Mine post personnel behavior identification method and system based on artificial intelligence

Abstract

The invention provides a mine post personnel behavior identification method and system based on artificial intelligence, and the method comprises the steps: constructing a mine post dynamic behavior logic reference model, combing post personnel standard operation behavior and process node association logic, and dynamically adjusting the weight. Acquiring personnel behavior acquisition data, and performing adaptive processing on the data and the model to obtain to-be-identified behavior data; and calling a pre-trained AI behavior difference adaptation model to carry out bidirectional feature adaptation, and generating dynamically optimized behavior feature adaptation parameters. And based on the parameters, performing enhancement processing on the to-be-identified behavior data, mining hidden logic association, and generating an enhanced behavior difference feature set. And performing compliance verification according to a preset rule base to obtain a compliance verification result. And finally, generating a post behavior control instruction according to the result and sending the post behavior control instruction to a mine operation control terminal. According to the invention, behaviors of mine post personnel can be accurately identified, and operation safety and high efficiency are guaranteed.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence, and more specifically, to a method and system for recognizing the behavior of personnel in mining positions based on artificial intelligence. Background Technology

[0002] In mining operations, the safety and standardization of personnel behavior play a crucial role in ensuring mine production safety and improving efficiency. Traditional methods of personnel behavior identification mainly rely on manual inspections or simple video surveillance. Manual inspections are not only costly in terms of manpower and resources, but also difficult to achieve real-time and comprehensive monitoring, and are prone to missed detections and misjudgments. While simple video surveillance can record personnel behavior, it lacks intelligent analysis capabilities and cannot automatically identify complex behavioral patterns and potential safety risks.

[0003] With the development of artificial intelligence (AI) technology, some AI-based behavior recognition methods have been gradually applied in the mining industry. However, most of these methods are based solely on static behavioral rules and simple feature matching, failing to fully consider the dynamic changes in the work processes of various positions in the mine and the complex logical relationships between personnel behavior and work process nodes. Furthermore, they lack in-depth analysis and logical mining of behavioral differences, making it difficult to accurately determine whether personnel behavior meets job requirements and thus unable to provide effective decision support for mine operation management. Therefore, a more intelligent and comprehensive method for identifying personnel behavior in mining positions is urgently needed. Summary of the Invention

[0004] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide an artificial intelligence-based method for recognizing the behavior of personnel in mining positions, the method comprising:

[0005] A dynamic behavioral logic benchmark model for mining positions is constructed. Based on the operational process requirements and personnel behavior norms of each position in the mine, the correlation logic between the standard operating behavior of personnel and the process nodes is sorted out. The weight of the behavioral logic dimension is dynamically adjusted in combination with the historical behavior identification results, resulting in a dynamic behavioral logic benchmark model for mining positions that includes dynamic behavioral benchmarks and correlation constraints.

[0006] The personnel behavior data of the mining work area is collected. The personnel behavior data includes continuous behavior records of personnel in different positions during the operation process. The personnel behavior data is matched with the dynamic behavior logic benchmark model of the mining position for behavior difference adaptation processing. The behavioral data to be identified is filtered out that differs from the dynamic behavior benchmark and meets the association constraint conditions.

[0007] The pre-trained AI behavior difference adaptation model is invoked to perform bidirectional feature adaptation on the behavior data to be identified and the dynamic behavior logic benchmark model of the mine position. The AI ​​behavior difference adaptation model dynamically generates feature processing sub-paths according to the associated constraints. The dynamic behavior logic benchmark model of the mine position updates the behavior logic dimension weights according to the feature adaptation feedback output by the AI ​​behavior difference adaptation model, thereby obtaining dynamically optimized behavior feature adaptation parameters.

[0008] Based on the behavioral feature adaptation parameters, the behavioral difference feature enhancement processing is performed on the behavioral difference data to be identified, integrating the behavioral difference information in the behavioral difference data to be identified with the logical association of the dynamic behavioral benchmark, mining the implicit logical association between behavioral differences, and generating a set of enhanced behavioral difference features with job behavior logic consistency.

[0009] Based on the preset rule library for compliance verification of mine job behavior, the set of enhanced behavioral differences is verified for compliance. The set of enhanced behavioral differences is judged to meet the job behavior requirements by referring to the associated constraints in the dynamic behavior logic benchmark model of the mine job. The compliance verification result of the mine job personnel behavior is obtained. The job behavior control instruction is generated according to the compliance verification result of the mine job personnel behavior and sent to the mine operation control terminal.

[0010] In another aspect, embodiments of the present invention also provide an artificial intelligence-based mine worker behavior recognition system, including a processor and a machine-readable storage medium connected to the processor. The machine-readable storage medium is used to store programs, instructions, or code, and the processor is used to execute the programs, instructions, or code in the machine-readable storage medium to implement the above-described method.

[0011] Based on the above, this embodiment of the invention constructs a dynamic behavioral logic benchmark model for mining positions, fully integrates the operational process requirements, personnel behavior norms, and historical behavior identification results of each mining position, and dynamically adjusts the weights of the behavioral logic dimensions. This enables the model to accurately reflect the standard behavioral logic of different positions in different operational scenarios. After acquiring personnel behavior data, behavioral difference adaptation processing is performed between the model and the dynamic behavioral logic benchmark model to filter out behavioral data with research value, thereby improving the efficiency and relevance of data processing. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of the execution flow of the artificial intelligence-based method for recognizing the behavior of mine workers provided in an embodiment of the present invention.

[0013] Figure 2This is a schematic diagram of exemplary hardware and software components of an artificial intelligence-based mine worker behavior recognition system provided in an embodiment of the present invention. Detailed Implementation

[0014] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a flowchart illustrating an artificial intelligence-based method for recognizing the behavior of mine workers, provided in one embodiment of the present invention. The following is a detailed description of this artificial intelligence-based method for recognizing the behavior of mine workers.

[0015] Step S110: Construct a dynamic behavioral logic benchmark model for mine positions. Based on the operational process requirements and personnel behavior norms of each position in the mine, sort out the correlation logic between the standard operating behavior of personnel and the process nodes, and dynamically adjust the weight of the behavioral logic dimension in combination with the historical behavior identification results to obtain a dynamic behavioral logic benchmark model for mine positions that includes dynamic behavioral benchmarks and correlation constraints.

[0016] In this embodiment, the operational process requirements for coal mining positions are first clarified, covering core processes such as coal cutting, support, and frame relocation, as well as the corresponding personnel behavior norms for each process. Standard operating procedures for each process node are then analyzed, such as equipment inspection before coal cutting and component installation during support, establishing a logical association between these procedures and process nodes. Next, historical results of personnel behavior identification in coal mining positions are combined to analyze the impact of temporal logic (order of behavior execution) and collaborative logic (personnel and equipment coordination) on identification accuracy, adjusting the weights of these two dimensions. Finally, the associated logic and adjusted weights are integrated, and associated constraints such as the order of behavior execution (e.g., warning before equipment activation) and equipment usage (e.g., using designated equipment for support) are added to form a dynamic behavioral logic benchmark model for coal mining positions.

[0017] Step S111: Extract the key process nodes from the work process requirements of each position in the mine. Each key process node corresponds to a set of standard operating behaviors of the personnel in the position. Associate and bind the key process nodes with the corresponding standard operating behaviors of the personnel in the position to generate a process-behavior association table.

[0018] From the coal mining workflow, key process nodes that play a crucial role in safety and efficiency are identified, such as pre-start inspections, setting up warning lines, starting the coal mining machine, and installing support components. Each key process node is matched with a corresponding set of standard operating procedures (SOPs). For example, the pre-start inspection node corresponds to checking the equipment power supply, cutting teeth, and cooling system. Each node and action is associated with a unique identifier, and a process-action association table is created to clearly present the correspondence between nodes and actions.

[0019] Step S112: Collect the behavioral norms documents of personnel in various positions in the mine, analyze the constraints on the work behavior of personnel in the behavioral norms documents of personnel in various positions in the mine, the constraints include the behavior execution sequence constraints and the behavior-related equipment constraints, match the constraints with the standard work behavior of personnel in the process-behavior association table, and generate the constraint-behavior correspondence relationship.

[0020] For example, collect documents such as safety regulations and equipment operation specifications for coal mining positions, and extract behavioral constraints from them. Constraints on the order of behavior execution include "complete equipment inspection before starting the coal mining machine," and constraints on the equipment associated with the behavior include "starting the coal mining machine requires a specified model of equipment." These constraints are then matched with standard operating procedures in the process-behavior relationship table to determine the specific behavior corresponding to each constraint. For instance, the constraint "inspect before starting" corresponds to the pre-start inspection and starting of the coal mining machine, generating a constraint-behavior correspondence and clarifying the constraint requirements that the behavior must follow.

[0021] Step S113: Determine the core descriptive attributes of the standard operating procedures of the personnel in the position. The core descriptive attributes include the performance period attribute and the associated object attribute. Extract the core descriptive attribute values ​​of the standard operating procedures of the personnel in the position under different key process nodes.

[0022] The core descriptive attributes for standard operating procedures in coal mining positions are defined as the execution time period (e.g., the process stage and time window of the action) and the associated object attributes (e.g., the equipment, tools, and work area involved). For each key process node, attribute values ​​are extracted. For example, the execution time for warning actions is set to before coal cutting and before equipment startup, and the associated objects are warning signs, warning tapes, and the work roadway entrance area. The attribute values ​​are then linked to and recorded with the actions.

[0023] Step S114: Based on the process-behavior association table, constraint-behavior correspondence and core descriptive attribute values, sort out the standard operating behavior of the personnel and the execution logic of key process nodes. The execution logic includes behavior-process sequence logic and behavior-equipment collaboration logic to form an initial behavior benchmark.

[0024] By combining the process-behavior association table, constraint-behavior correspondence, and core descriptive attribute values, the execution logic is clarified. The behavior-process sequence logic clarifies the execution order of process nodes and behaviors, such as performing pre-start checks and setting warnings before starting the coal mining machine; the behavior-equipment coordination logic clarifies the equipment status during behavior execution, such as requiring the equipment to be powered off and the cooling system to be normal before starting the coal mining machine. Integrating these two logics forms the initial behavioral benchmark for the coal mining position.

[0025] Step S115: Collect historical mine personnel behavior recognition results, extract the influence coefficients of each behavioral logic dimension on the recognition accuracy in the historical mine personnel behavior recognition results. The behavioral logic dimensions include temporal logic dimension and collaborative logic dimension. Adjust the weight values ​​of each behavioral logic dimension according to the influence coefficients.

[0026] Collect past records of personnel behavior identification in coal mining positions, and remove invalid records such as those related to equipment malfunctions or incorrect information. Analyze the impact of temporal logic and collaborative logic dimensions on identification accuracy, and extract the impact coefficients. If the impact coefficient of the temporal logic dimension is large, it indicates that it is more critical to identification accuracy, so the weight of this dimension is increased; conversely, it is decreased, thus completing the weight adjustment.

[0027] Step S116: Integrate the weighted execution logic with the initial behavioral benchmark, supplement the applicable job scope identifiers and logic activation conditions, and combine the constraint-behavior correspondence to form a dynamic behavioral logic benchmark model for mine positions that includes dynamic behavioral benchmarks and associated constraint conditions.

[0028] The adjusted execution logic is integrated into the initial behavioral benchmark, supplemented with job scope identification (clearly applicable to coal mining positions) and logic activation conditions (such as normal operating environment). Combining constraint-behavior correspondences, associated constraints are incorporated to form a complete dynamic behavioral logic benchmark model for mining positions, including dynamic behavioral benchmarks and associated constraints.

[0029] Step S117: Perform a logical self-consistency check on the dynamic behavior logic benchmark model of the mine positions, and determine whether there are conflicts between the dynamic behavior benchmarks of different positions and between the dynamic behavior benchmarks and related constraints. If there are conflicts, adjust the content of the dynamic behavior logic benchmark model of the mine positions according to the priority of the job operation process requirements and the accuracy of historical identification.

[0030] Check whether the benchmarks for coal mining positions conflict with those for other positions (such as tunneling positions), and whether the benchmarks for coal mining positions contradict their own associated constraints. If conflicts exist, adjust the model content based on the priority of the job's work processes (such as prioritizing safety-related processes) and the accuracy of historical identification to ensure that the model logic is consistent.

[0031] Step S120: Obtain personnel behavior data in the mining work area. The personnel behavior data includes continuous behavior records of personnel in different positions during the work process. Perform behavior difference adaptation processing on the personnel behavior data and the dynamic behavior logic benchmark model of the mining position, and filter out the behavior data to be identified that differs from the dynamic behavior benchmark and meets the association constraint conditions.

[0032] Raw behavioral data of personnel is acquired through data collection equipment in the coal mining work area, including behavioral actions and scene information. Based on the behavioral execution time period attributes in the dynamic behavioral logic benchmark model of the mine position, the raw data is divided into multiple continuous behavioral records, forming personnel behavior collection data. The collected data is compared with the dynamic behavioral benchmark to identify records with differences in behavioral execution order, etc., and then matched with associated constraint conditions to filter out the discrepancies that meet the constraints, forming the behavioral data to be identified.

[0033] Step S121: Receive raw behavioral data transmitted by behavior acquisition devices deployed in the mining work area. The raw behavioral data includes records of the behavior actions of the personnel during the work process and information on the scene where the behavior occurs. Each behavior acquisition device corresponds to a specific work area.

[0034] Data acquisition devices are deployed at key locations such as the coal mining face and equipment operating area, with each device corresponding to a specific area. The devices collect real-time data on personnel actions (such as checking equipment and pressing start buttons) and scene information (such as lighting conditions in the work area and equipment status), and transmit the raw behavioral data to the data processing terminal to ensure data integrity and real-time performance.

[0035] Step S122: Divide the original behavioral data into behavioral segments. Based on the behavioral execution time period attribute in the dynamic behavioral logic benchmark model of mine positions, divide the original behavioral data into multiple continuous personnel behavior records. Each personnel behavior record corresponds to a continuous work behavior time period.

[0036] The dynamic behavioral logic baseline model of mining positions is used to extract the behavioral execution time period attributes, clarifying the time period characteristics of different behaviors (such as the pre-start inspection corresponding to the pre-operation preparation stage). Based on these characteristics, the start and end points of behaviors in the raw data are identified, and the data is divided into multiple continuous behavior records, with each record corresponding to a complete work behavior time period.

[0037] Step S123: Store all the personnel behavior records after division according to the corresponding job operation area to form personnel behavior collection data containing personnel behavior records of different job operation areas.

[0038] Based on the identification of the data collection equipment, determine the corresponding work area (such as coal mining face, equipment operation area) for each behavior record, establish a classified storage structure, set up storage directories by area, classify and store the behavior records to form personnel behavior collection data.

[0039] Step S124: Extract the job scope identifier and dynamic behavior benchmark from the dynamic behavior logic benchmark model of the mine job, and determine the target job and target dynamic behavior benchmark that need to be processed.

[0040] Extract job scope identifiers (such as "coal mining job") from the dynamic behavior logic benchmark model of mining positions, determine the target job as coal mining job by combining the identification task requirements, and then extract the target dynamic behavior benchmark corresponding to the job.

[0041] Step S125: Match the personnel behavior records in the personnel behavior collection data with the target positions, and filter out the job matching behavior records whose job identifiers are consistent with the target positions.

[0042] Examine the job identifier of each record in the personnel behavior data collection, compare it with the identifier of the target job (coal mining job), and filter out the records with the same identifier, that is, the job matching behavior records.

[0043] Step S126: Compare the job matching behavior records with the target dynamic behavior benchmark to determine whether the execution order of the behaviors in the job matching behavior records is consistent with the behavior-process timing logic in the target dynamic behavior benchmark, and filter out the timing difference behavior records.

[0044] The behavior-process timing logic (such as checking first, then warning, and then starting the equipment) is obtained from the target dynamic behavior benchmark. The actual execution order of the job matching behavior record is analyzed and compared with the benchmark timing logic to filter out the timing difference behavior records with inconsistent execution order.

[0045] Step S127: Match the time-series difference behavior records with the association constraints in the dynamic behavior logic benchmark model of mine positions, determine whether the behavior-related equipment in the time-series difference behavior records meets the behavior-related equipment constraints in the association constraints, and filter out the time-series difference behavior records that meet the association constraints.

[0046] Extract behavioral-related equipment constraints (such as using a specified model to start a coal mining machine) from the dynamic behavioral logic benchmark model of mining positions, analyze the actual associated equipment in the time-series difference behavior records, match them with the constraints, and filter out the time-series difference behavior records that meet the constraints.

[0047] Step S128: Perform a completeness check on the time-series difference behavior records that meet the association constraint conditions, and determine whether each time-series difference behavior record contains all the core descriptive attribute values ​​required by the dynamic behavior logic benchmark model of the mine position. If there are any missing values, supplement the corresponding core descriptive attribute values ​​to form the behavior data to be identified.

[0048] Examine the time-series difference behavior records that meet the constraints to confirm whether they contain all core descriptive attribute values ​​such as the behavior execution period and associated objects. If missing, supplement the records by querying supplementary data and combining them with job logs to ensure completeness and form the behavior data to be identified.

[0049] Step S130: Call the pre-trained AI behavior difference adaptation model to perform bidirectional feature adaptation between the behavior data to be identified and the dynamic behavior logic benchmark model of the mine position. The AI ​​behavior difference adaptation model dynamically generates feature processing sub-paths according to the associated constraints. The dynamic behavior logic benchmark model of the mine position updates the weights of the behavior logic dimension according to the feature adaptation feedback output by the AI ​​behavior difference adaptation model, and obtains dynamically optimized behavior feature adaptation parameters.

[0050] A pre-trained AI behavior difference adaptation model is loaded, which includes modules such as a feature input layer and a logic parsing layer. The behavior data to be identified is input into the model to extract behavioral features. The dynamic behavior logic benchmark model for mining positions is input into the model to parse constraints and the dynamic benchmark. The AI ​​model generates feature processing sub-paths for different features based on the constraints, processes features through these sub-paths, calculates the fit degree, and generates feature adaptation feedback. The dynamic behavior logic benchmark model for mining positions adjusts the weights of the temporal and collaborative logic dimensions based on the feedback. The AI ​​model combines the adjusted weights to optimize the fusion parameters, generating behavioral feature adaptation parameters that include channel weights, dimension adaptation, and sub-path selection coefficients. These parameters are then validated and the final parameters are determined.

[0051] Step S131: Load the pre-trained AI behavior difference adaptation model. The AI ​​behavior difference adaptation model includes a feature input layer, a logic parsing layer, a dynamic sub-path generation layer, a parameter optimization layer, and a feedback output layer. Each layer achieves data interaction through a feature transfer channel.

[0052] Load a pre-trained AI behavior difference adaptation model for mining scenarios. The model includes a feature input layer (processing input data), a logic parsing layer (parsing the baseline model), a dynamic sub-path generation layer (generating processing paths), a parameter optimization layer (optimizing parameters), and a feedback output layer (transmitting feedback). Each layer interacts with data through an encrypted feature transmission channel to ensure data stability and security.

[0053] Step S132: Input the behavior data to be identified into the feature input layer of the AI ​​behavior difference adaptation model, perform initial feature extraction on each time-series difference behavior record in the behavior data to be identified, and obtain the initial feature vector of each time-series difference behavior record.

[0054] The behavioral data to be identified is input into the feature input layer according to the format. The data is first cleaned to remove interference information, and then features such as behavioral actions, time, and associated objects are extracted. Non-numerical features are converted into codes, and numerical features are normalized. They are combined in sequence to form the initial feature vector of each time-series difference behavioral record. The vector contains multi-dimensional feature values.

[0055] Step S133: Input the dynamic behavior logic benchmark model of the mining position into the logic parsing layer of the AI ​​behavior difference adaptation model, parse the associated constraints, dynamic behavior benchmarks and behavior logic dimension weights in the dynamic behavior logic benchmark model of the mining position, and determine the feature influence factor corresponding to each associated constraint.

[0056] The dynamic behavioral logic benchmark model for mining positions is input into the logic parsing layer. Within this layer, the model data is split, and associated constraints, dynamic behavioral benchmarks, and behavioral logic dimension weights are extracted. The impact of each constraint on different features is analyzed to determine feature influence factors. For example, execution order constraints have a large impact factor on time features, and equipment constraints have a large impact factor on equipment features. A constraint-feature influence factor comparison table is then created.

[0057] Step S134: Input the initial feature vector of behavior and the feature influence factor corresponding to each associated constraint into the dynamic sub-path generation layer of the AI ​​behavior difference adaptation model. Based on the feature influence factor corresponding to each associated constraint, dynamically generate a feature processing sub-path for each initial feature vector of behavior. The feature processing sub-path includes an associated constraint strengthening channel and an unassociated constraint filtering channel.

[0058] The dynamic sub-path generation layer receives the initial feature vector of the behavior and a constraint-feature influence factor lookup table, and divides the features into related features that are related to the constraints and unrelated features that are not. Related features are assigned a related constraint enhancement channel (to amplify difference information), and unrelated features are assigned an unrelated constraint filtering channel (to reduce noise interference), combining these to form the feature processing sub-path for each vector.

[0059] Step S135: The initial feature vector of the behavior is processed in a targeted manner through the feature processing sub-path, which strengthens the feature signals related to the associated constraints, filters the feature noise unrelated to the associated constraints, and generates the constraint adaptation feature vector for each time-series difference behavior record.

[0060] Following the sub-path order, associated features are passed to the enhancement channel to amplify the difference signal, while non-associated features are passed to the filtering channel to reduce noise. After processing, the two types of features are combined in their original dimensional order to generate a constraint-adaptive feature vector for each time-series difference behavior record, highlighting key difference information.

[0061] Step S136: Input the constraint adaptation feature vector into the parameter optimization layer of the AI ​​behavior difference adaptation model, calculate the degree of adaptation between each constraint adaptation feature vector and the dynamic behavior benchmark in the dynamic behavior logic benchmark model of the mine position, and obtain feature adaptation feedback.

[0062] The parameter optimization layer extracts the baseline feature vector of the dynamic behavior benchmark (with the same dimension and type as the constraint adaptation vector). It then uses a similarity calculation method, combining the current behavior logic dimension weights to weight the feature similarity of each dimension, and comprehensively calculates the overall fit. Feedback is generated based on the fit score, including the fit score value, descriptions of low-fit dimensions and differences, and dimension contribution.

[0063] Step S137: The feature adaptation feedback is transmitted to the weight update module of the dynamic behavior logic benchmark model of the mine position through the feedback output layer of the AI ​​behavior difference adaptation model. The weight update module adjusts the weight values ​​of each behavior logic dimension according to the feature adaptation feedback. The weight values ​​of behavior logic dimensions with high adaptation are increased, and the weight values ​​of behavior logic dimensions with low adaptation are decreased.

[0064] The feedback output layer converts and encrypts the feedback format before transmitting it to the weight update module. After decryption and verification, the module extracts the fit and dimension contribution information. If a dimension has a high fit and a large contribution, its weight is increased; otherwise, it is decreased. The adjustment range is controlled within a preset range. After the update, the weight is stored and logged.

[0065] Step S138: The parameter optimization layer of the AI ​​behavior difference adaptation model combines the updated behavior logic dimension weights to optimize the fusion parameters of the constraint adaptation feature vector, generating behavior feature adaptation parameters that include channel weight coefficients, dimension adaptation coefficients, and sub-path selection coefficients.

[0066] The parameter optimization layer obtains the updated weights, analyzes the importance of feature dimensions, and optimizes the fusion parameters. High-channel weight coefficients are assigned to important feature channels, the dimension adaptation coefficients are adjusted to adapt to the vector characteristics, the priority of sub-path selection coefficients is adjusted according to the weights, and the three types of coefficients are integrated to form behavioral feature adaptation parameters.

[0067] Step S139: Verify the rationality of the generated behavioral feature adaptation parameters, determine whether the channel weight coefficient allocation matches the updated behavioral logic dimension weight, and whether the dimension adaptation coefficient conforms to the dimensional characteristics of the initial behavioral feature vector. If the verification fails, readjust the processing parameters of the dynamic sub-path generation layer until the required behavioral feature adaptation parameters are generated.

[0068] Verify whether the channel weight coefficient matches the weight (e.g., a higher time-series weight corresponds to a higher time-series channel coefficient) and whether the dimension adaptation coefficient conforms to the characteristics of vector dimensions (e.g., the dimension adaptation coefficient is reasonable within the numerical set). If it fails, the parameter optimization layer instructs the dynamic sub-path generation layer to adjust the processing parameters (e.g., threshold, calculation model), regenerate the parameters, and verify them until they pass.

[0069] Step S140: Based on the behavioral feature adaptation parameters, perform behavioral difference feature enhancement processing on the behavioral data to be identified, integrate the behavioral difference information in the behavioral data to be identified with the logical relationship of dynamic behavioral benchmark, mine the implicit logical relationship between behavioral differences, and generate a set of enhanced behavioral difference features with job behavior logic consistency.

[0070] Three types of coefficients are extracted from the behavioral feature adaptation parameters to determine the processing rules. For each time-series difference behavior record, a sub-path is selected based on the coefficients chosen according to the sub-path. The vector dimension is adjusted to a uniform level based on the dimension adaptation coefficients, and key features are highlighted by weighting the vectors with channel weight coefficients. The weighted vectors are associated with the dynamic behavioral baseline to extract behavioral difference information and establish logical connections. Using the implicit logic mining module of the AI ​​model, combined with the inherent logic of the work process, potential connections between differences (such as causality or the same error) are analyzed. Redundant information is removed, key difference information is combined according to the work logic order, source identifiers are added, and logical consistency is verified to form a set of enhanced behavioral difference features.

[0071] Step S141: Extract the channel weight coefficient, dimension adaptation coefficient and sub-path selection coefficient from the behavior feature adaptation parameters, and determine the weight allocation ratio, dimension adjustment standard and feature processing sub-path selection basis of the initial behavior feature vector of each temporal difference behavior record in the behavior data to be identified in the reinforcement processing.

[0072] The three types of coefficients are separated from the behavioral feature adaptation parameters. The channel weight ratio of each feature channel is determined according to the channel weight coefficient. The vector dimension adjustment standard is set according to the dimension adaptation coefficient. The sub-path selection priority is determined by the sub-path selection coefficient.

[0073] Step S142: Select the corresponding feature processing sub-path based on the initial feature vector of each temporal difference behavior record in the behavior data to be identified according to the sub-path selection coefficient. The feature processing sub-path with the higher sub-path selection coefficient is selected first.

[0074] Define the type of feature processing sub-path (temporal constraint sub-path, device constraint sub-path, and hybrid constraint sub-path) and its corresponding sub-path selection coefficient. Iterate through the initial feature vector of each temporal difference behavior record in the behavior data to be identified, analyzing the significance of the core behavior difference attributes corresponding to the vector—the behavior execution order difference attribute and the behavior-related device difference attribute. If the execution order difference is more significant, prioritize matching the temporal constraint sub-path; if the device difference is more significant, prioritize matching the device constraint sub-path; if both are significant, prioritize matching the hybrid constraint sub-path. Then, combine the selection coefficient values ​​of each sub-path. If the coefficient of the preferred sub-path is higher than a preset threshold, it is directly selected; if it is lower than the threshold, select the other suitable sub-path with the highest coefficient, completing the sub-path selection for each vector.

[0075] Step S143: Adjust the dimensions of the initial feature vectors of the selected feature processing sub-paths according to the dimension adaptation coefficient, and adjust each initial feature vector of the behavior to an adapted feature vector of the same dimension. During the adjustment process, retain the behavior difference information identifier of each time-series difference behavior record.

[0076] Based on a unified standard set by the dimension adaptation coefficient, the initial feature vectors of the selected sub-paths are adjusted in terms of dimensions. If the vector dimensions exceed the standard, irrelevant feature dimensions are removed; if the dimensions are insufficient, necessary feature dimensions are added. During adjustment, the behavioral difference information identifiers (such as difference type and location) corresponding to each vector are retained, ultimately forming a dimensionally unified adapted feature vector.

[0077] Step S144: Weight each adaptation feature vector according to the channel weight coefficient, calculate the weight value of each feature channel in each adaptation feature vector to obtain the weighted feature vector, and record the contribution of behavioral difference information corresponding to each feature channel.

[0078] According to the proportion of channel weight coefficients allocated, the eigenvalues ​​of each channel in the adaptation feature vector are combined with their corresponding coefficients to calculate the weighted value. All channel weighted values ​​are integrated to form a weighted feature vector, and the contribution of each channel to the behavioral difference information is recorded (e.g., if the difference is more significant after weighting a certain channel, its contribution is high).

[0079] Step S145: Logically integrate the weighted feature vector with the dynamic behavior benchmark in the dynamic behavior logic benchmark model of the mine position, extract the behavior difference feature information related to the dynamic behavior benchmark from the weighted feature vector, and perform cross-correlation processing on the extracted behavior difference feature information by combining the behavior-process timing logic and behavior-equipment collaboration logic in the dynamic behavior benchmark.

[0080] The weighted feature vector is compared with the dynamic behavior benchmark to filter out behavioral difference features that do not conform to the benchmark (such as execution order and equipment usage differences). Combining the temporal logic (process sequence) and collaborative logic (equipment and behavior matching) in the benchmark, the correlation between the difference information is analyzed, such as whether a certain equipment difference is related to the difference in the sequence of subsequent processes. Logically related difference information is bound to form a correlation group.

[0081] Step S146: Using the implicit logic mining module of the AI ​​behavior difference adaptation model, based on the inherent logic of the mine job operation process, analyze the potential logical relationships between the behavioral difference feature information after cross-correlation processing, and supplement the implicit logical connections between behavioral differences.

[0082] The system loads the inherent logical knowledge base of the mine's operational processes (including process dependencies and equipment coordination rules), and inputs the cross-correlated difference feature information into the implicit logic mining module. The module then analyzes the potential relationships between differences using the knowledge base, such as whether a difference is the cause of another difference (e.g., failure to check equipment leading to abnormal parameter adjustments), or whether multiple differences originate from the same operational error. The mined implicit relationships are then added to the difference feature information.

[0083] Step S147: Redundancy removal is performed on the behavioral difference feature information after supplementing implicit logical associations. Key behavioral difference feature information that is directly related to the job behavior logic is retained, and duplicate feature information or feature information that does not contribute to the identification of job behavior is removed.

[0084] Examine and supplement the implicit correlations of the discrepancy information one by one to determine whether it is related to the core behavioral logic of the coal mining position (process execution, equipment operation) and is useful for compliance identification. Remove duplicate records (such as multiple annotations of the same difference) and irrelevant information (such as minor differences in actions unrelated to safety), and retain key difference information.

[0085] Step S148: Combine the key behavioral difference feature information according to a unified feature arrangement order, and generate feature source identifiers by combining the contribution of behavioral difference information corresponding to each feature channel, forming a set of enhanced behavioral difference features with job behavior logic consistency and source traceability.

[0086] Key difference feature information is combined according to a preset order (such as the sequence of procedures, importance of differences), and a source identifier (including the channel to which it belongs, contribution level, and original record number) is added to each feature to ensure traceability of the source. After combination, a set of enhanced behavioral difference features is formed, which not only conforms to the job logic but also allows for the traceability of feature sources.

[0087] Step S149: Perform a logical consistency check on the generated set of reinforced behavior difference features to determine whether the logical relationship of each feature information in the set of reinforced behavior difference features is consistent with the dynamic behavior benchmark in the dynamic behavior logical benchmark model of the mine position. If there is any inconsistency, correct the logical expression of the feature information according to the dynamic behavior benchmark to ensure the logical consistency of the set of reinforced behavior difference features.

[0088] The set of enhanced behavioral difference features is compared with the logical rules (time sequence, collaborative logic) of the dynamic behavioral benchmark. If the logical expression of the difference information in the set conflicts with the benchmark (such as contradiction in the process sequence), the expression is corrected according to the benchmark to ensure that the logic of the set is consistent with the benchmark and there is no contradiction.

[0089] Step S150: Based on the preset rule library for compliance verification of mine job behavior, perform compliance verification on the set of enhanced behavioral difference features, and determine whether the set of enhanced behavioral difference features meets the job behavior requirements by referring to the associated constraint conditions in the dynamic behavior logic benchmark model of mine job, so as to obtain the compliance verification result of mine job personnel behavior, generate job behavior control instructions based on the compliance verification result of mine job personnel behavior and send them to the mine operation control terminal.

[0090] The system accesses the mine job compliance verification rule library to extract compliance standards and verification conditions (such as safe operation and equipment specifications) corresponding to coal mining positions. It compares the enhanced set of behavioral difference characteristics with the verification conditions, marking compliant and non-compliant differences, and determines the severity of non-compliance (general / severe) based on associated constraints. The results are summarized to determine the compliance level, a verification report is generated, and control instructions (warning / intervention) are generated based on the report, encrypted, and sent to the mine operation control terminal to ensure instruction reception and confirmation.

[0091] Step S151: Call the preset mine job behavior compliance verification rule library, parse the job behavior compliance standards contained in the mine job behavior compliance verification rule library, each job behavior compliance standard corresponds to a set of compliance verification conditions, and the compliance verification conditions are formulated based on the mine job operation safety specifications and operation process requirements.

[0092] The rule base call program is started, the contents of the library are parsed, the compliance standards for different positions are distinguished, the compliance standards for coal mining positions are extracted, and the compliance verification conditions corresponding to each standard are broken down. These conditions are all formulated based on mine safety regulations and coal mining operation procedures (such as "inspection is required before starting the equipment").

[0093] Step S152: Extract the feature information and feature source identifier from the enhanced behavior difference feature set, determine the target position corresponding to the enhanced behavior difference feature set, and retrieve the job behavior compliance standards and corresponding compliance verification conditions corresponding to the target position from the mine job behavior compliance verification rule library.

[0094] Job information is obtained from the source identifiers of the enhanced behavioral difference feature set, identifying the target job as a coal mining position. The compliance standards and verification conditions corresponding to the coal mining position are retrieved from the rule base to ensure a match with the target job.

[0095] Step S153: Match each behavioral difference feature in the enhanced behavioral difference feature set with the corresponding compliance verification conditions one by one, and determine whether each behavioral difference feature meets the corresponding compliance verification conditions. If it does, mark it as a compliant difference feature; otherwise, mark it as a non-compliant difference feature.

[0096] Iterate through each difference feature in the set of enhanced behavior difference features, and match each feature with the corresponding compliance verification condition (e.g., "cooling system not checked" matches "cooling system needs to be checked before startup"). Compare the actual situation of the feature with the condition requirements, and mark the difference feature as compliant or non-compliant.

[0097] Step S154: Record the matching results for each behavioral difference feature and generate a preliminary verification record containing feature identifier, matching result identifier, and description of non-compliance reasons. The matching result identifier includes two types: compliant and non-compliant.

[0098] Each discrepancy feature is assigned a unique identifier, and preliminary verification records are generated based on the matching results (compliant / non-compliant). Non-compliant records must describe the reason (e.g., "cooling system not checked, violating pre-start inspection conditions"), while compliant records are simplified to "meets verification conditions," clearly presenting the verification status of each feature.

[0099] Step S155: Referring to the associated constraints in the dynamic behavior logic benchmark model of mine positions, further analyze the non-compliant difference features in the preliminary verification records, and determine whether the non-compliant difference features violate the behavior execution order constraints or behavior-related equipment constraints in the associated constraints. If they violate them, they are marked as serious non-compliant difference features; if they do not violate them, they are marked as general non-compliant difference features.

[0100] Extract the associated constraints (execution order, equipment constraints) from the dynamic behavioral logic baseline model of mining positions, and compare the non-compliance differences with the constraints. If a constraint is violated (such as starting equipment directly without warning), it is marked as a serious non-compliance; if only ordinary conditions are violated (such as omitting non-critical checks), it is marked as a general non-compliance, thus achieving non-compliance classification.

[0101] Step S156: Summarize and statistically analyze the preliminary verification records, and calculate the proportion of compliant difference features, the proportion of non-compliant difference features, the number of serious non-compliant difference features, and the number of general non-compliant difference features in the set of enhanced behavioral difference features.

[0102] The system calculates the total number of compliant and non-compliant differences, the compliance percentage (number of compliant items / total number of items), and the non-compliance percentage (number of non-compliant items / total number of items). It also calculates the specific number of serious and general non-compliance items and generates a statistical report to present the overall compliance situation in a visual way.

[0103] Step S157: Based on the statistical results and the preset compliance judgment criteria, determine the compliance level of the mine personnel's behavior corresponding to the set of enhanced behavioral difference characteristics. The compliance level is used to indicate the degree to which the mine personnel's behavior meets the requirements of their job.

[0104] Pre-defined compliance criteria are established (e.g., "high compliance rate with no serious non-compliance is excellent, medium compliance rate with few serious non-compliance is good, low compliance rate or many serious non-compliance is unqualified"). The statistical results are compared with these criteria to determine the compliance level of personnel in coal mining positions, reflecting the degree to which their behavior meets the requirements of their positions.

[0105] Step S158: Integrate the compliance level, the percentage of compliance difference features, the percentage of non-compliance difference features, the details of serious non-compliance difference features, and the details of general non-compliance difference features to generate the compliance verification results of the behavior of personnel in mining positions.

[0106] The compliance level, the proportion of compliance and non-compliance, and the details of serious and general non-compliance (characteristics and reasons) are integrated and organized into a structured verification result document in a fixed format.

[0107] Step S159: Analyze the compliance level in the compliance verification results of the mine personnel's behavior, and determine the corresponding behavior compliance mark based on the compliance level. Behaviors whose compliance level meets the preset compliance standards are compliant marks, and behaviors whose compliance level does not meet the preset compliance standards are non-compliant marks.

[0108] By reviewing the compliance level in the verification results and comparing it with the preset standards (e.g., excellent or good for compliance, and unqualified for non-compliance), the system assigns a "compliant" label to those that meet the standards and a "non-compliant" label to those that do not, thus quickly distinguishing the overall compliance status.

[0109] Step S1510: Call the preset mine job behavior control measures library, retrieve the corresponding control measures according to the behavior compliance mark and non-compliance difference characteristics details. If the behavior compliance mark is a compliance mark, retrieve the routine behavior monitoring measures; if the behavior compliance mark is a non-compliance mark, retrieve the emergency behavior intervention measures according to the serious non-compliance difference characteristics details, and retrieve the behavior warning guidance measures according to the general non-compliance difference characteristics details.

[0110] Call the control measures library and retrieve routine monitoring measures (such as periodic records and routine inspections) according to compliance indicators; retrieve emergency intervention measures (such as suspension of operations and on-site rectification) for serious non-compliance according to non-compliance indicators, and retrieve warning and guidance measures (such as sending warnings and operation guidelines) for general non-compliance.

[0111] Step S1511: Link and integrate the behavior compliance identifier, corresponding control measures, details of non-compliance difference characteristics (if any), and feature source identifiers of the enhanced behavior difference characteristic set to generate basic data for job behavior control.

[0112] Link and bind compliance / non-compliance labels, control measures, non-compliance details (only when non-compliance occurs), and characteristic source labels one by one to ensure that the information corresponds and form the basic data for job behavior control.

[0113] Step S1512: Standardize the format of the basic data for job behavior control. In accordance with the data format that can be recognized by the mine operation control terminal, supplement the data generation time information and job identification information to form job behavior control instructions.

[0114] Adjust the basic data according to the preset format of the mine operation control terminal, supplement the data generation time and job identification (such as "coal mining job - working face"), improve the instruction content, and form a standardized job behavior control instruction.

[0115] Step S1513: Establish an encrypted communication connection with the mine operation control terminal, and send the job behavior control instructions to the corresponding mine operation control terminal through an encrypted transmission protocol.

[0116] Initiate an encrypted communication program to establish an encrypted connection (including key verification) with the target control terminal (such as the coal mining face terminal), and use a dedicated encryption protocol to transmit control commands to prevent data theft or tampering.

[0117] Step S1514: Receive instruction reception confirmation information from the mine operation control terminal. If no instruction reception confirmation information is received within the preset feedback waiting time, resend the job behavior control instruction until instruction reception confirmation information is received from the mine operation control terminal.

[0118] After sending the command, start feedback listening and wait for the terminal to return confirmation. If no confirmation is received within the preset time, resend the command and repeat the process until confirmation is received to ensure successful delivery of the command.

[0119] Step S1515: Link and store the sending records of job behavior control instructions, the receiving confirmation records of the mine operation control terminal, and the compliance verification results of mine personnel behavior to the mine job behavior control log database.

[0120] The command sending time, number of times, terminal confirmation time, and compliance verification results are integrated and stored in the log database.

[0121] Figure 2 The illustration shows exemplary hardware and software components of an AI-based mine worker behavior recognition system 100 that can implement the ideas of this application, according to some embodiments of this application. For example, a processor 120 can be used in the AI-based mine worker behavior recognition system 100 and to perform the functions described in this application.

[0122] The AI-based mine worker behavior recognition system 100 can be a general-purpose server or a special-purpose server; both can be used to implement the AI-based mine worker behavior recognition method of this application. Although only one server is shown in this application, for convenience, the functions described in this application can be implemented in a distributed manner on multiple similar platforms to balance the load.

[0123] For example, the AI-based mine worker behavior recognition system 100 may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and various forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the AI-based mine worker behavior recognition system 100 may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The methods of this application can be implemented according to these program instructions. The AI-based mine worker behavior recognition system 100 also includes an I / O interface 150 between the computer and other input / output devices.

[0124] For ease of explanation, only one processor is described in the AI-based mine worker behavior recognition system 100. However, it should be noted that the AI-based mine worker behavior recognition system 100 of this application may also include multiple processors. Therefore, the steps performed by one processor as described in this application may also be performed jointly or individually by multiple processors. For example, if the processor of the AI-based mine worker behavior recognition system 100 performs steps A and B, it should be understood that steps A and B may also be performed jointly by two different processors or individually by one processor. For example, the first processor performs step A, the second processor performs step B, or the first processor and the second processor jointly perform steps A and B.

[0125] Furthermore, this embodiment of the invention also provides a readable storage medium, wherein computer-executable instructions are preset in the readable storage medium, and when the processor executes the computer-executable instructions, the above-mentioned artificial intelligence-based method for recognizing the behavior of personnel in mining positions is implemented.

[0126] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A method for recognizing the behavior of personnel in mining positions based on artificial intelligence, characterized in that, The method includes: A dynamic behavioral logic benchmark model for mining positions is constructed. Based on the operational process requirements and personnel behavior norms of each position in the mine, the correlation logic between the standard operating behavior of personnel and the process nodes is sorted out. The weight of the behavioral logic dimension is dynamically adjusted in combination with the historical behavior identification results, resulting in a dynamic behavioral logic benchmark model for mining positions that includes dynamic behavioral benchmarks and correlation constraints. The personnel behavior data of the mining work area is collected. The personnel behavior data includes continuous behavior records of personnel in different positions during the operation process. The personnel behavior data is matched with the dynamic behavior logic benchmark model of the mining position for behavior difference adaptation processing. The behavioral data to be identified is filtered out that differs from the dynamic behavior benchmark and meets the association constraint conditions. The pre-trained AI behavior difference adaptation model is invoked to perform bidirectional feature adaptation on the behavior data to be identified and the dynamic behavior logic benchmark model of the mine position. The AI ​​behavior difference adaptation model dynamically generates feature processing sub-paths according to the associated constraints. The dynamic behavior logic benchmark model of the mine position updates the behavior logic dimension weights according to the feature adaptation feedback output by the AI ​​behavior difference adaptation model, thereby obtaining dynamically optimized behavior feature adaptation parameters. Based on the behavioral feature adaptation parameters, the behavioral difference feature enhancement processing is performed on the behavioral difference data to be identified, integrating the behavioral difference information in the behavioral difference data to be identified with the logical association of the dynamic behavioral benchmark, mining the implicit logical association between behavioral differences, and generating a set of enhanced behavioral difference features with job behavior logic consistency. Based on the preset rule library for compliance verification of mine job behavior, the set of enhanced behavioral differences is verified for compliance. The set of enhanced behavioral differences is judged to meet the job behavior requirements by referring to the associated constraints in the dynamic behavior logic benchmark model of the mine job. The compliance verification result of the mine job personnel behavior is obtained. The job behavior control instruction is generated according to the compliance verification result of the mine job personnel behavior and sent to the mine operation control terminal.

2. The method for identifying the behavior of mine workers based on artificial intelligence according to claim 1, characterized in that, The construction of the dynamic behavioral logic benchmark model for mining positions involves, based on the operational process requirements and personnel behavior norms of each position in the mine, sorting out the correlation logic between the standard operating behaviors of personnel and process nodes, and dynamically adjusting the weights of the behavioral logic dimensions in conjunction with historical behavior identification results. This results in a dynamic behavioral logic benchmark model for mining positions that includes dynamic behavioral benchmarks and correlation constraints, including: Extract the key process nodes from the work process requirements of each position in the mine. Each key process node corresponds to a set of standard operating behaviors of the personnel in the position. Associate and bind the key process nodes with the corresponding standard operating behaviors of the personnel in the position to generate a process-behavior association table. Collect the behavioral norms documents of personnel in various positions in the mine, analyze the constraints on the work behavior of personnel in the behavioral norms documents of various positions in the mine, the constraints include the behavior execution sequence constraints and the behavior associated equipment constraints, match the constraints with the standard work behavior of personnel in the process-behavior association table, and generate the constraint-behavior correspondence relationship; Determine the core descriptive attributes of the standard operating behavior of personnel in different positions. The core descriptive attributes include the behavior execution time period attribute and the behavior associated object attribute. Extract the core descriptive attribute values ​​of the standard operating behavior of personnel in different key process nodes. Based on the process-behavior association table, the constraint-behavior correspondence, and the core description attribute values, the execution logic of the standard operating behavior of the personnel and the key process nodes is sorted out. The execution logic includes behavior-process timing logic and behavior-equipment collaboration logic to form an initial behavior benchmark. Collect historical mine personnel behavior recognition results, extract the influence coefficients of each behavioral logic dimension on the recognition accuracy from the historical mine personnel behavior recognition results, the behavioral logic dimensions include temporal logic dimension and collaborative logic dimension, and adjust the weight values ​​of each behavioral logic dimension according to the influence coefficients; The weighted execution logic is integrated with the initial behavior benchmark, and the applicable job scope identifier and logic activation conditions are added. Combined with the constraint-behavior correspondence, a dynamic behavior logic benchmark model for mine positions is formed, which includes dynamic behavior benchmarks and associated constraints. The logical consistency of the dynamic behavior logic benchmark model for the mining positions is checked to determine whether there are conflicts between the dynamic behavior benchmarks of different positions and between the dynamic behavior benchmarks and the associated constraints. If there are conflicts, the content of the dynamic behavior logic benchmark model for the mining positions is adjusted according to the priority of the job operation process requirements and the accuracy of historical identification.

3. The method for identifying the behavior of mine workers based on artificial intelligence according to claim 1, characterized in that, The process involves acquiring personnel behavior data from the mining work area. This data includes continuous behavioral records of personnel in different positions during their work. The acquired personnel behavior data is then compared with the dynamic behavioral logic baseline model for mining positions to undergo behavior difference adaptation processing. Behavioral data that differs from the dynamic behavioral baseline but meets the association constraints is then selected for identification, including: Receive raw behavioral data transmitted by behavioral acquisition devices deployed in the mining work area. The raw behavioral data includes records of the behavioral actions of the personnel during the work process and information on the scene where the behavior occurs. Each behavioral acquisition device corresponds to a specific work area. The original behavioral data is divided into behavioral segments. Based on the behavioral execution time period attribute in the dynamic behavioral logic benchmark model of the mine position, the original behavioral data is divided into multiple continuous personnel behavior records, and each personnel behavior record corresponds to a continuous work behavior time period. The behavior records of all personnel in the divided positions are stored according to their corresponding work areas, forming personnel behavior collection data that includes the behavior records of personnel in different work areas. Extract the job scope identifier and dynamic behavior benchmark from the dynamic behavior logic benchmark model of the mining job, and determine the target job and target dynamic behavior benchmark that need to be processed now; The personnel behavior records in the personnel behavior collection data are matched with the target positions, and the job matching behavior records that are consistent with the job identifier of the target position are filtered out. The job matching behavior record is compared with the target dynamic behavior benchmark by behavior logic comparison. It is determined whether the behavior execution order in the job matching behavior record is consistent with the behavior-process timing logic in the target dynamic behavior benchmark. Time difference behavior records with timing differences are filtered out. The time-series difference behavior records are matched with the association constraints in the dynamic behavior logic benchmark model of the mine position. It is determined whether the behavior-related devices in the time-series difference behavior records meet the behavior-related device constraints in the association constraints, and the time-series difference behavior records that meet the association constraints are selected. Completeness checks are performed on time-series difference behavior records that meet the association constraints to determine whether each time-series difference behavior record contains all the core descriptive attribute values ​​required by the dynamic behavior logic benchmark model of the mine position. If there are any missing values, the corresponding core descriptive attribute values ​​are added to form the behavior data to be identified.

4. The method for identifying the behavior of mine workers based on artificial intelligence according to claim 1, characterized in that, The process involves calling a pre-trained AI behavior difference adaptation model to perform bidirectional feature adaptation on the behavior data to be identified and the dynamic behavior logic benchmark model for mining positions. The AI ​​behavior difference adaptation model dynamically generates feature processing sub-paths based on the associated constraints. The dynamic behavior logic benchmark model for mining positions updates the behavior logic dimension weights based on the feature adaptation feedback output by the AI ​​behavior difference adaptation model, resulting in dynamically optimized behavior feature adaptation parameters, including: Load a pre-trained AI behavior difference adaptation model, which includes a feature input layer, a logic parsing layer, a dynamic sub-path generation layer, a parameter optimization layer, and a feedback output layer. Each layer achieves data interaction through a feature transmission channel. The behavioral data to be identified is input into the feature input layer of the AI ​​behavioral difference adaptation model, and initial feature extraction is performed on each temporal difference behavioral record in the behavioral data to be identified to obtain the initial feature vector of each temporal difference behavioral record. The dynamic behavior logic benchmark model of the mining position is input into the logic parsing layer of the AI ​​behavior difference adaptation model to parse the associated constraints, dynamic behavior benchmarks and behavior logic dimension weights in the dynamic behavior logic benchmark model of the mining position, and to determine the feature influence factor corresponding to each associated constraint. The initial feature vector of behavior and the feature influence factor corresponding to each associated constraint are input into the dynamic sub-path generation layer of the AI ​​behavior difference adaptation model. Based on the feature influence factor corresponding to each associated constraint, a feature processing sub-path is dynamically generated for each initial feature vector of behavior. The feature processing sub-path includes an associated constraint strengthening channel and an unassociated constraint filtering channel. The initial feature vector of the behavior is processed in a targeted manner through the feature processing sub-path, which strengthens the feature signals related to the associated constraints, filters the feature noise unrelated to the associated constraints, and generates the constraint adaptation feature vector for each time-series difference behavior record. The constraint adaptation feature vector is input into the parameter optimization layer of the AI ​​behavior difference adaptation model, and the degree of adaptation of each constraint adaptation feature vector with the dynamic behavior benchmark in the dynamic behavior logic benchmark model of the mine position is calculated to obtain feature adaptation feedback. The feature adaptation feedback is transmitted to the weight update module of the dynamic behavior logic benchmark model of the mine position through the feedback output layer of the AI ​​behavior difference adaptation model. The weight update module adjusts the weight value of each behavior logic dimension according to the feature adaptation feedback. The weight value of the behavior logic dimension with high adaptation is increased, and the weight value of the behavior logic dimension with low adaptation is decreased. The parameter optimization layer of the AI ​​behavior difference adaptation model combines the updated behavior logic dimension weights to optimize the fusion parameters of the constraint adaptation feature vector, generating behavior feature adaptation parameters that include channel weight coefficients, dimension adaptation coefficients, and sub-path selection coefficients. The generated behavioral feature adaptation parameters are validated for rationality. It is determined whether the channel weight coefficient allocation matches the updated behavioral logic dimension weight and whether the dimension adaptation coefficient conforms to the dimensional characteristics of the initial feature vector of the behavior. If the validation fails, the processing parameters of the dynamic sub-path generation layer are readjusted until the required behavioral feature adaptation parameters are generated.

5. The method for identifying the behavior of mine workers based on artificial intelligence according to claim 1, characterized in that, The process of enhancing the behavioral difference features of the target behavior data based on the behavioral feature adaptation parameters integrates the logical association between the behavioral difference information in the target behavior data and the dynamic behavioral benchmark, mines the implicit logical associations between behavioral differences, and generates a set of enhanced behavioral difference features with job-specific behavioral logic consistency, including: Extract the channel weight coefficient, dimension adaptation coefficient, and sub-path selection coefficient from the behavior feature adaptation parameters, and determine the weight allocation ratio, dimension adjustment standard, and feature processing sub-path selection basis of the initial behavior feature vector of each temporal difference behavior record in the behavior data to be identified in the reinforcement processing. Based on the sub-path selection coefficient, a corresponding feature processing sub-path is selected for each temporal difference behavior record in the behavior data to be identified, with the feature processing sub-path having a higher sub-path selection coefficient being selected first. The initial feature vectors of the selected feature processing sub-paths are adjusted according to the dimension adaptation coefficients, and each initial feature vector of the behavior is adjusted to an adapted feature vector of the same dimension. During the adjustment process, the behavior difference information identifier of each time-series difference behavior record is retained. The adaptive feature vectors are weighted according to the channel weight coefficients, and the weighted value of each feature channel in each adaptive feature vector is calculated to obtain the weighted feature vector. At the same time, the contribution of the behavioral difference information corresponding to each feature channel is recorded. The weighted feature vector is logically associated and integrated with the dynamic behavior benchmark in the dynamic behavior logic benchmark model of the mine position. The behavioral difference feature information related to the dynamic behavior benchmark in the weighted feature vector is extracted. The extracted behavioral difference feature information is cross-associated by combining the behavior-process timing logic and behavior-equipment collaboration logic in the dynamic behavior benchmark. Through the implicit logic mining module of the AI ​​behavior difference adaptation model, based on the inherent logic of the mine job operation process, the potential logical relationship between the behavior difference feature information after cross-correlation processing is analyzed, and the implicit logical relationship between behavior differences is supplemented. Redundancy is eliminated from the behavioral difference feature information after supplementing implicit logical associations. Key behavioral difference feature information that is directly related to the job behavior logic is retained, while duplicate feature information or feature information that does not contribute to the identification of job behavior is eliminated. The key behavioral difference feature information is combined in a unified feature arrangement order, and the feature source identifier is generated by combining the contribution of behavioral difference information corresponding to each feature channel, forming a set of enhanced behavioral difference features with job behavior logic consistency and source traceability. A logical consistency check is performed on the generated set of enhanced behavior difference features to determine whether the logical relationship of each feature information in the set of enhanced behavior difference features is consistent with the dynamic behavior benchmark in the dynamic behavior logical benchmark model of the mine position. If there is an inconsistency, the logical expression of the feature information is corrected according to the dynamic behavior benchmark to ensure the logical consistency of the set of enhanced behavior difference features.

6. The method for identifying the behavior of mine workers based on artificial intelligence according to claim 5, characterized in that, The step of selecting a corresponding feature processing sub-path for each temporal difference behavior record in the behavior data to be identified based on the sub-path selection coefficient, with the feature processing sub-path having a higher sub-path selection coefficient being selected first, includes: Extract the sub-path selection coefficients from the behavior feature adaptation parameters, and determine the selection coefficient value corresponding to each feature processing sub-path. The feature processing sub-paths include time-constrained sub-paths, device-constrained sub-paths, and hybrid-constrained sub-paths. Traverse the initial feature vector of each temporal difference behavior record in the behavior data to be identified, and analyze the core behavior difference attributes in the temporal difference behavior record corresponding to each initial feature vector. The core behavior difference attributes include behavior execution order difference attributes and behavior associated device difference attributes. Based on the core behavioral difference attributes, determine the constraint type that is more suitable for the initial feature vector of the behavior. If the significance of the behavioral execution order difference attribute is higher than that of the behavioral associated device difference attribute, then the time-series constraint sub-path is matched first; if the significance of the behavioral associated device difference attribute is higher than that of the behavioral execution order difference attribute, then the device constraint sub-path is matched first; if the significance of both attributes meets the high fit requirement, then the hybrid constraint sub-path is matched. The final selection of the appropriate feature processing sub-path is made by combining the selection coefficient value corresponding to each feature processing sub-path. If the selection coefficient value of the preferred feature processing sub-path is higher than the preset sub-path selection threshold, the preferred feature processing sub-path is directly selected. If the selection coefficient value of the preferred feature processing sub-path is lower than the preset sub-path selection threshold, the other feature processing sub-path with the highest selection coefficient value is selected. Record the feature processing sub-path selection results and selection criteria corresponding to the initial feature vector of each behavior. The selection criteria include the core behavior difference attribute analysis results and feature processing sub-path selection coefficient values. Cross-validate the feature processing sub-path selection results of all initial feature vectors of behavior to determine whether there is a logical conflict between the feature processing sub-paths selected by different initial feature vectors of behavior in the same job area. If there is a conflict, the feature processing sub-path is re-selected according to the updated behavior logic dimension weight in the dynamic behavior logic benchmark model of the mine job to ensure the logical consistency of the feature processing sub-paths in the same job area. The feature processing sub-path selection result and the verified feature processing sub-path corresponding to each behavior's initial feature vector are associated and bound with the corresponding behavior's initial feature vector.

7. The method for identifying the behavior of mine workers based on artificial intelligence according to claim 5, characterized in that, The implicit logic mining module of the AI ​​behavior difference adaptation model, based on the inherent logic of the mine job operation process, analyzes the potential logical relationships between the cross-correlation processed behavior difference feature information, and supplements the implicit logical relationships between behavior differences, including: Load the internal logic knowledge base of the mining job operation process. The internal logic knowledge base of the mining job operation process includes the standard operation step sequence of each job, the dependency relationship between the standard operation steps, and the description of the typical behavior action corresponding to each standard operation step. The cross-correlation processed behavioral difference feature information is input into the feature mapping submodule of the implicit logic mining module of the AI ​​behavioral difference adaptation model. The behavioral difference features in the cross-correlation processed behavioral difference feature information are mapped and matched with the typical behavioral action descriptions in the internal logic knowledge base of the mine job operation process to determine the standard operation steps corresponding to each behavioral difference feature in the cross-correlation processed behavioral difference feature information. Based on the dependencies between standard operation steps in the internal logical knowledge base of the mining job operation process, the logical relationships between standard operation steps corresponding to different behavioral difference features in the behavioral difference feature information after cross-association processing are analyzed. If the standard operation steps corresponding to two behavioral difference features have a sequential dependency relationship, then these two behavioral difference features are marked as having a step-dependent implicit logical relationship. Extract the time series information from the behavioral difference feature information after cross-association processing, analyze the occurrence order and interval pattern of different behavioral difference features in the time dimension, and if two behavioral difference features maintain a fixed time interval and occurrence order in multiple consecutive operation periods, then mark that these two behavioral difference features have a time-series-related implicit logical association. Analyze the equipment association information in the behavioral difference feature information after cross-association processing, and combine it with the association relationship between standard operation steps and equipment in the internal logical knowledge base of the mining job operation process to determine whether there is a collaborative operation relationship between the equipment corresponding to different behavioral difference features. If there is a collaborative operation relationship, mark the corresponding behavioral difference feature as having a collaborative implicit logical association with equipment. The labeling results of step-dependent implicit logical associations, time-series implicit logical associations, and device-cooperative implicit logical associations are associated and stored with the corresponding behavioral difference features in the behavioral difference feature information after cross-association processing, forming an implicit logical association description table. The implicit logical association description table is validated for rationality. Referring to the actual operation scenario of the mine, it is determined whether the marked implicit logical associations conform to the internal logic of the actual operation process. If there are implicit logical associations that do not conform to the actual logic, the corresponding markers are deleted and implicit logical association markers that conform to the actual logic are added. The verified implicit logical association information is added to the behavioral difference feature information after cross-association processing to form a complete set of behavioral difference feature information containing both explicit and implicit logical associations.

8. The method for identifying the behavior of mine workers based on artificial intelligence according to claim 1, characterized in that, The rule base for verifying compliance of mine job behavior is based on a pre-set rule library. The set of enhanced behavioral difference features is then verified for compliance. The set of enhanced behavioral difference features is judged against the associated constraints in the dynamic behavioral logic benchmark model for mine jobs to determine whether it meets the job behavior requirements. This yields the compliance verification results for mine job personnel behavior, including: The system calls upon a pre-defined rule library for compliance verification of mine job behaviors, parses the compliance standards for each job behavior contained in the rule library, and each job behavior compliance standard corresponds to a set of compliance verification conditions. The compliance verification conditions are formulated based on the mine job operation safety specifications and operating procedure requirements. Extract the feature information and feature source identifier from the enhanced behavior difference feature set, determine the target position corresponding to the enhanced behavior difference feature set, and retrieve the job behavior compliance standard and corresponding compliance verification conditions corresponding to the target position from the mine job behavior compliance verification rule library; Each behavioral difference feature in the enhanced behavioral difference feature set is matched one by one with the corresponding compliance verification conditions. It is determined whether each behavioral difference feature meets the corresponding compliance verification conditions. If it does, it is marked as a compliant difference feature; otherwise, it is marked as a non-compliant difference feature. The matching results for each behavioral difference feature are recorded to generate a preliminary verification record containing feature identifiers, matching result identifiers, and descriptions of non-compliance reasons. The matching result identifiers include two types: compliant and non-compliant. Referring to the associated constraints in the dynamic behavior logic benchmark model of the mining position, the non-compliant difference features in the preliminary verification record are further analyzed to determine whether the non-compliant difference features violate the behavior execution order constraints or behavior-related equipment constraints in the associated constraints. If they violate them, they are marked as serious non-compliant difference features; if they do not violate them, they are marked as general non-compliant difference features. The preliminary verification records are summarized and statistically analyzed to calculate the proportion of compliant difference features, the proportion of non-compliant difference features, the number of serious non-compliant difference features, and the number of general non-compliant difference features in the enhanced behavior difference feature set. Based on the statistical results and in conjunction with the preset compliance judgment criteria, the compliance level of the mine personnel's behavior corresponding to the set of enhanced behavioral difference characteristics is determined. The compliance level is used to indicate the degree to which the mine personnel's behavior meets the requirements of their job. The compliance level, the percentage of compliance difference features, the percentage of non-compliance difference features, the details of serious non-compliance difference features, and the details of general non-compliance difference features are integrated to generate the compliance verification result of the behavior of personnel in mining positions.

9. The method for identifying the behavior of mine workers based on artificial intelligence according to claim 1, characterized in that, The process of generating a job behavior control instruction based on the compliance verification results of the mine personnel's behavior, wherein the job behavior control instruction includes a behavior compliance identifier and corresponding control measures, and sending the job behavior control instruction to the mine operation control terminal, includes: The compliance level in the compliance verification results of the personnel's behavior in the mining positions is analyzed, and the corresponding behavior compliance mark is determined according to the compliance level. The behavior compliance mark that meets the preset compliance standard is a compliance mark, and the behavior compliance mark that does not meet the preset compliance standard is a non-compliance mark. The system calls upon a pre-defined database of mine job behavior control measures. Based on the compliance flag and details of non-compliance differences, it retrieves the corresponding control measures. If the compliance flag is compliant, it retrieves routine behavior monitoring measures. If the compliance flag is non-compliant, it retrieves emergency behavior intervention measures based on details of serious non-compliance differences, and retrieves behavior warning and guidance measures based on details of general non-compliance differences. By associating and integrating the behavioral compliance identifiers, corresponding control measures, details of non-compliance difference characteristics (if any), and the feature source identifiers of the enhanced behavioral difference characteristic set, basic data for job behavior control is generated. The basic data for job behavior control is formatted and standardized according to the data format that can be recognized by the mine operation control terminal. Data generation time information and job identification information are added to form job behavior control instructions. Establish an encrypted communication connection with the mine operation control terminal, and send the job behavior control instructions to the corresponding mine operation control terminal through an encrypted transmission protocol; The system receives confirmation information from the mine operation control terminal. If no confirmation information is received within the preset feedback waiting time, the system resends the job behavior control instruction until confirmation information is received from the mine operation control terminal. Records of sending job behavior control instructions, receipt confirmation records of mine operation control terminals, and compliance verification results of mine personnel behavior are linked and stored in the mine job behavior control log database.

10. A mine worker behavior recognition system based on artificial intelligence, characterized in that, The AI-based mine worker behavior recognition system includes a processor and a memory, the memory and the processor being connected. The memory is used to store programs, instructions or code, and the processor is used to execute the programs, instructions or code in the memory to implement the AI-based mine worker behavior recognition method according to any one of claims 1-9.

Images