rate of each pixel within a continuous time window, and a threshold condition is set to determine whether abnormal heating occurs. When the temperature in a certain area rises by more than 15°C within 5 minutes and the rate of change significantly deviates from the surrounding background level, it is marked as an abnormal equipment heating area, generating corresponding equipment anomaly mask data. This mask, based on pixel-level heat source focusing, indicates the location of overload, jamming, or electrical faults in the transportation equipment. Based on this, anomaly identification of personnel states is performed using the temperature distribution matrix. Specifically, temperature pixels within the typical human body temperature range (e.g., 36°C to 38°C) are selected, and a binary image representation of personnel activity hotspots is constructed based on these pixels. Through pixel clustering analysis and spatial density calculation, it is identified whether there is an abnormally dense crowd gathering in the current hotspot. When the number of such pixels per unit area exceeds a set threshold, or the hotspot area suddenly increases beyond the historical average limit in a short period, this is considered a potential personnel gathering anomaly, and corresponding personnel anomaly mask data is generated. This mask is used to reflect issues such as personnel gathering and lingering, passageway blockage, or delayed evacuation, and is suitable for improving accident early warning and personal safety protection capabilities. Simultaneously, it assesses the thermal uniformity of cargo storage areas in the mine to identify accumulation blockages or transportation delays. This process, based on the temperature gradient matrix in the temperature feature matrix, calculates the temperature difference amplitude and direction of change between adjacent pixels within the cargo storage area. When a spatially uneven distribution of the temperature gradient is found in a certain area—i.e., disordered temperature difference direction, concentrated local hot spots, and severe edge gradients—it is determined that the area has problems such as cargo accumulation, sensor obstruction, or abnormal heat exchange, and cargo anomaly mask data is generated. This process combines temperature gradient statistical parameters and local consistency indicators to construct a thermal stability evaluation model. Pixel-level spatial mapping and merging analysis are performed on equipment anomaly masks, personnel anomaly masks, and cargo anomaly masks. The types of anomalies are classified and labeled according to the overlapping area, distribution area, and duration of each mask. The severity level of the anomaly is comprehensively calculated by combining the location coordinates of the anomaly area in the mine thermal imaging map, the temperature amplitude characteristics within the anomaly time window, and the historical statistical weights of the identification model. The output is a set of transportation anomaly information that includes an anomaly type identifier, anomaly spatial location coordinates, and severity level values.

[0066] Step S103: Perform temperature pattern matching based on transportation anomaly information to generate a transportation operation instruction set;

[0067] Specifically, the anomaly type, location coordinates, and severity level are extracted from the transportation anomaly information. The anomaly type is identified according to predefined categories such as "equipment overheating," "personnel gathering," or "cargo accumulation," and encoded as corresponding anomaly identifiers, each corresponding to a different position in the encoding vector. For example, the equipment overheating identifier is represented as [1,0,0], the personnel gathering identifier as [0,1,0], and the cargo accumulation identifier as [0,0,1]. The location coordinates are then appended to the vector in two-dimensional or three-dimensional numerical form, and the severity level is added as a normalized value to form a complete anomaly feature encoding vector containing semantic labels and numerical attributes. Based on the anomaly feature encoding vector, a pre-defined transportation instruction matching dictionary is queried. This dictionary contains a set of instruction templates, each of which has undergone feature vectorization. The dimensional structure of the vector is consistent with the anomaly feature encoding vector to facilitate efficient similarity calculation. The instruction templates in the dictionary are categorized into multiple types based on historical scheduling experience and safety regulations, such as "suspend equipment," "arrange manual inspection," "evacuate personnel," "switch to backup lanes," and "reassign vehicles." Each type of template is stored in a standard format, including applicable anomaly type codes, parameterization rules, and execution scenario labels. A vector matching algorithm is used to calculate the cosine similarity between the current anomaly feature encoding vector and the feature vector of each instruction template in the candidate template set, measuring the degree of semantic matching between the current anomaly state and each template. When the cosine similarity between a template and the current anomaly vector exceeds a preset threshold (e.g., 0.7), the template is marked as an "activated template," indicating that its content is highly relevant to the current anomaly situation. For activated instruction templates, parameterized reconstruction is performed based on the location coordinates and severity level in the anomaly information. Abstract placeholders in the template are replaced with specific numerical values ​​or area identifiers, generating a transportation operation instruction set with spatial positioning capabilities and risk level adaptability.

[0068] Step S104: Using the transportation operation instruction set as constraints, perform mine tunnel temperature safety path planning to obtain the transportation execution plan;

[0069] Specifically, the transportation operation instruction set is structured and parsed to extract execution area information related to spatial location and time requirement data related to scheduling rhythm. Areas marked as operation centers or work sites in the instructions are marked as "no-entry zones" or "restricted passage zones" on the path planning map to prevent subsequent paths from crossing areas with concentrated personnel or equipment interference, thus avoiding new scheduling conflicts or thermal interference. Simultaneously, time requirements are parsed into passage time constraint windows for path nodes. Semantic descriptions such as "complete within 5 minutes" or "execute with a 30-minute delay" are transformed into passage time interval parameters for corresponding edges or nodes in the path map, constructing a set of path constraint conditions with spatial exclusivity and time scheduling characteristics. The latest mine tunnel temperature distribution map, collected by thermal imaging cameras, is simultaneously accessed. This map is used to construct a high-resolution temperature field model through continuous image processing and spatial interpolation. Key feature values, including average temperature, temperature gradient direction, and rate of change, are extracted from each path node and edge segment. The temperature risk level of each path segment at the current moment is further calculated, and a temperature-weighted path network is constructed accordingly. The path constraints and temperature-weighted path network are input into a mixed-integer programming solver with multi-objective optimization capabilities. The model's objective function consists of four parts: shortest transportation time, lowest temperature risk, minimum congestion, and optimal energy cost. These four objectives are assigned adjustable weight parameters, for example, 0.3, 0.4, 0.2, and 0.1 respectively, ensuring that the model balances transportation efficiency and operational safety during the solution process. The solver uses a heuristic search algorithm based on branch-and-bound and temperature gradient guidance to quickly iterate and generate a set of optimal paths while satisfying all spatial and temporal constraints. This set of paths covers all vehicle scheduling lines, personnel operation routes, and cargo flow trajectories required for the transportation task, forming a structured multi-objective optimization path solution. Based on obtaining the optimal path solution, resources involved in the path are automatically configured. The task objectives of each transportation node are matched with the time period of the path to form a vehicle scheduling schedule. Vehicle start and stop times are allocated according to the order of passage on each path. Combining the location of work points and task content involved in the path, personnel are assigned to positions based on their work ability, job responsibilities, and historical scheduling records to ensure that personnel execute corresponding instructions at reasonable times and in appropriate locations. For goods that need to be moved or handed over, their transfer routes and loading / unloading times are determined according to their classification, weight, and safety level. Equipment operating status and mine shift arrangements are included in the scheduling priority judgment to ensure that high-risk goods are executed first and critical resources are not interrupted. All scheduling information is summarized to generate a transportation execution plan that includes a transportation route map, task timetable, job allocation table, vehicle call records, and emergency plan.

[0070] Step S105: After locating the specific position of the mine workers in real time using a thermal imaging camera, the transportation execution plan is converted into standardized voice commands and pushed to the corresponding workers via walkie-talkie network partitions, and scheduling feedback data including command reception confirmation and execution progress is obtained in real time.

[0071] Specifically, thermal imaging cameras deployed above the work site or at key visual locations such as corners of passageways collect image data in real time. Combined with the characteristic reflection curves of human body temperature in the thermal infrared band, image segmentation algorithms identify target areas with temperatures higher than the background environment and contours conforming to human anatomy. Edge detection, region tracking, and dynamic contour updates are used to extract the contours of mine workers. Based on displacement calculations across multiple frames, the spatial trajectory of individual heat sources is constructed, and coordinate mapping is performed using the mine's internal three-dimensional geographic coordinate system to achieve real-time location tracking of each worker. The obtained location information is synchronously transmitted to the location database in the dispatch center in coordinate encoding form. After location tracking, information including vehicle scheduling times, personnel assignments, and cargo transfer routes is extracted according to the transportation execution plan. Following a pre-defined standard mine operation voice protocol, the textual task description is converted into a voice-reading instruction text, uniformly organized in the order of "recipient—task type—operation content—time requirement," while controlling word conciseness, speech rate adaptability, and redundancy suppression. The generated text commands are input into a deep neural network-based speech synthesis system. This system uses the Tacotron2 model combined with a mining operation terminology dictionary for end-to-end speech synthesis, outputting standardized voice commands in audio format with a speech rate controlled at approximately 150 words per minute and a stable pitch frequency within the range of 300 to 800 Hz. The generated voice commands are then pushed to the corresponding target personnel via a zoned communication mechanism. Based on the personnel coordinates obtained from the aforementioned thermal imaging positioning, the system queries the device geographic distribution table and communication channel allocation table in the walkie-talkie network to identify the walkie-talkie terminal address and communication channel bound to each personnel's current location. The voice commands are then bound to the corresponding channel according to the zone number, forming a zoned push command package with a geographic distribution tag. This push mechanism supports multi-task parallel scheduling, allowing the system to prioritize sending emergency commands and supports retransmission and overlay mechanisms to ensure complete delivery of the voice signal. After receiving the voice command, each target personnel's walkie-talkie terminal responds with a voice reception signal or a button response signal according to the set confirmation process. The system collects this feedback through the walkie-talkie gateway and automatically identifies the corresponding reception time and executor identifier, thereby generating a command reception confirmation record. During the operation, the walkie-talkie terminal continuously uploads the execution progress based on manual operation or sensor integration status, such as standard execution status codes like "Task started," "Arrived at designated location," and "Goods transfer completed." This information is uniformly received, parsed, and updated to the scheduling progress console by the scheduling management system, forming a real-time scheduling feedback data stream. Based on this data stream, the system dynamically adjusts task priorities, reissues instructions, or activates emergency response mechanisms, constructing a closed-loop scheduling control system.

[0072] In one specific embodiment, the process of performing step S101 may specifically include the following steps:

[0073] Temperature data of mine transportation nodes are collected by thermal imaging cameras, and the temperature data of mine transportation nodes are normalized according to the mine environmental reference temperature to obtain standardized temperature data.

[0074] A sliding time window algorithm is used to calculate the temperature weighted average between consecutive frames in the standardized temperature data, and the standardized temperature data is then subjected to time-series smoothing based on the temperature weighted average to obtain time-series smoothed temperature data.

[0075] Calculate the temperature difference between each pixel and its neighboring pixels in the time-smoothed temperature data, and perform spatial denoising on the time-smoothed temperature data based on the temperature difference to obtain spatially denoised temperature data;

[0076] Based on the spatially denoised temperature data, the temperature distribution value of the current frame, the temperature change rate between previous and next frames, and the temperature gradient value between pixels are calculated respectively, and a temperature feature matrix containing the temperature distribution matrix, the temperature change rate matrix, and the temperature gradient matrix is ​​constructed.

[0077] Specifically, raw thermal images acquired by thermal imaging cameras deployed at various transportation nodes in the mine are used as input. These images are a sequence of infrared images composed of consecutive time frames. Each pixel in each frame corresponds to the infrared radiation intensity at that location at a specific time, which is converted into a temperature value defined internally by the thermal imaging equipment. Due to airflow disturbances, multi-source thermal interference, and equipment noise within the mine, these temperature values ​​contain significant dynamic drift and high-frequency fluctuations. Therefore, a unified temperature reference system is established through normalization processing. Based on the environmental reference temperature curve obtained from long-term monitoring of the mine, a dynamic reference temperature range is set (e.g., a reference of 18℃~28℃). The pixel temperature values ​​in each frame of the thermal image are linearly transformed according to their deviation from the reference range, converting them into standardized temperature data. A sliding time window algorithm is used to calculate the weighted average temperature across consecutive frames in the standardized temperature data. By selecting a reasonable time window length (e.g., 3 to 7 frames) and setting the inter-frame step size, the system selects standardized temperature maps of consecutive frames within each sliding window and performs a weighted average of the temperature values ​​of each corresponding pixel on the time axis. The weighting strategy assigns higher weights to frames closer to the center based on time distance. Weighting functions include linear weights, trigonometric function weights, or Gaussian distribution weights. The result is a set of weighted temperature images that integrate short-term thermal change trends and suppress occasional jumps, forming time-stationary smooth temperature data. Local anomalies are identified using the temperature difference information between pixels. For each pixel in each frame of the image, the temperature difference between it and its 8 neighbors (i.e., adjacent pixels in the top, bottom, left, right, and four diagonal directions) is analyzed. When the average temperature difference between a pixel and its neighborhood exceeds a preset threshold (e.g., 3℃), it is determined that there is a local anomaly at that point, caused by measurement error, foreign object occlusion, or thermal reflection. To eliminate this type of local high-frequency noise, the system uses bilateral filtering or median interpolation to reconstruct and replace the anomalous pixel. The interpolation is based on the stability of temperature changes and structural continuity within the neighborhood. The final generated image is the spatially denoised temperature data after spatial consistency optimization. Based on the spatially denoised temperature data, the temperature distribution value of the current frame, the temperature change rate between previous and subsequent frames, and the temperature gradient value between pixels are calculated. The temperature distribution value of the current frame is extracted. This value is a two-dimensional matrix composed of the absolute temperatures of all pixels in the spatially denoised temperature image at the current time point, representing the thermal distribution state of the mine node at the current moment. The difference between the corresponding pixels of the current frame and the previous and next time frames is calculated to obtain the temperature change rate matrix. This matrix reflects the dynamic trend of the mine thermal field changing over time and can capture the heat rise process generated by the operation of transportation equipment or the temperature change caused by collective activities in densely populated areas. The amplitude and direction of the temperature difference between each pixel in the image and its neighboring pixels are calculated in the spatial domain to obtain the temperature gradient matrix. This matrix reflects the directionality and density clustering characteristics of the temperature difference distribution in space.The system combines the three matrices into a structured data object to obtain the temperature feature matrix.

[0078] In one specific embodiment, the process of performing step S102 may specifically include the following steps:

[0079] A time-series analysis is performed on the temperature change rate matrix in the temperature feature matrix to determine the location of abnormal heating in the transportation equipment, and equipment abnormality mask data is generated based on the location of abnormal heating in the transportation equipment.

[0080] Temperature pixels within the target temperature range are selected based on the temperature distribution matrix in the temperature feature matrix, and pixel distribution analysis of densely populated areas is performed based on the temperature pixels to obtain personnel anomaly mask data.

[0081] The temperature distribution uniformity within the cargo storage area is calculated based on the temperature gradient matrix in the temperature feature matrix, and cargo anomaly mask data is determined based on the temperature distribution uniformity.

[0082] Spatial location mapping and severity calculation are performed on equipment anomaly mask data, personnel anomaly mask data, and cargo anomaly mask data to generate transportation anomaly information that includes anomaly type, location coordinates, and severity level.

[0083] Specifically, the system performs abnormal temperature rise analysis on transportation equipment based on the temperature change rate matrix in the temperature feature matrix. Using the change rate matrix corresponding to each time point as input, the system constructs a pixel-level time series vector according to the change trend of each pixel over multiple time frames. By setting a temperature rise threshold (e.g., the cumulative temperature change rate exceeds 15℃ within 5 consecutive minutes), the system performs trend fitting and extreme value judgment on this vector. When a pixel continuously exhibits a high-amplitude temperature rise trend and its location matches the known equipment distribution map in the mine, the system marks it as an abnormal heat source. Using this point as the center, the system performs density clustering analysis on the surrounding pixel blocks using a spatial neighborhood expansion mechanism, and generates equipment anomaly mask data encoded in Boolean logic form in the thermally dense areas. This mask map is used to indicate the spatial location of the transportation equipment where there is continuous high heat output or electrical / mechanical fault. After equipment anomaly identification is completed, the system calls the temperature distribution matrix in the temperature feature matrix to perform anomaly detection in densely populated areas. It focuses on pixels within the normal human body temperature range (set to 36℃ to 38℃) in the temperature distribution map. The system filters the entire temperature map by setting upper and lower thresholds, extracting all target pixel sets that meet the conditions and mapping their positions to the mine's spatial coordinate system. Then, it uses a density-based clustering algorithm (such as DBSCAN or KMeans) to determine the spatial clustering of these target pixels. When the number of target pixels per unit area exceeds a historical baseline, or when the centroids of multiple clusters are close and showing a merging trend, the system determines that excessive personnel gathering has occurred in that area, indicating problems such as blocked evacuation routes, overstaffing, or concentrated waiting. Therefore, the system automatically generates personnel anomaly mask data using the boundary contours of the clustered areas as a reference and compares it with the mine's job distribution map to determine whether the current personnel gathering is within a designated area and outside the safety control range. Simultaneously, to ensure visualized monitoring of the transportation status of cargo storage areas, the system performs anomaly identification tasks based on the temperature gradient matrix in the temperature feature matrix. Under normal conditions, cargo storage areas exhibit strong thermal uniformity due to their large heat capacity and static physical state, appearing as stable regions with low gradients and strong continuity in thermal images. However, when uneven stacking, localized congestion, equipment obstruction, or heat source influence occur, abnormally drastic temperature difference sections or directional vector changes appear in the gradient image. To address this, the system selects all valid pixels within the known cargo area outline and calculates the absolute gradient difference between adjacent pixels to determine the overall temperature distribution uniformity index, such as standard deviation, skewness, or maximum / minimum gradient ratio. When this index exceeds a preset safety tolerance range (e.g., standard deviation exceeding 1.5 times the normal baseline), it is determined to indicate thermal distribution disorder. This generates cargo anomaly mask data covering the area, recording its center point location and gradient anomaly weight level.Equipment anomaly mask data, personnel anomaly mask data, and cargo anomaly mask data are used as multi-channel inputs for spatial location fusion. Each mask point is then spatially mapped using a unified coordinate system. Simultaneously, the severity level of each anomaly point is calculated based on its type label, the thermal characteristics of its location, duration, and trend. This calculation employs a weighted scoring method for quantitative assessment, using factors such as the rate of thermal change of the anomaly area, its size, and the degree of mask overlap as primary factors. Each factor has a maximum value of 100 points and is assigned different weight combinations, ultimately yielding a risk score for each anomaly point. Based on the score range, anomalies are categorized into three levels: "minor anomaly," "moderate anomaly," and "severe anomaly." A structured transportation anomaly information dataset is generated based on the three types of anomaly mask results, location coordinates, and severity levels. Each data record includes key fields such as anomaly type (e.g., high equipment temperature, personnel gathering, cargo accumulation), spatial coordinates (e.g., X, Y or X, Y, Z), and severity level (e.g., levels 1-3).

[0084] In one specific embodiment, the process of calculating the temperature distribution uniformity within the cargo storage area based on the temperature gradient matrix in the temperature feature matrix, and determining cargo anomaly mask data based on the temperature distribution uniformity, can specifically include the following steps:

[0085] Extract regions from the temperature gradient matrix of the temperature feature matrix that have a temperature gradient smaller than a preset gradient range as cargo storage areas;

[0086] The temperature distribution uniformity is calculated based on the cargo storage area, and candidate areas of abnormal temperature distribution are identified based on the temperature distribution uniformity.

[0087] By using candidate regions with abnormal temperature distribution and tracking temperature changes, abnormal areas of cargo accumulation are identified. These abnormal areas are then converted into binary codes to construct cargo anomaly mask data corresponding to the temperature gradient matrix.

[0088] Specifically, a temperature gradient matrix is ​​extracted from the temperature feature matrix. This matrix reflects the magnitude and direction of the temperature gradient around each pixel in the spatial direction. In a typical mining environment, the cargo storage area is a low-gradient region with uniform heat conduction and a relatively stable spatial structure. This is because the stacked cargo, due to its large heat capacity and lack of significant heat source fluctuations, will not experience drastic temperature gradient changes in a short period of time. Based on this physical characteristic, a gradient threshold range (e.g., 0℃ / pixel ~ 1.2℃ / pixel) is set as the gradient screening criterion for cargo storage areas. All connected regions that meet this gradient condition are identified by scanning all pixels in the image, thus initially extracting a set of candidate cargo storage areas. Area filtering and boundary rule judgment operations are performed on the initially screened gradient regions to remove pseudo-regions with excessively small areas, irregular boundaries, or abrupt shape changes. At the same time, the remaining valid regions are compared with the spatial layout model of the mining operation scenario to confirm whether they fall within the designed cargo stacking area. If a match is successful, the region is marked as a valid cargo storage area map. This map is in the form of a two-dimensional Boolean matrix, where a value of 1 indicates that the current pixel belongs to the cargo storage area, and a value of 0 indicates that it is not a cargo area. After obtaining the cargo storage areas, the system performs statistical analysis on the temperature distribution of these areas based on spatially denoised temperature data, calculating the corresponding temperature distribution uniformity index. This index is modeled using multiple dimensions, such as temperature standard deviation (reflecting fluctuation amplitude), root mean square gradient (reflecting detailed undulations), and center-edge difference (reflecting local concentration). If the temperature standard deviation of a certain area is too high spatially (e.g., exceeding 0.8℃), or its internal gradient variance deviates significantly from the average value of the background area, the system considers that the temperature uniformity of that area has been disrupted. This result is caused by abnormal stacking structure, obstruction, abnormal cargo flow leading to changes in heat source distribution, etc. Therefore, the system isolates these sub-regions with indices exceeding the threshold from the cargo storage area map, constructing a set of candidate regions for abnormal temperature distribution. The system incorporates a time dimension for dynamic verification. Based on candidate regions with abnormal temperature distribution, a temperature change pattern tracking algorithm is executed. This algorithm extracts the temperature change trend vectors of pixels in these regions over multiple consecutive time frames and compares them with historical thermal change models of normal cargo areas. If the abnormal region exhibits non-periodic cumulative temperature increases, continuous disruption of the thermal distribution structure, and unstable migration trajectories of hot and cold spots, it is further confirmed as a genuine cargo accumulation anomaly region. After confirming the cargo accumulation anomaly region, the system generates a cargo anomaly mask image with the same size as the temperature gradient matrix, based on its spatial location in the image. This image is organized in binary encoding, with each pixel corresponding to a Boolean value: 1 indicates that the point belongs to the cargo accumulation anomaly region, and 0 indicates a normal region or an undefined region.

[0089] In one specific embodiment, the process of executing step S103 may specifically include the following steps:

[0090] Extract the anomaly type, location coordinates, and severity level from the transportation anomaly information, and encode the anomaly type as equipment overheating identifier, personnel gathering identifier, and cargo stacking identifier to obtain an anomaly feature encoding vector;

[0091] Based on the abnormal feature encoding vector, a pre-set transportation instruction matching dictionary is queried to obtain a set of candidate instruction templates;

[0092] The cosine similarity between the abnormal feature encoding vector and the feature vector of each instruction template in the candidate instruction template set is calculated. When the similarity value exceeds the similarity threshold, the corresponding instruction template is activated and a transportation operation instruction set is generated parameterized according to the location coordinates and severity level.

[0093] Specifically, the system extracts the anomaly type, location coordinates, and severity level from the transportation anomaly information. The anomaly type field is the result of the system's classification and identification, including three basic types: "equipment overheating," "personnel gathering," and "cargo accumulation." Each type corresponds to a fixed semantic identifier within the system. The location coordinate field records the center location of the anomaly occurrence area in two-dimensional or three-dimensional spatial coordinates for performing area localization. The severity level field is a level label generated by the system based on comprehensive factors such as the anomaly's impact range, duration, and thermal fluctuation amplitude. It ranges from level 1 to level 3, representing mild, moderate, and severe anomalies, respectively. The system uses these three types of fields as main parameters, extracts their semantic content to construct anomaly feature encoding vectors. The anomaly type field is converted into three binary identifier bits through one-hot encoding, corresponding to equipment overheating, personnel gathering, and goods accumulation, respectively. For example, if the current anomaly is equipment overheating, the corresponding vector prefix is ​​[1,0,0]. The location coordinate field is compressed into two or three floating-point values ​​through spatial encoding and appended to the middle of the vector. The severity level field is normalized and appended to the end, forming a set of structured, multi-dimensional numerical semantic feature vectors to describe the quantitative expression of the abnormal event in the three dimensions of type, location, and severity. The system invokes a pre-defined transportation instruction matching dictionary, which contains a set of transportation instruction templates constructed from expert systems or historical scheduling data. Each template describes a standard operating procedure for handling a specific abnormal situation, such as "immediately suspend equipment in a designated area," "evacuate personnel from a designated area," or "initiate cargo transfer and diversion." Each instruction template includes a description of the applicable abnormality type, semantic scenario, execution action, and applicable level range, along with a feature vector. This feature vector is constructed using the same structure as the abnormality feature encoding vector, ensuring consistency in vector space dimensions and dimensional meaning. The system compares the current abnormality feature encoding vector one-to-one with the feature vector of each instruction template in the dictionary, and calculates the cosine value of the angle between their vectors using the cosine similarity formula, serving as a semantic matching metric between the two semantic vectors. The closer the cosine value is to 1, the more consistent the directions of the two vectors in the semantic space, meaning the more similar the current abnormal scenario is to the instruction logic designed in the template. The system filters the results by setting a similarity threshold (e.g., 0.75 or 0.8), and all instruction templates with similarity exceeding the threshold are considered as the set of candidate instruction templates activated by the current abnormality. After the instruction templates are activated, the system generates these templates parametrically based on the specific spatial location and severity level of the anomaly.The instruction template itself exists in a structured instruction text format, containing multiple replaceable placeholders such as "{location}", "{time}", and "{level}". The system directly replaces the location placeholder field in the instruction template with the location coordinates extracted from the exception information. Simultaneously, it determines the urgency label based on the severity level; for example, level 1 corresponds to "complete within 30 minutes", level 2 to "execute immediately", and level 3 to "emergency priority dispatch". The system then replaces the execution time description or priority field in the instruction accordingly. If the template contains parameter fields such as vehicle dispatch, personnel allocation, and route direction, the system further semantically fills in the information based on the current transportation resource pool status and channel topology. For example, in "dispatch the nearest available vehicle to {location}", the vehicle information is filled with "vehicle number 5", and the route information is filled with "channel A → junction area → storage and transportation point", thus constructing a transportation operation instruction set.

[0094] In one specific embodiment, the process of querying a preset transportation instruction matching dictionary based on the abnormal feature encoding vector to obtain a set of candidate instruction templates can specifically include the following steps:

[0095] Parse the anomaly type identifier in the anomaly feature encoding vector. When the anomaly type identifier is the equipment overheat identifier, query the equipment class instruction branch. When the anomaly type identifier is the personnel gathering identifier, query the personnel class instruction branch. When the anomaly type identifier is the cargo stacking identifier, query the cargo class instruction branch to obtain the transportation instruction matching dictionary.

[0096] The location coordinates are read from the transportation instruction matching dictionary, matched with the mine tunnel topology database to determine the connecting passages and adjacent facilities in the abnormal area, and the location-matching instruction templates are filtered out.

[0097] Based on the abnormal feature encoding vector, the location adaptation instruction template is filtered according to the emergency response intensity to obtain the level matching instruction template;

[0098] Cross-validate the level matching instruction template with the timeliness requirements of exception handling to form a set of candidate instruction templates.

[0099] Specifically, the system performs semantic parsing on the abnormal feature encoding vector. This vector consists of multiple fields, with the first three bits using one-hot encoding to correspond to three core abnormality types: equipment overheating, personnel gathering, and cargo accumulation. The system determines the current abnormality type by detecting the identifier values ​​in the vector. For example, a vector prefix of [1,0,0] corresponds to equipment overheating, [0,1,0] to personnel gathering, and [0,0,1] to cargo accumulation. The system performs type-branch targeted query operations by matching the abnormality type identifier. That is, it calls the corresponding sub-branch content from the unified stored transportation instruction matching dictionary. If it is an equipment-related abnormality, it only queries the "equipment instruction branch," which contains operation templates related to equipment response, such as "shutdown," "maintenance," and "dispatch electrician." If it is a personnel-related abnormality, it queries the "personnel instruction branch," which includes crowd control templates such as "evacuation," "ventilation," and "broadcast notification." If it is a cargo-related abnormality, it accesses the "cargo instruction branch," which includes operation templates such as "transfer," "stacking optimization," and "cargo area sealing." The system needs to perform a spatial matching process on all instruction templates selected from the above-mentioned branches, based on the location coordinate field contained in the anomaly feature encoding vector. The system parses the scope or applicable area description from the instruction template and compares it with the location coordinates in the anomaly code. To enhance matching accuracy, a mine tunnel topology database is introduced. This database stores the connection relationships of spatial elements such as mine transportation nodes, intersecting passages, equipment locations, and personnel positions in a graph structure, supporting queries for the passage connection area, adjacent facility nodes, and interactive operation areas corresponding to any coordinate point. The system inputs the anomaly coordinates into the database to retrieve the passage number, connection path, and facility layout within its radius. Then, it matches the search results with the applicable area label bound to each instruction template. When the execution area required by the template intersects with the anomaly passage or its reachable path, it can be marked as a location-adapted instruction template. For example, if a template is applicable to "passage C section to main ventilation opening", and the anomaly coordinates are located in passage C section or reachable from the main ventilation opening in one hop, then the spatial location adaptation of the template is considered valid. After spatial adaptation, the system performs emergency response level filtering on the adapted instruction templates based on the severity level field in the anomaly feature encoding vector. This involves comparing the minimum response level threshold set by the instruction template with the current anomaly level to see if it matches or exceeds it. For example, if template A is set to be used only for a Level 1 mild response, but the current anomaly is a Level 2 moderate response, template A will be excluded; while template B, applicable to Levels 1-3 responses, will be retained as a level-matching template. This process determines the overlap between the anomaly level and the template level range, using whether the response intensity meets the minimum requirements as the filtering criterion. The system cross-validates the set of level-matching templates with the processing timeliness rules required by the anomaly type.The timeliness requirement consists of two parts: first, the risk level of the anomaly itself and its impact on response time. For example, an overheating anomaly with a severity level of 3 requires control actions to be executed within 3 minutes; second, the execution delay definition of the template itself. If a template defines an execution preparation time of 10 minutes, it cannot adapt to a 3-minute response scenario. The system compares and analyzes the timeliness label of each template with the maximum allowable response delay required by the anomaly, eliminating all templates that do not meet the response time limit and retaining all templates that meet the timeliness requirement, ultimately forming a candidate instruction template set.

[0100] In one specific embodiment, the process of executing step S104 may specifically include the following steps:

[0101] The execution area and time requirements of the transportation operation instruction set are analyzed. The execution area is marked as a prohibited area in the route planning. At the same time, the time requirements are converted into time window parameters for route planning to obtain the constraints.

[0102] Based on the real-time monitoring of temperature distribution data in mine tunnels by thermal imaging cameras, and based on the temperature distribution data in mine tunnels, feature extraction is performed to construct a temperature weighted path network.

[0103] The constraints and temperature-weighted path network are input into a mixed-integer programming solver for multi-objective optimization, resulting in a multi-objective optimized path solution.

[0104] Based on the multi-objective optimization path solution, specific vehicle scheduling times, personnel assignments, and cargo transfer routes are allocated. Combined with the mine work shift schedule and equipment operating status, a detailed time plan and resource allocation table are generated to form a transportation execution plan.

[0105] Specifically, the transportation operation instruction set uses structured semantics to describe the key parameters of each scheduling action, including the type of operation (such as dispatching vehicles, evacuating personnel, transferring goods, etc.), the corresponding spatial coordinates, and specific execution time limits. The instruction set is parsed item by item, extracting the execution area coordinates for each instruction. These coordinates are then spatially mapped using the mine tunnel topology map to identify all tunnel segments directly connected to or affected by the execution area. These segments are marked as prohibited areas within the corresponding time window, with time boundaries set for prohibition. Simultaneously, the system converts semantic time limit fields such as "execute immediately," "complete within 10 minutes," and "complete before operation begins" into standardized time window parameters. These parameters are uniformly represented as intervals of [start time, end time] using a time parser, forming the time constraint dimension in the path planning process. By parsing the execution area and time requirements of all instructions, a set of path planning constraints, including spatial exclusion constraints and time interval limitations, is constructed. The system uses real-time thermal imaging data of mine tunnels acquired by thermal imaging cameras to model the temperature distribution of the entire mine tunnel grid. Image processing operations are performed on the thermal images of each tunnel segment, including denoising, normalization, and time-series smoothing. Then, statistical temperature characteristics such as average temperature, maximum temperature, temperature change rate, and thermal fluctuation amplitude are calculated for each tunnel segment and converted into temperature cost factors. Each tunnel edge is assigned a temperature weight value; a higher value indicates lower thermal safety for that tunnel segment. Based on this, the system establishes a temperature-weighted path network graph with tunnel nodes as vertices and tunnel segments as edges. This graph is stored in the form of an adjacency matrix, where edge weights are determined by geometric distance, thermal cost, and accessibility. If a tunnel segment experiences overheating or overlaps with a restricted area in the current or predetermined time, its weight is dynamically increased or it is marked as unusable. The system continuously updates the thermal accessibility of this network, forming a realistic, variable thermal environment graph structure that is linked to scheduling commands. The constraints and temperature-weighted path network are then input into a mixed-integer programming solver for optimal path planning. The solution model is based on a multi-objective function, which includes multiple sub-objectives such as minimizing total transportation time, minimizing corridor thermal risk, minimizing route congestion, and minimizing transportation energy consumption. Each sub-objective is assigned a weight coefficient (e.g., transportation time 0.3, thermal risk 0.4, congestion 0.2, energy consumption 0.1). The system integrates multiple objectives through weighted linear combination or hierarchical optimization strategies, while introducing a previously constructed set of constraints to strictly limit route feasibility. In terms of variable definition, the system uniformly models route node selection, corridor segment activation status, time period passage identifiers, vehicle numbers, etc., as integer or binary variables, and establishes a logical expression of time consistency, resource exclusivity, and task scheduling sequence through constraint relationships.The model is solved using algorithms such as branch and bound, Lagrange relaxation, or heuristic search, ultimately outputting a set of optimal paths, resulting in a multi-objective optimized path solution. This solution includes the start and end nodes, travel time, path number, and thermal cost assessment results for each task. Based on this optimized path solution, the system organizes resource-level scheduling information, assigning path segments to specific vehicles. It then uses path length and travel time to deduce vehicle scheduling timetables, ensuring vehicles enter designated channels at the most suitable time and complete tasks within the period of least channel congestion risk. For personnel assignments, the system analyzes the work points, equipment nodes, and safety checkpoints involved in the path, matching each scheduling task to the corresponding duty post or personnel group, forming a personnel assignment plan. For tasks involving cargo transfer, the system maps the path to a cargo transfer route map based on the storage node or unloading area number connected to the path's endpoint, combined with transportation priority and cargo type, and marks the task execution order. To ensure the overall transportation system operates in sync with actual mine operations, the system integrates mine shift scheduling data and equipment operation status monitoring data to avoid conflicts between route planning and planned times. If equipment is under maintenance or out of service during the scheduling period, the accessibility of the corresponding route is reassessed. The system integrates information such as multi-objective route solutions, scheduling times, job assignments, cargo flow, and operation time constraints to generate a structured transportation execution plan. This plan includes multiple scheduling forms, including a time plan, resource allocation table, channel usage table, risk heat map, and list of backup emergency routes, and supports delivery to personnel via graphical interface or voice commands.

[0106] In one specific embodiment, the process of inputting the constraint conditions and temperature-weighted path network into a mixed-integer programming solver for multi-objective optimization to obtain the multi-objective optimized path solution can specifically include the following steps:

[0107] Based on the constraints, the restricted area is represented as a binary variable constraint, the time window parameter is represented as an inequality constraint, and the temperature weighted path network is converted into a 0-1 integer variable matrix for path selection, thus obtaining a mixed integer programming solver.

[0108] The objective functions are: transportation time objective function to calculate the ratio of total route length to speed; temperature safety objective function to calculate the cumulative value of route temperature risk; route congestion objective function to count the frequency of route use during the same period; and energy cost objective function to accumulate the power consumption of vehicles and equipment, thus obtaining a comprehensive objective function.

[0109] The root node of the branch and bound algorithm is initialized based on the mixed integer programming solver, and the upper bound of the search is set to positive infinity and the lower bound is set to the current optimal solution. The branch operation is performed on the integer variables one by one to generate child nodes. The linear relaxation solution of each child node is calculated as the lower bound estimate of the corresponding child node to obtain the branch search tree.

[0110] Traverse the active nodes that have not been pruned in the branch search tree. When the lower bound of an active node exceeds the current optimal upper bound, prune the corresponding branch. When an active node obtains an integer feasible solution and the objective function value of the combined objective function is greater than the current optimal solution, update the optimal solution. Repeat the branch search until all nodes have been processed to obtain the multi-objective optimized path solution.

[0111] Specifically, the restricted areas and time requirements parsed from the scheduling instructions are converted into formal constraint expressions, and the temperature-weighted path network generated by thermal imaging is structured into a variable matrix that can be recognized by the optimizer. In the restricted area modeling, the system encodes the edges involved in the restricted areas in the path graph with a Boolean variable, representing whether a path segment is enabled. If a path segment is in a restricted state, a fixed-value constraint is forcibly added to exclude it from the planning process. In terms of time requirement modeling, the system defines a passage time variable for each path node and adds inequality constraints to it according to the task execution requirements, such as stipulating that the node's passage time must not be earlier than a certain moment or must be completed within a certain time window. The system establishes continuity constraints through the time sequence logic between path segments to ensure the time transmission relationship between path segments. Through this step, the path passage status and scheduling rhythm are uniformly encoded as a combination of 0-1 integer variables and continuous time variables, forming the variable domain of the mixed integer programming model. The system constructs a multi-objective evaluation mechanism to support path selection, that is, defining four types of objective functions: transportation time, temperature risk, channel congestion, and energy consumption cost, and integrating them into a comprehensive optimization objective. In the modeling of the transportation time objective function, the system uses the ratio of total path length to transportation speed as an evaluation index to measure the timeliness of the transportation path. The temperature safety objective function is based on the temperature risk coefficient of each path segment in the thermal imaging path map, accumulating the thermal safety cost of the path to evaluate the heat exposure degree of the entire path. The channel congestion function is used to count the frequency of multiple transportation tasks using a certain path segment simultaneously within the same time period; the larger the value, the higher the degree of resource competition. The energy consumption cost function is obtained by accumulating the power consumption of the corresponding vehicles and equipment on the path segment. These four objectives are assigned corresponding weights according to actual operational preferences and combined into a comprehensive objective function, which serves as the search index for the optimizer, controlling energy consumption and resource usage conflicts while ensuring thermal safety and scheduling efficiency. After constructing the variable set and objective function, the system starts the mixed integer programming solver and initializes the branch and bound algorithm, taking the root node as the starting point for all variables in an undetermined state, setting the upper bound of the search to positive infinity, and the lower bound to the currently known optimal solution or the initial feasible solution result. The system selects an unassigned integer variable and performs branching operations on it under two possible value states, generating two child nodes for each. Within each child node, the integer condition is temporarily relaxed, and a linear problem is solved to obtain a linear relaxation estimate. This estimate serves as the lower bound of the target for that node, used to determine if there is a potential for a better solution. All child nodes are added to the search tree, and their status is recorded as active. The system traverses the active nodes in the tree one by one. For each node, if its lower bound exceeds the upper bound of the current optimal solution, it is considered to have no solution space and is pruned directly; if its solution is an integer feasible solution, and the corresponding comprehensive objective function value is better than the current optimal solution, then the current optimal solution is updated.This process is repeated cyclically. The system continuously updates the search boundary, splits variable branches, evaluates subproblem estimates, and prunes invalid paths, gradually converging to a set of integer solutions that satisfy all constraints and have global optimality. As the search tree expands and is pruned for optimization, the system eventually obtains a set of path structures with the enabled path segment variables in an active state. This structure is a thermally safe and feasible path combination that has a reasonable timing arrangement, controlled energy consumption, and minimal resource conflicts—that is, a multi-objective optimized path solution.

[0112] In one specific embodiment, the process of executing step S105 may specifically include the following steps:

[0113] Based on thermal imaging cameras, the specific location of miners can be determined by detecting human body temperature characteristics and recognizing human body contours.

[0114] The vehicle scheduling time, personnel division of labor and cargo transfer routes in the transportation execution plan are converted into voice command text according to the standard voice specifications for mine operations, and standardized voice commands are generated based on the voice command text.

[0115] Based on the specific location of the mine workers, the walkie-talkie network is queried, the walkie-talkie channel and device address corresponding to the area of ​​each worker are matched, and standardized voice commands are pushed to the target workers in different areas.

[0116] Monitor the voice confirmation signals and execution status reports returned by each walkie-talkie terminal, and obtain scheduling feedback data including instruction reception confirmation and execution progress in real time.

[0117] Specifically, high-precision infrared thermal imaging cameras deployed in the mine collect real-time thermal image data from different areas. These cameras have long-wave infrared recognition capabilities, identifying the distribution of human body temperature characteristics. By segmenting and extracting bright pixels in the image frame with temperatures between 36 and 38 degrees Celsius, and combining connected component analysis, thermal density aggregation judgment, and contour edge detection algorithms, the system identifies heat source areas with typical human-shaped contour structures. The system further combines an image stabilization tracking module to match and update targets between consecutive frames, enabling continuous spatial positioning of individual workers. Simultaneously, the system assigns a unique identification code to each identified human heat source and maps its location coordinates to the pre-modeled channel grid structure in the mine map system, thereby achieving precise location marking of each worker's current location. This location data is continuously updated over time, forming a thermal image data structure with personnel numbers and location tags. The system initiates a dispatch instruction conversion module. This module takes a pre-generated transportation execution plan as input, which includes structured data across multiple dimensions, such as vehicle dispatch times, personnel assignments, and cargo transfer routes and time sequences. The system transcribes this information into semantically clear, complete, and redundancy-controlled voice instruction text according to the mine dispatch standard voice protocol, following a template of "recipient-task content-operation area-execution time limit." To ensure the voice expression is adaptable to the high-noise environment of the mine and maintains structural standardization, a specially trained voice generation model performs speech synthesis on the aforementioned text instructions. This model is based on a deep neural network speech synthesis architecture and incorporates a mine-specific terminology vocabulary, controlling the speech rate to approximately 150 words per minute and adjusting the pitch frequency to a suitable audible range under ventilation equipment noise. The resulting voice instruction output is a standardized voice instruction audio stream. Based on the specific location of the mine worker, the system accesses the walkie-talkie area index table and device address binding database of the mine communication network to find the corresponding communication channel number and physical device code for the worker's work area within the communication system. For example, if a person is located in the B zone, the system retrieves the currently active intercom channel in B and the unique address code of the intercom device carried by the person from the database, and binds the standardized voice command audio to this communication identifier to achieve zoned and targeted push of voice data. Simultaneously, to ensure real-time performance and reliability during communication, the push system transmits all task commands according to priority levels. Emergency dispatch tasks are set as first-level channels for priority transmission, while ordinary tasks are broadcast in a time-based polling manner by area, and a multi-channel redundant broadcast mechanism ensures command delivery rate.After a command is successfully sent to the target operator, the walkie-talkie terminal will automatically parse or manually confirm the received command content and generate a voice or digital confirmation signal through a predefined response mechanism. This signal is transmitted back to the central dispatch and control system via the walkie-talkie's reverse channel. Upon receiving the confirmation signal, the system marks the operator's "command reception status" as "confirmed" and records the response time and signal strength, forming a command reception confirmation data record. During the operator's execution, if the walkie-talkie terminal is integrated with external auxiliary sensing modules (such as displacement sensors, cargo status sensors, and personnel position signals), it will also periodically transmit task execution status, such as key status codes like "arrived at task point," "cargo loaded," and "equipment commissioned." The dispatch system receives these in real time, updates the execution progress indicator, and issues risk warnings for tasks that fail to respond within the time limit, thus constructing a closed-loop dispatch feedback data link.

[0118] The intelligent mine transportation method in the embodiments of the present invention has been described above. The intelligent mine transportation system in the embodiments of the present invention will be described below. Please refer to [link / reference]. Figure 2 One embodiment of the intelligent mine transportation system in this invention includes:

[0119] The denoising module 201 is used to perform time-series filtering and gradient denoising on the temperature data of mine transportation nodes collected by the thermal imaging camera to obtain a temperature feature matrix.

[0120] The identification module 202 is used to identify transportation anomaly information based on the temperature feature matrix;

[0121] Matching module 203 is used to perform temperature pattern matching based on transportation anomaly information and generate a transportation operation instruction set;

[0122] Planning module 204 is used to plan the temperature safety path of the mine passage using the transportation operation instruction set as constraints, and obtain the transportation execution plan;

[0123] The push module 205 is used to locate the specific location of the mine workers in real time through the thermal imaging camera, convert the transportation execution plan into standardized voice commands and push them to the corresponding workers through the walkie-talkie network, and obtain scheduling feedback data including command reception confirmation and execution progress in real time.

[0124] Through the collaborative efforts of the aforementioned components, temperature data is collected by a thermal imaging camera and processed with temporal filtering and gradient denoising to construct a three-dimensional temperature feature matrix containing temperature distribution, rate of change, and gradient. Compared to traditional point-based monitoring methods, this approach can comprehensively perceive the temperature distribution in the mine's transportation area. Temperature temporal masking technology is used to accurately identify three types of anomalies: equipment overheating, personnel gathering, and cargo accumulation. Specialized algorithms, such as 30-second sliding window analysis, 36-38°C human body temperature detection, and temperature gradient uniformity analysis, enable automated conversion from temperature features to abnormal states, overcoming the subjectivity and lag of traditional manual inspections. Based on anomaly feature encoding vectors and a transportation instruction matching dictionary, cosine similarity calculation and priority allocation mechanisms are employed to achieve intelligent mapping from temperature anomaly patterns to standardized transportation instructions. Compared to traditional manual decision-making or simple rule engines, this significantly improves the intelligence level and response speed of instruction generation. Transportation operation instructions are transformed into path constraints. A temperature-weighted path network is established based on real-time temperature monitoring data. Mixed-integer programming and branch-and-bound algorithms are used to simultaneously optimize four objectives: transportation time, temperature safety, path congestion, and energy cost, achieving intelligent path planning that considers temperature safety factors. Thermal imaging cameras detect human body temperature characteristics, and connected component analysis algorithms accurately locate the positions of workers. This is combined with a walkie-talkie network topology to achieve zoned push notifications. A three-level priority mechanism ensures the timely transmission of important instructions, overcoming the blindness and inefficiency of traditional broadcasting methods. Voice recognition technology is used to analyze worker feedback, track instruction execution progress and anomalies in real time, and generate scheduling feedback data including reception confirmation rate, execution completion rate, and anomaly feedback.

[0125] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0126] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an intelligent mine transportation device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0127] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for intelligent transportation in mines, characterized in that, include: Temporal filtering and gradient denoising were performed on the temperature data of mine transportation nodes acquired by thermal imaging cameras to obtain the temperature feature matrix. The temperature feature matrix includes the temperature distribution matrix, the temperature rate of change matrix, and the temperature gradient matrix; Identifying transportation anomaly information based on the temperature feature matrix includes: performing time-series analysis on the temperature change rate matrix in the temperature feature matrix to determine the location of abnormal heating in the transportation equipment, and generating equipment anomaly mask data based on the abnormal heating location of the transportation equipment; filtering temperature pixels within a target temperature range based on the temperature distribution matrix in the temperature feature matrix, and performing pixel distribution analysis on densely populated areas based on the temperature pixels to obtain personnel anomaly mask data; calculating the temperature distribution uniformity within the cargo storage area based on the temperature gradient matrix in the temperature feature matrix, and determining cargo anomaly mask data based on the temperature distribution uniformity; and performing spatial location mapping and anomaly severity calculation on the equipment anomaly mask data, the personnel anomaly mask data, and the cargo anomaly mask data. The system generates transportation anomaly information containing anomaly type, location coordinates, and severity level. Based on this information, it performs temperature pattern matching to generate a transportation operation instruction set, including: extracting the anomaly type, location coordinates, and severity level from the anomaly information, and encoding the anomaly type as an equipment overheating identifier, a personnel gathering identifier, and a cargo stacking identifier to obtain an anomaly feature encoding vector; querying a preset transportation instruction matching dictionary based on the anomaly feature encoding vector to obtain a candidate instruction template set; calculating the cosine similarity between the anomaly feature encoding vector and the feature vectors of each instruction template in the candidate instruction template set; when the similarity value exceeds a similarity threshold, the corresponding instruction template is activated and a transportation operation instruction set is generated parameterized according to the location coordinates and the severity level. The transportation operation instruction set is used as a constraint to plan the temperature safety path of the mine passage, resulting in a transportation execution plan. After locating the specific positions of mine workers in real time using thermal imaging cameras, the transportation execution plan is converted into standardized voice commands and pushed to the corresponding workers via walkie-talkie network partitions, and scheduling feedback data including command reception confirmation and execution progress is obtained in real time.

2. The intelligent mine transportation method according to claim 1, characterized in that, The temperature feature matrix is ​​obtained by performing time-series filtering and gradient denoising on the temperature data of mine transportation nodes acquired by the thermal imaging camera, including: Temperature data of mine transportation nodes are collected by thermal imaging cameras, and the temperature data of the mine transportation nodes are normalized according to the mine environmental reference temperature to obtain standardized temperature data. A sliding time window algorithm is used to calculate the temperature weighted average value between consecutive frames in the standardized temperature data, and the standardized temperature data is then subjected to time-series smoothing based on the temperature weighted average value to obtain time-series smoothed temperature data. Calculate the temperature difference between each pixel and its neighboring pixels in the time-smoothed temperature data, and perform spatial denoising on the time-smoothed temperature data based on the temperature difference to obtain spatially denoised temperature data. Based on the spatially denoised temperature data, the temperature distribution value of the current frame, the temperature change rate between previous and next frames, and the temperature gradient value between pixels are calculated respectively, and a temperature feature matrix including the temperature distribution matrix, the temperature change rate matrix, and the temperature gradient matrix is ​​constructed.

3. The intelligent mine transportation method according to claim 1, characterized in that, The step of calculating the temperature distribution uniformity within the cargo storage area based on the temperature gradient matrix in the temperature feature matrix, and determining cargo anomaly mask data based on the temperature distribution uniformity, includes: The regions with temperature gradients less than a preset gradient range are extracted from the temperature gradient matrix of the temperature feature matrix and designated as cargo storage areas. The temperature distribution uniformity is calculated based on the cargo storage area, and candidate areas of abnormal temperature distribution are identified based on the temperature distribution uniformity. By using the candidate regions of abnormal temperature distribution and adopting a temperature change tracking mode, abnormal areas of cargo accumulation are identified, and the abnormal areas of cargo accumulation are converted into binary code form to construct cargo abnormality mask data corresponding to the temperature gradient matrix.

4. The intelligent mine transportation method according to claim 1, characterized in that, The step of querying a preset transportation instruction matching dictionary based on the abnormal feature encoding vector to obtain a candidate instruction template set includes: Parse the anomaly type identifier in the anomaly feature encoding vector. When the anomaly type identifier is an equipment overheat identifier, query the equipment class instruction branch. When the anomaly type identifier is a personnel gathering identifier, query the personnel class instruction branch. When the anomaly type identifier is a cargo stacking identifier, query the cargo class instruction branch to obtain the transportation instruction matching dictionary. Location coordinates are read from the transportation instruction matching dictionary, and the connecting passages and adjacent facilities of the abnormal area are determined by matching them with the mine tunnel topology database. Location-adaptive instruction templates are then selected. Based on the abnormal feature encoding vector, the location-adaptive instruction templates are graded according to the emergency response intensity to obtain grade-matching instruction templates. The level matching instruction template is cross-validated with the timeliness requirements for exception handling to form a set of candidate instruction templates.

5. The intelligent mine transportation method according to claim 1, characterized in that, The step of using the transportation operation instruction set as a constraint to plan the temperature safety path of the mine tunnel to obtain a transportation execution plan includes: The execution area and time requirements of the transportation operation instruction set are analyzed, the execution area is marked as a prohibited area in the route planning, and the time requirements are converted into time window parameters for route planning to obtain the constraints. Based on the real-time monitoring of mine tunnel temperature distribution data by thermal imaging cameras, and based on the mine tunnel temperature distribution data, feature extraction is performed to construct a temperature weighted path network. The constraints and the temperature-weighted path network are input into a mixed-integer programming solver for multi-objective optimization to obtain the multi-objective optimized path solution. Based on the multi-objective optimized path solution, specific vehicle scheduling times, personnel assignments, and cargo transfer routes are allocated. Combined with the mine work shift schedule and equipment operating status, a detailed time plan and resource allocation table are generated to form a transportation execution plan.

6. The intelligent mine transportation method according to claim 5, characterized in that, The step of inputting the constraints and the temperature-weighted path network into a mixed-integer programming solver for multi-objective optimization to obtain a multi-objective optimized path solution includes: Based on the constraints, the restricted area is represented as a binary variable constraint, the time window parameter is represented as an inequality constraint, and the temperature weighted path network is converted into a 0-1 integer variable matrix for path selection, thus obtaining a mixed integer programming solver. The objective functions are: transportation time objective function to calculate the ratio of total route length to speed; temperature safety objective function to calculate the cumulative value of route temperature risk; route congestion objective function to count the frequency of route use during the same period; and energy cost objective function to accumulate the power consumption of vehicles and equipment, thus obtaining a comprehensive objective function. The root node of the branch and bound algorithm is initialized based on the mixed integer programming solver and the upper bound of the search is set to positive infinity and the lower bound is set to the current optimal solution. Branch operations are performed on integer variables one by one to generate child nodes. The linear relaxation solution of each child node is calculated as the lower bound estimate of the corresponding child node to obtain the branch search tree. Traverse the active nodes that have not been pruned in the branch search tree. When the lower bound of an active node exceeds the current optimal upper bound, prune the corresponding branch. When an active node obtains an integer feasible solution and the objective function value of the integrated objective function is greater than the current optimal solution, update the optimal solution. Repeat the branch search until all nodes have been processed to obtain the multi-objective optimized path solution.

7. The intelligent mine transportation method according to claim 1, characterized in that, After locating the specific positions of mine workers in real time using thermal imaging cameras, the transportation execution plan is converted into standardized voice commands and pushed to the corresponding workers via walkie-talkie network partitions. Real-time dispatch feedback data, including command reception confirmation and execution progress, is obtained, including: Based on thermal imaging cameras, the specific location of miners can be determined by detecting human body temperature characteristics and recognizing human body contours. The vehicle scheduling time, personnel division of labor and cargo transfer route in the transportation execution plan are converted into voice command text according to the standard voice specification for mine operations, and standardized voice commands are generated based on the voice command text. Based on the specific location of the mine workers, the walkie-talkie network is queried, the walkie-talkie channel and device address corresponding to the area where each worker is located are matched, and the standardized voice commands are pushed to the target workers in the partition. Monitor the voice confirmation signals and execution status reports returned by each walkie-talkie terminal, and obtain scheduling feedback data including instruction reception confirmation and execution progress in real time.

8. An intelligent transportation system for mines, characterized in that, The intelligent mine transportation method for implementing any one of claims 1-7 includes: The denoising module is used to perform time-series filtering and gradient denoising on the temperature data of mine transportation nodes acquired by the thermal imaging camera to obtain the temperature feature matrix. The identification module is used to identify transportation anomaly information based on the temperature feature matrix; the matching module is used to perform temperature pattern matching based on the transportation anomaly information and generate a transportation operation instruction set. The planning module is used to plan the temperature safety path of the mine passage as a constraint condition using the transportation operation instruction set, so as to obtain the transportation execution plan; The push module is used to locate the specific location of the mine workers in real time using a thermal imaging camera, convert the transportation execution plan into standardized voice commands, and push them to the corresponding workers through the walkie-talkie network. It also obtains scheduling feedback data in real time, including command reception confirmation and execution progress.

Images