A transmission line image monitoring optimization method

Abstract

The application discloses a power transmission line image monitoring optimization method to solve the problems of insufficient multi-source feature correlation, threshold value adaptation without self-learning ability and missing cross-link collaborative regulation in the prior art. The method constructs and runs a feature coupling model and a dynamic threshold value self-learning model, outputs a multi-source feature coupling correlation result and an adaptive threshold value system, inputs a cross-link linkage compensation mechanism to complete abnormality identification, multi-link parameter synchronous adjustment, coupling verification and feedback optimization, and finally integrates in an operation and maintenance platform to realize whole-process early warning and data sedimentation. The application realizes multi-source feature deep correlation, threshold value dynamic self-adaptation and cross-link collaborative regulation, significantly improves the adaptability, operation stability and abnormality disposal timeliness of the monitoring system to complex scenes, and meets the unmanned, accurate and efficient monitoring requirements of the power transmission line.

Description

Technical Field

[0001] This invention relates to the field of power transmission line monitoring technology, specifically to an optimization method for power transmission line image monitoring. Background Technology

[0002] As the core carrier of electrical energy transmission in the power system, the stable and reliable operation of transmission lines directly affects the safe power supply and the normal operation of social production and life. With the continuous expansion of power grid coverage, transmission lines are mostly distributed in complex environments such as mountainous areas, suburbs, and coastal areas, facing various risks such as meteorological disasters, environmental interference, and equipment aging. Therefore, real-time and accurate image monitoring of transmission lines to promptly detect line defects and abnormal operating conditions has become a crucial aspect of power operation and maintenance. Currently, transmission line image monitoring largely relies on various monitoring devices to collect images and operational data, and optimizes the monitoring process through relevant algorithms and mechanisms to improve monitoring effectiveness and reliability.

[0003] Existing methods for optimizing image monitoring of transmission lines generally employ multi-dimensional data acquisition and single-stage parameter adjustments. For image acquisition optimization, parameters such as camera exposure and resolution are adjusted to adapt to environmental changes; for power supply stability optimization, fixed threshold-controlled charging and discharging strategies are used; and for communication transmission optimization, switching transmission links or adjusting compression ratios ensures data transmission. While these methods improve the operational stability of individual stages to some extent, they do not fully consider the inherent connections between monitoring stages. They are mostly independent optimization models, lacking deep coupling analysis of multi-dimensional characteristics such as environment, power supply, communication, and equipment operation, resulting in insufficient synergy in the optimization process.

[0004] Meanwhile, existing threshold adaptation technologies mostly rely on fixed settings or manual adjustments, lacking the ability to learn from historical data and real-time scenarios. When the monitoring scenario undergoes dynamic changes, such as sudden weather changes or line load adjustments, the threshold system cannot adapt in a timely manner, easily leading to problems such as false positives or false negatives in image recognition, unstable power supply, and communication transmission interruptions. Furthermore, for abnormal features observed during monitoring, existing methods mostly only adjust the link to which the anomaly belongs independently, without establishing a cross-linkage compensation mechanism. This means that optimizing a single link may trigger operational conflicts in other links, further reducing the adaptability and reliability of the overall monitoring system, and increasing operation and maintenance costs and the risk of delays in fault handling.

[0005] In summary, existing methods for optimizing transmission line image monitoring suffer from technical deficiencies such as insufficient multi-source feature correlation, lack of self-learning capability in threshold adaptation, and lack of cross-link collaborative control. These deficiencies result in poor adaptability of the monitoring system to complex dynamic scenarios and insufficient overall operational stability, making it difficult to meet the actual needs of unattended, accurate, and efficient monitoring of transmission lines. Summary of the Invention

[0006] The purpose of this invention is to provide an optimization method for image monitoring of power transmission lines to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for optimizing image monitoring of transmission lines, comprising the following steps: Construct and run a feature coupling model and a dynamic threshold self-learning model based on multi-source data perception; wherein, the feature coupling model receives multi-dimensional operational data during the transmission line monitoring process and outputs multi-source feature coupling correlation results; the dynamic threshold self-learning model receives the feature coupling correlation results, historical optimization data and monitoring scenario requirements, and outputs an adaptive threshold system for each monitoring link; The feature coupling correlation results and the adaptive threshold system are input into the cross-linkage compensation mechanism, which performs the following operations to achieve collaborative optimization of the monitoring parameters: The system compares the current multi-source operating data with the adaptive threshold system in real time to identify abnormal feature values ​​that exceed the threshold range; for the abnormal feature values, it synchronously adjusts the image acquisition parameters, power output parameters, and communication transmission parameters to achieve collaborative compensation for defects in each link. Based on the aforementioned feature coupling model, the operating status after parameter adjustment is coupled and verified to analyze the correlation and influence of parameter adjustment in each link, and to locate the parameter deviations and core causes that still exist after regulation. The parameter deviation and core cause are used as feedback to drive the dynamic threshold self-learning model to update the adaptive threshold system, while optimizing the feature association logic of the feature coupling model and improving the model adaptation accuracy. The optimized feature coupling model, dynamic threshold self-learning model, and cross-linkage compensation mechanism are integrated and deployed on the transmission line monitoring and operation platform. The integrated model is driven by real-time collected multi-dimensional operation data to achieve full-process early warning and data accumulation in the monitoring process. The multi-dimensional operational data includes at least environmental perception data, image acquisition data, power supply status data, communication status data, and equipment operating condition data; the adaptation threshold system covers at least image acquisition adaptation threshold, power supply stability threshold, and communication transmission qualification threshold; and the historical optimization data includes at least historical parameter adjustment data, operational status verification data, deviation processing data, and early warning handling data.

[0008] Preferably, a feature coupling model based on multi-source data perception is constructed and run, including: Multi-dimensional operational data is collected by the built-in sensing module of the power transmission line image monitoring device, and the multi-dimensional operational data is preprocessed to remove data noise and redundant information. Extract the core representational features of each dimension of the preprocessed data, establish the mapping relationship between features of different dimensions through feature correlation analysis, and construct a multi-source feature coupling model. The coupling features in the coupling model are weighted and determined based on the degree of influence of the features on the monitoring effect. The weights can be dynamically adjusted according to the actual monitoring scenario, and the multi-source feature coupling correlation results are output.

[0009] Preferably, building and running a dynamic threshold self-learning model includes: Based on the gradient descent algorithm, the optimal fit range of each parameter in the historical optimization data is calculated iteratively to generate an initial fit threshold system. Establish a threshold update trigger mechanism to automatically trigger the threshold update process when the monitoring scenario changes or the accumulated deviation data reaches a preset level; Based on the feature coupling correlation results and the monitoring scenario requirements, the initial adaptation threshold system is dynamically adjusted to output an adaptation threshold system that is adapted to the current monitoring conditions.

[0010] Preferably, the feature coupling model receives multi-dimensional operational data during the transmission line monitoring process and outputs multi-source feature coupling correlation results. The dynamic threshold self-learning model receives the feature coupling correlation results, historical optimization data, and monitoring scenario requirements, and outputs an adapted threshold system, including: The preprocessed multi-dimensional operational data is input into the feature coupling model. Through feature mapping and weight calculation within the model, the multi-source feature coupling correlation result is output. The correlation result represents the intrinsic correlation between environmental features, equipment operation features, power supply features and communication features. The feature coupling and correlation results, historical optimization data, and monitoring scenario requirements are input into the dynamic threshold self-learning model. Through threshold iterative calculation and adaptation adjustment, an adaptation threshold system is output, which includes image acquisition adaptation threshold, power supply stability threshold, and communication transmission qualification threshold.

[0011] Preferably, the real-time comparison of the current multi-source operating data with the adaptive threshold system identifies abnormal feature values ​​that exceed the threshold range, including: The feature values ​​of each dimension corresponding to the current multi-source running data are compared one by one with the corresponding threshold range in the adaptive threshold system; For each dimension's feature value, determine whether it falls within the corresponding threshold range. If the feature value exceeds the threshold range, mark it as an abnormal feature value and record the value, dimension, and occurrence time of the abnormal feature value. The marked abnormal feature values ​​are classified and summarized to form an abnormal feature list, which provides a basis for subsequent cross-linked linkage compensation and control.

[0012] Preferably, for the abnormal feature value, the image acquisition parameters, power output parameters, and communication transmission parameters are adjusted synchronously and in conjunction, including: Based on the feature coupling correlation results, the correlation weight between the link to which the abnormal feature value belongs and other monitoring links is determined; Based on the priority of the associated weights, the operating parameters of the associated links are adjusted synchronously: if the communication transmission characteristic value is abnormal, the image acquisition parameters are optimized simultaneously to reduce the data transmission volume and ensure that the power supply module gives priority to the communication module; if the power supply characteristic value is abnormal, the power supply output parameters are adjusted simultaneously to optimize the power consumption parameters of image acquisition and communication transmission. The parameter adjustment process does not require any modification to the hardware structure of the monitoring device; parameter adaptation is achieved solely through software algorithms.

[0013] Preferably, the operating state after parameter adjustment is coupled and verified based on the feature coupling model, the correlation and influence of parameter adjustment in each link are analyzed, and the parameter deviations and core causes that still exist after regulation are located, including: Data is collected on the operational status of each stage after parameter adjustment to obtain multi-dimensional operational data after adjustment. The adjusted multi-dimensional operational data is input into the feature coupling model to verify the effect of a single link and the correlation of multiple links, and to determine whether there are any new abnormal feature values ​​or parameter correlation conflicts. The feature source analysis method is adopted to trace the initial features and fundamental factors that cause parameter deviations based on the feature correlation relationship in the feature coupling model. The fundamental factors include at least environmental interference factors, parameter correlation conflict factors, and threshold adaptation deviation factors.

[0014] Preferably, the learning rate of the gradient descent algorithm is dynamically adjusted based on historical optimization results. When the historical optimization deviation is small, the learning rate is reduced to ensure threshold stability; when the historical optimization deviation is large, the learning rate is increased to accelerate threshold adaptation. After the initial adaptation threshold system is constructed, it needs to be verified in a small-scale scenario. Only after the verification is passed can it be applied to actual monitoring. After the threshold update process is triggered, the parameter changes and adaptation effects before and after the threshold adjustment are recorded synchronously to form a threshold adjustment log, which provides data support for subsequent model optimization.

[0015] Preferably, the image acquisition parameter adjustment includes at least resolution adjustment, frame rate adjustment, exposure adjustment, and night vision mode switching; the power supply output parameter adjustment includes at least power supply mode switching, output current adjustment, and charging / discharging strategy adjustment; and the communication transmission parameter adjustment includes at least transmission link switching, data compression ratio adjustment, and transmission rate adaptation. The magnitude and direction of parameter adjustments in each stage are determined based on the correlation weights in the feature coupling correlation results. The higher the correlation weight of a stage, the stronger the synergy of parameter adjustments, ensuring that the operating states of each stage are compatible after adjustment, and avoiding new anomalies in other stages caused by optimization of a single stage.

[0016] Preferably, the correlation verification is implemented by coupling degree calculation. When the coupling degree after the adjustment of parameters of multiple links reaches the preset qualified standard, it is determined that the verification is passed; otherwise, it is determined that there is an abnormal correlation and a secondary control is triggered. The feature source tracing analysis method specifically involves tracing the associated feature chain of abnormal feature values ​​based on a multi-source feature coupling model. By analyzing the changing patterns and mutual influences of each feature in the associated feature chain, the starting feature and core cause of the abnormality can be located, providing accurate feedback for threshold iteration and model optimization.

[0017] Compared with existing technologies, the beneficial effects of this invention are: by constructing a feature coupling model, a dynamic threshold self-learning model, and a cross-linkage compensation mechanism, it realizes deep correlation analysis of multi-source data, dynamic adaptation of thresholds, and collaborative optimization across links. It has the advantages of improving the overall adaptability, operational reliability, and synergy of the monitoring system, effectively responding to the dynamic changes of complex field environments, and achieving efficient operation and maintenance. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the image monitoring optimization method according to an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1 This paper proposes an optimization method for image monitoring of power transmission lines. The method aims to effectively solve the aforementioned problems by constructing and running a feature coupling model based on multi-source data perception and a dynamic threshold self-learning model, combined with a cross-linkage compensation mechanism, to achieve coordinated optimization of monitoring parameters. The method includes the following steps: For ease of understanding, the following explains some key terms in this embodiment: Feature Coupling Model: This model is configured to receive multi-dimensional operational data during the transmission line monitoring process and output multi-source feature coupling correlation results by analyzing the inherent correlations between data from different dimensions. Its function is to reveal the mutual influence and dependencies between data from various monitoring stages, providing a data foundation for subsequent dynamic threshold adaptation and coordinated control.

[0021] Dynamic Threshold Self-Learning Model: This model is configured to receive feature coupling correlation results, historical optimization data, and monitoring scenario requirements. Through a learning and iterative process, it outputs an adapted threshold system for each monitoring stage. Its core function is to dynamically adjust and optimize thresholds based on real-time monitoring status and historical experience to adapt to complex and ever-changing monitoring environments and ensure the accuracy of anomaly identification.

[0022] Multi-dimensional operational data refers to various types of data collected during the monitoring of transmission lines, including at least environmental sensing data, image acquisition data, power supply status data, communication status data, and equipment operating condition data. These data collectively reflect the operational status of the transmission lines.

[0023] Multi-source feature coupling correlation results: Output by the feature coupling model, representing the inherent correlations revealed after in-depth analysis of operational data from different dimensions. This result quantifies the degree of mutual influence between features and serves as an important basis for achieving cross-stage collaborative optimization.

[0024] The adaptive threshold system, output by a dynamic threshold self-learning model, is a set of dynamic thresholds for each stage of transmission line monitoring, covering at least image acquisition adaptation thresholds, power supply stability thresholds, and communication transmission qualification thresholds. This system is adjusted based on real-time monitoring scenarios and historical optimization experience to determine whether the current operating data is within the normal range.

[0025] Cross-stage linkage compensation mechanism: This mechanism is configured to receive multi-source feature coupling correlation results and adapt the threshold system, and is responsible for real-time comparison of operating data with thresholds, identifying abnormal feature values, and synchronously adjusting image acquisition parameters, power supply output parameters, and communication transmission parameters in response to anomalies. Its goal is to achieve collaborative compensation for defects in each monitoring stage, avoiding problems in other stages caused by optimization of a single stage.

[0026] Transmission Line Monitoring and Maintenance Platform: A comprehensive platform integrating an optimized feature coupling model, a dynamic threshold self-learning model, and a cross-linkage compensation mechanism. This platform utilizes real-time, multi-dimensional operational data to drive the integrated model, enabling full-process early warning, data accumulation, and operation and maintenance management during monitoring.

[0027] This embodiment provides a method for optimizing image monitoring of power transmission lines, the specific implementation process of which includes the following steps: First, a feature coupling model and a dynamic threshold self-learning model based on multi-source data perception are constructed and run. As one implementation, the feature coupling model is a simple statistical correlation analysis module used to calculate the linear correlation coefficient between different data dimensions. By analyzing the relationship between historical ambient temperature and image acquisition clarity, a preliminary correlation model is established. This model is configured to receive multi-dimensional operational data during transmission line monitoring, including environmental perception data collected by the monitoring device's sensors, image acquisition data obtained by the image acquisition module, power supply status data fed back by the power supply module, communication status data transmitted by the communication module, and equipment operating condition data provided by the equipment itself. This data is input into the model, processed through preset calculation logic, and outputs the multi-source feature coupling correlation results. As another implementation, the feature coupling model is a logical judgment system based on predefined rules. It judges the correlation between different features through a series of rules; if the wind speed exceeds a certain threshold, it is considered to potentially affect image stability. The dynamic threshold self-learning model is a system that sets thresholds based on historical data averages and a fixed deviation range. Its initial threshold is manually set based on long-term operational experience. The model is configured to receive the aforementioned feature coupling correlation results, historical optimization data, and monitoring scenario requirements. Based on these inputs, and through a pre-defined algorithm logic, it outputs an adaptation threshold system for each monitoring stage. The upper and lower limits of the image acquisition adaptation threshold are calculated based on the average and standard deviation of historical image acquisition data.

[0028] Furthermore, the aforementioned feature coupling correlation results and the adapted threshold system are input into the cross-stage linkage compensation mechanism. This cross-stage linkage compensation mechanism is configured to perform the following operations to achieve collaborative optimization of monitoring parameters.

[0029] Specifically, this mechanism compares the current multi-source operational data with the adapted threshold system in real time to identify abnormal feature values ​​that exceed the threshold range. Each feature value corresponding to the currently collected environmental perception data, image acquisition data, power supply status data, communication status data, and equipment operating condition data is compared one by one with the corresponding threshold range in the adapted threshold system. If a feature value is lower than its lower threshold limit or higher than its upper threshold limit, it is identified as an abnormal feature value. In one implementation, the comparison process is a simple numerical comparison module; when the input data exceeds a preset range, the system generates an anomaly marker. When the communication signal strength is lower than a preset minimum acceptable value, the system marks it as abnormal.

[0030] For identified abnormal feature values, image acquisition parameters, power supply output parameters, and communication transmission parameters are adjusted synchronously and in a coordinated manner to achieve defect compensation in each stage. When an abnormality in the image acquisition data is detected, the system, according to preset correlation logic, adjusts not only the exposure parameters of the image acquisition module but also the output power of the power supply module to ensure stable operation of the image acquisition module, and adjusts the data transmission rate of the communication module to accommodate potentially increased data processing demands. In one implementation, parameter adjustment is based on a preset lookup table. When a specific type of anomaly is detected, the lookup table indicates the parameters that need to be adjusted and their adjustment amounts. When abnormal fluctuations occur in the power supply voltage, the lookup table indicates that the frame rate of image acquisition should be reduced by a fixed step to reduce power consumption.

[0031] Subsequently, a coupling verification is performed on the operational status after parameter adjustment based on the feature coupling model. This analyzes the correlational impact of parameter adjustments on each stage, pinpointing the remaining parameter deviations and their core causes after adjustment. After parameter adjustment, the system re-collects operational data from each stage and inputs it into the feature coupling model to assess whether the adjustment introduced new correlation problems or completely resolved the original anomaly. In one implementation, coupling verification involves independent monitoring of each parameter after adjustment, checking whether the image acquisition clarity, power supply voltage stability, and communication transmission success rate have returned to normal. If any parameter still has not returned to normal, a parameter deviation is considered to exist.

[0032] Based on this, the aforementioned parameter deviations and core causes are used as feedback to drive the dynamic threshold self-learning model to update the adaptation threshold system. Simultaneously, the feature association logic of the feature coupling model is optimized to improve model adaptation accuracy. If the verification finds that the adjusted image acquisition parameters still have deviations, and the power supply status also fluctuates slightly, this deviation information will be fed back to the dynamic threshold self-learning model, prompting it to adjust the image acquisition adaptation threshold and power supply stability threshold. At the same time, the feature coupling model will also reassess the correlation strength between environmental factors and image acquisition and power supply based on new operating data and deviation information to optimize its feature association logic. In one implementation, the feedback mechanism is a simple log recording system, with manual review of the logs and manual updates to the model parameters.

[0033] Finally, the optimized feature coupling model, dynamic threshold self-learning model, and cross-linkage compensation mechanism are integrated and deployed on the transmission line monitoring and maintenance platform. This integrated model is driven by real-time, multi-dimensional operational data, enabling full-process early warning and data accumulation during monitoring. These models and mechanisms, as independent software modules, are installed on monitoring devices or cloud servers and exchange data through standard interfaces. The platform is responsible for the unified management of these modules and provides a user interface for monitoring and data querying. In one implementation, the integrated deployment runs each model as an independent application on different servers, interacting with data through file sharing or a simple API. Real-time, multi-dimensional operational data is periodically sent to these applications for processing.

[0034] This embodiment effectively solves the problems of insufficient multi-source feature correlation, lack of self-learning ability in threshold adaptation, and lack of cross-link collaborative control in transmission line image monitoring by constructing a multi-source feature coupling model and a dynamic threshold self-learning model, and introducing a cross-link linkage compensation mechanism. As a result, the monitoring system can achieve collaborative optimization and dynamic adaptation of parameters according to complex and ever-changing monitoring scenarios, improving the overall operational stability and the timeliness of anomaly handling, thereby meeting the actual needs of unattended, accurate, and efficient monitoring of transmission lines.

[0035] In some of the embodiments described above in this application, a feature coupling model is proposed to be constructed and run to output multi-source feature coupling correlation results. In its implementation process, data acquisition may be subject to noise and redundant information interference, feature extraction may be inaccurate, resulting in the loss of core representations, inaccurate feature correlation mapping relationship affects model construction, and fixed weight allocation cannot adapt to changes in dynamic monitoring scenarios, resulting in inaccurate coupling results, thereby reducing the synergy and accuracy of subsequent monitoring optimization processes.

[0036] In response, this application further proposes to construct and run a feature coupling model based on multi-source data perception, specifically including: Multi-dimensional operational data is collected by the built-in sensing module of the power transmission line image monitoring device, and the multi-dimensional operational data is preprocessed to remove data noise and redundant information. Extract the core representational features of each dimension of the preprocessed data, establish the mapping relationship between features of different dimensions through feature correlation analysis, and construct a multi-source feature coupling model. The coupling features in the coupling model are weighted and determined based on the degree of influence of the features on the monitoring effect. The weights can be dynamically adjusted according to the actual monitoring scenario, and the multi-source feature coupling correlation results are output.

[0037] Specifically, multi-dimensional operational data is collected through the built-in sensing module of the transmission line image monitoring device. This sensing module typically integrates multiple sensors, including temperature, humidity, and wind speed sensors for acquiring environmental parameters; visible light cameras and infrared thermal imagers for capturing line images; voltage and current sensors for monitoring power supply status; and signal strength detection modules for assessing communication quality. These sensors work together to comprehensively and in real-time acquire environmental sensing data, image acquisition data, power supply status data, communication status data, and equipment operating condition data during transmission line operation, providing rich and comprehensive raw data input for subsequent monitoring and optimization. The multi-dimensional operational data undergoes preprocessing to remove noise and redundant information, aiming to improve data quality and ensure the accuracy of subsequent analysis. Kalman filtering and wavelet denoising algorithms are used to filter out random noise during sensor acquisition. Simultaneously, principal component analysis, independent component analysis, or feature selection methods based on information entropy are used to identify and eliminate redundant information in the data, thereby reducing data dimensionality and improving processing efficiency.

[0038] Based on this, core representational features of the preprocessed data from each dimension are extracted. Through feature correlation analysis, mapping relationships between features of different dimensions are established, constructing a multi-source feature coupling model. Core representational features refer to key attributes that effectively reflect the operating status of transmission lines and environmental changes. For image data, edge features, texture features, color histograms, or deep features extracted through convolutional neural networks are extracted. For environmental data, statistical features of temperature, humidity, and wind speed are extracted. Feature correlation analysis aims to reveal the intrinsic connections and mutual influence mechanisms between features of different dimensions. Statistical methods such as mutual information and Pearson correlation coefficients are used to quantify the linear or nonlinear correlation strength between features, or deep learning models such as graph neural networks are used to construct complex feature maps to capture implicit dependencies between multi-source features, thereby forming a multi-source feature coupling model that can comprehensively represent the operating status of transmission lines.

[0039] Furthermore, weights are assigned to the coupling features in the coupled model. These weights are determined based on the degree of influence of each feature on the monitoring effect and can be dynamically adjusted according to the actual monitoring scenario, outputting multi-source feature coupling and correlation results. The purpose of weight assignment is to highlight key features that affect the monitoring effect and reduce the interference of secondary features. Initial weights can be set based on expert experience, or the importance of features can be evaluated and weights assigned using machine learning algorithms. An attention mechanism can also be used to allow the model to automatically focus on more important features when dealing with different scenarios. Simultaneously, to adapt to the dynamic changes in transmission line monitoring scenarios, the weight assignment mechanism has dynamic adjustment capabilities. A reinforcement learning algorithm is introduced to adaptively adjust the weights of each feature based on historical optimization results and real-time monitoring feedback, ensuring that the model always outputs multi-source feature coupling and correlation results that best reflect the current monitoring conditions.

[0040] Through the aforementioned technical solution, this application effectively eliminates data noise and redundant information by performing refined preprocessing on multi-dimensional operational data, ensuring the purity and reliability of the input data. Based on this, the core representational features of each dimension are accurately extracted, and feature correlation analysis is used to establish accurate mapping relationships between features of different dimensions. This constructs a structurally sound and logically clear multi-source feature coupling model, enhancing the model's comprehensive perception capability of complex transmission line operating states. Furthermore, by using a weight allocation mechanism based on the degree of impact of monitoring effects and dynamically adjusting it according to actual monitoring scenarios, the model can intelligently highlight key features and adapt to environmental changes, thereby outputting more accurate and timely multi-source feature coupling correlation results. This not only solves the problems of inaccurate data, inaccurate feature correlation, and rigid weights, but more importantly, it provides high-quality, high-precision input for subsequent dynamic threshold self-learning models and cross-linkage compensation mechanisms. This greatly improves the synergy and accuracy of the entire transmission line image monitoring optimization method, enabling the system to more effectively identify anomalies, adjust parameters, and ultimately achieve full-process early warning and data accumulation during monitoring.

[0041] In some of the solutions mentioned above in this application, a dynamic threshold self-learning model is proposed to output the adaptive threshold system for each monitoring link. In this process, the generation of the initial threshold lacks intelligent iterative calculation based on historical data, the threshold update depends on manual triggering and is not timely, and the dynamic adjustment does not fully consider the feature coupling correlation results and the needs of real-time monitoring scenarios, resulting in insufficient threshold adaptation accuracy.

[0042] In response, this application further proposes an optimization method for transmission line image monitoring, which involves constructing and running a dynamic threshold self-learning model, including the following steps: based on the gradient descent algorithm, the optimal adaptation range of each parameter in historical optimization data is iteratively calculated to generate an initial adaptation threshold system; a threshold update triggering mechanism is constructed to automatically trigger the threshold update process when the monitoring scene changes or the accumulated deviation data reaches a preset level; and the initial adaptation threshold system is dynamically adjusted by combining the feature coupling correlation results with the monitoring scene requirements to output an adaptation threshold system that adapts to the current monitoring conditions.

[0043] Specifically, in the step of generating an initial fitness threshold system by iteratively calculating the optimal fit range of each parameter in historical optimization data based on the gradient descent algorithm, gradient descent is a commonly used optimization method. Its core lies in iteratively adjusting parameters along the direction of gradient descent of the objective function to gradually approach the local minimum of the function. In this application, this algorithm is used to learn and determine the optimal value range of each monitoring parameter from a large amount of historical optimization data. Batch gradient descent is used, where all historical data are used to calculate the gradient and update the parameters in each iteration; alternatively, stochastic gradient descent is used, where a sample is randomly selected in each iteration for gradient calculation and parameter update to accelerate convergence and avoid local optima. Historical optimization data refers to the verified and adjusted operational data accumulated during transmission line monitoring, including historical parameter adjustment data, operational status verification data, deviation handling data, and early warning response data. The optimal fit range refers to the allowable optimal value interval for each monitoring parameter while ensuring monitoring effectiveness and system stability. Iterative computation refers to the process of repeatedly calculating and progressively optimizing parameters to approximate the optimal solution. A loss function is defined to measure the deviation between the current threshold system and historical optimization data. Then, using the gradient descent algorithm, the threshold parameters are continuously adjusted to minimize this loss function, thereby obtaining the optimal fit range. The initial fit threshold system generated through the above process provides a data-driven starting point for subsequent real-time monitoring and parameter adjustment, avoiding the subjectivity and inaccuracy caused by relying entirely on manual experience to set thresholds.

[0044] In constructing a threshold update trigger mechanism, which automatically triggers the threshold update process when the monitoring scenario changes or the accumulated deviation data reaches a preset level, the threshold update trigger mechanism aims to ensure that the adapted threshold system can respond promptly to changes in the monitoring environment and system operating status. This mechanism automatically initiates the threshold update process through preset condition judgments, thereby avoiding the lag of manual intervention. The mechanism is triggered based on time periods, system performance indicators, or environmental sensor data. Changes in the monitoring scenario refer to changes in the external environment or internal operating conditions that affect the monitoring effect of transmission lines, such as extreme weather, seasonal changes in illumination, adjustments to transmission line loads, and maintenance or upgrades of monitoring equipment. These changes may render the original adapted threshold system inapplicable. Accumulated deviation data refers to the cumulative amount of abnormal feature values ​​identified by the system during the monitoring process that exceed the current adapted threshold range. The preset level is the number of abnormal events that occur, such as five consecutive detections of image blurring anomalies, the degree to which abnormal feature values ​​deviate from the threshold center (e.g., the average deviation exceeding a certain percentage), or the duration of the anomaly (e.g., a certain abnormal state lasting more than 30 minutes). When these cumulative deviations reach a preset threshold, it indicates that the current threshold system may no longer be suitable for actual operating conditions and needs to be updated. Automatic triggering means that the system autonomously initiates the threshold update calculation and adjustment process based on preset logic and conditions without manual intervention. This automation capability improves the system's response speed and efficiency to dynamic changes, reduces the workload of maintenance personnel, and ensures the continuity and accuracy of monitoring.

[0045] Based on this, and combining the feature coupling correlation results with the monitoring scenario requirements, the initial adaptation threshold system is dynamically adjusted to output an adaptation threshold system suitable for the current monitoring conditions. The feature coupling correlation results are the inherent relationships between various dimensions of features output by the feature coupling model after in-depth analysis of multi-dimensional operational data during transmission line monitoring. These relationships include environmental perception data, image acquisition data, power supply status data, communication status data, and equipment operating condition data. These correlation results are used to guide the direction and magnitude of threshold adjustments during dynamic adjustment. If the feature coupling correlation results show a strong correlation between ambient temperature and image acquisition parameters, the system will prioritize adjusting the image acquisition-related thresholds when the ambient temperature changes. Monitoring scenario requirements refer to the specific performance requirements of the monitoring system under specific monitoring conditions. In critical transmission lines or high-risk areas, higher image recognition accuracy and lower false alarm rates may be required. In low-power mode, some thresholds may need to be relaxed to save energy; at night or in severe weather, thresholds may need to be adjusted to adapt to low-light or high-noise environments. These requirements serve as important references for dynamically adjusting thresholds. Dynamic adjustment refers to continuous, real-time fine-tuning and optimization based on the initial adaptive threshold system, according to real-time monitoring data, feature coupling and correlation results, and the requirements of the current monitoring scenario. This adjustment can be rule-based (e.g., adjusting the image exposure threshold to Y if the ambient brightness is below X) or model-based (using a small predictive model to predict the optimal threshold based on real-time input). The final output adaptive threshold system, adapted to the current monitoring conditions, not only considers historical data and initial settings but also incorporates factors such as real-time environmental changes, equipment operating status, and monitoring targets, thereby ensuring that the monitoring system maintains optimal performance and reliability under various complex dynamic scenarios.

[0046] Through the above technical solution, this application specifically addresses the problem of insufficient threshold adaptation accuracy by constructing and running a dynamic threshold self-learning model. First, based on the gradient descent algorithm, an initial adaptation threshold system is generated by iteratively calculating the optimal adaptation range of each parameter in historical optimization data. This method utilizes the iterative characteristics of gradient descent to intelligently learn the optimal parameter range from historical data, ensuring that the initial threshold system is data-driven and avoiding the subjectivity and inaccuracy of manual settings. Second, a threshold update trigger mechanism is constructed. When the monitoring scenario changes or the accumulated deviation data reaches a preset level, the threshold update process is automatically triggered. This mechanism automatically responds to changes based on preset conditions without manual intervention, ensuring the timeliness and dynamism of threshold updates and preventing adaptation failure caused by delayed updates. Finally, combining the feature coupling correlation results with the monitoring scenario requirements, the initial adaptation threshold system is dynamically adjusted to output an adaptation threshold system suitable for the current monitoring conditions. This step integrates the inherent correlation of the feature coupling results and real-time requirements for dynamic fine-tuning, ensuring that the threshold system not only inherits historical optimizations but also fits the characteristics of the current operating conditions, thereby outputting a more accurate adaptation threshold system and improving overall monitoring accuracy and stability.

[0047] In some of the solutions mentioned above in this application, a feature coupling model and a dynamic threshold self-learning model are proposed to provide correlation results and a threshold system. In their implementation process, without specific feature mapping and weight calculation mechanisms, the model may not be able to accurately capture the inherent correlation between environmental features, equipment operation features, power supply features and communication features. At the same time, without the threshold iterative calculation and adaptation adjustment process, the threshold system is difficult to dynamically respond to changes in the monitoring scenario requirements, resulting in distorted correlation results and lagging threshold adaptation, which in turn affects the accuracy and real-time performance of cross-link collaborative optimization.

[0048] To address this, this application further proposes a feature coupling model that receives multi-dimensional operational data during transmission line monitoring and outputs multi-source feature coupling correlation results. A dynamic threshold self-learning model receives the feature coupling correlation results, historical optimization data, and monitoring scenario requirements and outputs an adaptive threshold system. Specifically, this includes: inputting preprocessed multi-dimensional operational data into the feature coupling model, and outputting multi-source feature coupling correlation results through feature mapping and weight calculation within the model. The correlation results represent the inherent correlation between environmental features, equipment operation features, power supply features, and communication features. Furthermore, the application proposes inputting the feature coupling correlation results, historical optimization data, and monitoring scenario requirements into the dynamic threshold self-learning model, and outputting an adaptive threshold system including image acquisition adaptation thresholds, power supply stability thresholds, and communication transmission qualification thresholds through threshold iterative calculation and adaptation adjustment.

[0049] Specifically, inputting preprocessed multi-dimensional running data into the feature coupling model aims to provide high-quality input data and ensure the accuracy of subsequent feature association analysis. The preprocessed multi-dimensional running data has been cleaned to remove noise, redundancy, or errors, thus avoiding the negative impact of low-quality data on the model's output. One approach is to perform real-time or batch processing of the raw data using data cleaning algorithms at the data acquisition end or data aggregation layer, and then stream or batch load the processed data into the feature coupling model. Another approach is to integrate a data preprocessing module within the feature coupling model. When the raw multi-dimensional running data is received, it first performs data standardization, normalization, and missing value imputation operations before sending the processed data to the feature analysis module.

[0050] The core of the feature coupling model lies in its internal feature mapping and weight calculation, which reveal the deep correlations between features of different dimensions and quantify their impact on the overall monitoring effect. Feature mapping aims to establish logical connections between different types of data, while weight calculation assigns different weights based on the importance of these features or their degree of influence on the monitoring target. One implementation method is to use a deep learning model, such as a graph neural network (GNN) or an attention mechanism, taking features of different dimensions as nodes or sequence inputs. The network layers learn the complex mapping relationships between features and automatically assign weights. GNNs can capture the topological relationships between features, while attention mechanisms highlight the contribution of key features to the output results. Another implementation method is to use statistical methods or machine learning algorithms, such as principal component analysis, canonical correlation analysis, or random forests, to perform dimensionality reduction and correlation analysis on features, and then determine the weights of each feature through expert experience or regression models trained on historical data. Weights are then assigned based on the predictive ability of features to monitor abnormal events.

[0051] The system outputs multi-source feature coupling correlation results, which characterize the intrinsic relationships between environmental features, equipment operation features, power supply features, and communication features. These results are a direct output of the feature coupling model and aim to provide a comprehensive quantitative indicator reflecting the operational status of various aspects of the transmission line monitoring system and their mutual influences. These correlation results serve as crucial bases for subsequent dynamic threshold self-learning models to adjust thresholds and for cross-linkage compensation mechanisms to regulate parameters. One implementation method is to represent the correlation results as a multi-dimensional vector or matrix, where each element represents the coupling strength or influence factor between specific environmental features, equipment operation features, power supply features, or communication features. An correlation matrix is ​​output, with values ​​representing the degree of influence of different features on each other. Another implementation method is to represent the correlation results as a set of rules or model parameters that describe how equipment operation, power supply, and communication status influence each other under specific environmental conditions. For example, the correlation between the output power of the power supply module and the transmission rate of the communication module strengthens when the ambient temperature rises.

[0052] The feature coupling correlation results, historical optimization data, and monitoring scenario requirements are input into the dynamic threshold self-learning model, providing it with multifaceted information input to ensure that it can comprehensively consider the current operating status, historical experience, and actual needs to generate an adapted threshold system. The feature coupling correlation results provide real-time, in-depth system operation insights, historical optimization data provides valuable lessons learned, and monitoring scenario requirements clarify the focus of the current monitoring task. One implementation method is to encapsulate the correlation results output in real-time by the feature coupling model, historical optimization data queried from the database, and monitoring scenario requirements obtained through the user interface or preset strategies into a unified data packet, and send it to the dynamic threshold self-learning model asynchronously or synchronously, via a data interface or message queue. Another implementation method is for the dynamic threshold self-learning model to actively pull the required data from the feature coupling model, historical database, and configuration service, and perform data fusion to form a unified input dataset.

[0053] The core function of a dynamic threshold self-learning model is its iterative threshold calculation and adaptive adjustment, designed to dynamically optimize and adjust thresholds at each monitoring stage based on input information. Iterative calculation ensures continuous learning and improvement of the thresholds, while adaptive adjustment guarantees their flexible response to changes in the monitoring scenario. One implementation approach is to use reinforcement learning algorithms, treating threshold adjustment as a decision-making process. Through interaction with the environment, the threshold strategy is iteratively optimized based on feedback from monitoring results. Another approach is to use global optimization methods such as Bayesian optimization or genetic algorithms to search for the optimal threshold combination within a preset threshold range. The search process is guided and constrained by feature coupling correlation results, historical optimization data, and monitoring scenario requirements, thereby achieving iterative threshold calculation and adaptive adjustment.

[0054] The output includes an adaptation threshold system encompassing image acquisition adaptation thresholds, power supply stability thresholds, and communication transmission qualification thresholds. This is the final output of the dynamic threshold self-learning model, aiming to provide a comprehensive, dynamic, and accurate threshold standard to guide cross-linkage compensation mechanisms in anomaly identification and parameter adjustment. This threshold system covers key aspects of transmission line image monitoring, ensuring overall optimization of the monitoring process. One implementation method is to output the adaptation threshold system in a structured data format, including the dynamic threshold range for image acquisition parameters, the stable threshold range for power supply parameters, and the qualification threshold range for communication transmission parameters. Another implementation method is to directly integrate the adaptation threshold system into the configuration management module of the monitoring and maintenance platform for real-time invocation by the cross-linkage compensation mechanism. This system can be a dynamically updated lookup table or a set of executable threshold judgment functions.

[0055] Through the above technical solution, this application can input preprocessed multi-dimensional operational data into a feature coupling model, and utilize the model's internal feature mapping and weight calculation mechanism to accurately capture the inherent correlation between environmental features, equipment operation features, power supply features, and communication features, thereby outputting high-precision multi-source feature coupling correlation results. These correlation results provide a solid foundation for subsequent threshold calculations, avoiding misjudgments caused by inaccurate feature correlation.

[0056] Based on this, the feature coupling and correlation results, historical optimization data, and monitoring scenario requirements are input into the dynamic threshold self-learning model. Through iterative threshold calculation and adaptation adjustment, the threshold system can dynamically respond to changes in the monitoring scenario. This dynamic adjustment capability solves the problem that traditional fixed thresholds or manually adjusted thresholds cannot adapt to complex and changing monitoring environments, ensuring that the output adaptive threshold system, including image acquisition adaptation thresholds, power supply stability thresholds, and communication transmission qualification thresholds, always highly matches the current operating conditions. This application effectively solves the problems of inaccurate feature correlation and non-dynamic threshold adaptation, improving the synergy, accuracy, and real-time performance of the transmission line image monitoring optimization method. High-precision feature coupling and correlation results provide more reliable input for dynamic threshold self-learning, while the dynamically adapted threshold system can more effectively guide the cross-linkage compensation mechanism to adjust parameters, thereby avoiding the negative impact of distorted correlation results and lagging threshold adaptation on the overall monitoring optimization effect, ultimately ensuring the stable and reliable operation of the transmission line monitoring system.

[0057] In some of the embodiments described above in this application, a cross-linkage compensation mechanism is proposed to achieve coordinated optimization of monitoring parameters. However, in its implementation, there is a lack of an effective mechanism to compare data with thresholds in real time, accurately identify abnormal feature values, and record detailed information. This may lead to low efficiency in anomaly handling and insufficient basis for compensation control, affecting the overall coordination and reliability of the monitoring system.

[0058] To address this, this application further proposes a method for real-time comparison of the matching status of current multi-source operational data with the adaptive threshold system, and for identifying abnormal feature values ​​that exceed the threshold range. Specifically, the method includes: comparing the feature values ​​of each dimension corresponding to the current multi-source operational data with the corresponding threshold range in the adaptive threshold system one by one; determining whether the feature value of each dimension is within the corresponding threshold range; if the feature value exceeds the threshold range, it is marked as an abnormal feature value, and the value, dimension, and occurrence time of the abnormal feature value are recorded; the marked abnormal feature values ​​are classified and summarized to form an abnormal feature list, providing a basis for subsequent cross-linkage linkage compensation and control.

[0059] Specifically, real-time comparison of current multi-source operational data with the adapted threshold system identifies abnormal characteristic values ​​exceeding the threshold range. This aims to quickly and accurately detect potential operational anomalies by comparing real-time monitoring data with preset, dynamically adjusted thresholds. This comparison process is implemented by deploying a real-time data processing module within the transmission line monitoring and maintenance platform. This module continuously receives multi-dimensional operational data from the transmission line monitoring devices and compares it in parallel with the adapted threshold system output by the dynamic threshold self-learning model. Streaming data processing technology ensures low latency in data processing. Alternatively, a periodic batch processing approach can be used, performing a centralized comparison of the latest collected multi-source operational data every preset time interval, such as 1 minute or 5 minutes. This method is suitable for scenarios where real-time requirements are not extremely stringent but a large amount of historical data needs to be processed. The comparison program is triggered by a scheduled task system.

[0060] The process of comparing the feature values ​​of each dimension corresponding to the current multi-source operational data with the corresponding threshold ranges in the adaptive threshold system one by one ensures a comprehensive check of all key monitoring dimensions, leaving no potential anomalies unchecked. This step guarantees the accuracy of refined management and anomaly localization. This process can be implemented programmatically using a loop or iterator to traverse each feature dimension in the current multi-source operational data, including ambient temperature, image brightness, power supply voltage, and communication packet loss rate, and for each dimension, retrieve its corresponding upper and lower thresholds from the adaptive threshold system. Then, the current feature value is numerically compared with the threshold range. Alternatively, a rule-based engine approach can be used, pre-defining a series of comparison rules, each rule corresponding to a feature dimension and its threshold range. When new multi-source operational data arrives, the data is input into the rule engine, which automatically executes all matching rules to complete the one-to-one comparison.

[0061] For each dimension's feature value, it is determined whether it falls within the corresponding threshold range. If a feature value exceeds the threshold range, it is marked as an anomalous feature value, and the specific logic for anomaly detection is recorded, including the value of the anomalous feature value, its corresponding dimension, and the time of occurrence. Detailed recording of anomaly information is emphasized to provide foundational data for subsequent analysis and processing. During the comparison process, if a feature value is less than its lower threshold or greater than its upper threshold, the system marks it as an anomaly. Simultaneously, the specific value of the anomalous feature value, its corresponding dimension identifier, and the timestamp of the anomaly are stored in structured data records, database tables, or log files. Alternatively, anomaly detection operators in a data stream processing framework can be used to automatically trigger an anomaly event when a data point falls outside a preset threshold range. This event carries the anomalous data point itself, its metadata, and a timestamp, encapsulated into an anomaly record object, and sent to a message queue or persistent storage.

[0062] The system categorizes and summarizes marked abnormal feature values ​​to form an abnormal feature list, providing a basis for subsequent cross-process coordinated compensation and control. The aim is to structure and organize scattered abnormal information, enabling maintenance personnel or automated systems to quickly grasp the full picture of abnormalities and providing a clear and centralized basis for subsequent decision-making. An abnormality management module is designed to periodically query the latest abnormal data from a database storing abnormal records. This module categorizes abnormalities based on their dimensions and types, such as environmental abnormalities and power supply abnormalities, and summarizes abnormalities of the same type or within the same time period, generating a highly readable abnormal feature list presented in JSON, XML, or report format. Additionally, in the real-time data processing workflow, once an abnormal feature value is marked, it can be immediately sent to a dedicated abnormality aggregation service. This service is responsible for receiving, classifying, and summarizing abnormalities in real time and maintaining a dynamically updated abnormal feature list. When cross-process coordinated compensation and control is required, the latest abnormal list is directly obtained from this service.

[0063] Through the above technical solution, this application can compare current multi-source operating data with the adapted threshold system in real time, instantly identifying abnormal feature values ​​exceeding the threshold range. By recording the numerical value, dimension, and occurrence time of abnormal feature values ​​in detail and classifying and summarizing them to form a structured list of abnormal features, this solves the problem of lacking effective means to identify anomalies in real time and accurately and provide detailed evidence in cross-linked compensation mechanisms. This mechanism ensures real-time monitoring of the transmission line operating status, avoids delays in anomaly response, and enables earlier intervention to prevent problems from escalating. Comparing feature values ​​of each dimension with the corresponding threshold range ensures that all key monitoring points are checked without omission, improving the comprehensiveness of anomaly identification. At the same time, detailed recording of anomaly information provides a solid data foundation for subsequent anomaly tracing, cause analysis, and precise control. By forming a structured list of abnormal features, a clear and centralized basis is provided for cross-linked compensation control, making subsequent parameter adjustments more targeted and avoiding blind adjustments, thereby improving the overall efficiency and reliability of the collaborative optimization of the monitoring system. By combining the constructed feature coupling model and dynamic threshold self-learning model, this implementation method can use the adaptive threshold system provided by the self-learning model for comparison to ensure the dynamic adaptability of the threshold. At the same time, the identified abnormal feature values ​​will be used as input to drive the cross-linkage compensation mechanism for coordinated adjustment and feed back to the dynamic threshold self-learning model for optimization, forming a closed-loop optimization process, which improves the intelligence level and operational stability of transmission line image monitoring.

[0064] In some of the solutions mentioned above in this application, the synchronous correlation adjustment of image acquisition parameters, power supply output parameters and communication transmission parameters is proposed to achieve collaborative compensation for defects in each link. In this process, the adjustment may lack a weight priority mechanism based on the feature coupling correlation results, resulting in uncoordinated parameter adjustment, causing operational conflicts or insufficient optimization in other links. When communication is abnormal, the amount of image data is not reduced synchronously, thus increasing the power supply burden, or when power supply is abnormal, power consumption is not optimized, thus affecting communication stability.

[0065] In response, this application further proposes a specific implementation method for synchronously adjusting image acquisition parameters, power supply output parameters, and communication transmission parameters in response to abnormal feature values. This includes: determining the correlation weight between the abnormal feature value's component and other monitoring components based on the feature coupling correlation results; and synchronously adjusting the operating parameters of the correlated components according to the priority of the correlation weights. If a communication transmission-related feature value is abnormal, while adjusting the communication transmission parameters, the image acquisition parameters are simultaneously optimized to reduce data transmission volume and ensure priority power supply from the power supply module to the communication module. If a power supply-related feature value is abnormal, while adjusting the power supply output parameters, the power consumption parameters for image acquisition and communication transmission are simultaneously optimized. The parameter adjustment process does not require modification of the monitoring device's hardware structure; parameter adaptation is achieved solely through software algorithms.

[0066] Based on the feature coupling correlation results, the correlation weights between the link to which the abnormal feature value belongs and other monitoring links are determined, aiming to quantify the impact of abnormal feature values ​​on other monitoring links. The feature coupling correlation results are obtained after in-depth analysis of multi-dimensional operational data by the feature coupling model, revealing the inherent interdependencies between different monitoring links. By utilizing these correlations, the potential chain reaction or impact strength on other links when an anomaly occurs in one link can be accurately assessed. The correlation weights are determined by analyzing the connection strength or path length between feature nodes in the feature coupling model. If the model uses a graph structure to represent feature relationships, the weights can be calculated based on edge weights between nodes or shortest path algorithms. Alternatively, machine learning methods, such as regression analysis or neural networks, can be used to train on historical anomaly data and parameter adjustment effects to learn and predict the impact of different anomaly feature values ​​on parameters of other links, thereby outputting quantified correlation weights.

[0067] Synchronously adjusting the operating parameters of related links based on correlation weight priority refers to the coordinated adjustment of the operating parameters of other monitoring links related to the abnormal link after identifying abnormal characteristic values, according to the pre-determined correlation weight. The priority mechanism ensures that when resources are limited or there are adjustment conflicts, the links with the greatest impact on the overall monitoring effect can be addressed first, thus achieving efficient and coordinated parameter optimization. A threshold is set, and synchronous adjustment is only performed when the correlation weight exceeds this threshold. Alternatively, all related links can be sorted by weight from high to low, with the link with the highest weight adjusted first, and the decision to continue adjusting other links dynamically based on the adjustment effect. Simultaneously, a multi-objective optimization algorithm can be used, treating the operating parameters of each link as optimization variables and the correlation weight as part of the objective function. Through iterative calculation, an optimal set of parameter adjustment schemes can be found, resulting in optimal overall monitoring performance and minimal conflicts between links.

[0068] If communication transmission characteristic values ​​are abnormal, the system adjusts communication transmission parameters while simultaneously optimizing image acquisition parameters to reduce data transmission volume and ensuring priority power supply from the power supply module to the communication module. This describes the coordinated compensation strategy adopted by the system when communication transmission anomalies occur. Communication transmission anomalies may lead to data loss, delays, or interruptions, affecting the timeliness and integrity of monitoring data. Synchronously optimizing image acquisition parameters reduces data volume, alleviates the burden on communication transmission, and improves transmission success rate. Simultaneously, prioritizing power supply from the power supply module to the communication module ensures sufficient energy support for the communication module at critical moments, maintaining its basic functions and preventing complete communication interruption due to insufficient power. When a decrease in communication transmission rate or an increase in packet loss rate is detected, the system automatically triggers the image acquisition module to switch the image resolution from high-definition mode to standard-definition mode and reduce the frame rate. Simultaneously, the power supply management unit adjusts the power distribution strategy to ensure that the voltage and current of the communication module remain stable within the operating range.

[0069] If power supply characteristics are abnormal, the system adjusts power supply output parameters while simultaneously optimizing power consumption parameters for image acquisition and communication transmission. This describes the system's coordinated compensation strategy when power supply issues arise. Power supply anomalies may manifest as unstable voltage, insufficient current, or low battery power, directly impacting the normal operation of the entire monitoring device. The system stabilizes power supply by adjusting output parameters and simultaneously optimizes power consumption parameters for image acquisition and communication transmission. This includes reducing image acquisition frequency and decreasing standby power consumption of the communication module, effectively extending the device's runtime or maintaining core monitoring functions under power constraints. When battery power falls below a warning threshold or external power is interrupted, the system automatically switches to low-power mode. The image acquisition module reduces acquisition frequency and image quality, the communication module enters intermittent transmission mode, and the power management unit adjusts output voltage and current to adapt to low-power operation.

[0070] The parameter adjustment process requires no modification to the monitoring device's hardware structure; parameter adaptation is achieved solely through software algorithms. This emphasizes the method of parameter adjustment, which is accomplished through software-level algorithmic control rather than physical hardware alterations. This means the optimization method offers high flexibility, deployability, and low cost. Through software algorithms, adjustment strategies can be remotely updated, iterated, and optimized without on-site manual intervention or equipment replacement, significantly improving operational efficiency and system adaptability. The control unit inside the monitoring device runs a pre-set software program. This program receives adjustment commands from the cross-linkage compensation mechanism and sends instructions to the image sensor, power management chip, communication module, etc., via the control interface to modify their working registers or configuration parameters.

[0071] Through the above technical solution, this application effectively solves the problems of lack of coordination in parameter adjustment and the tendency to cause secondary problems in traditional methods. The weighted priority mechanism based on feature coupling and correlation results makes parameter adjustment decisions more scientific and accurate, avoiding the negative impacts of blind adjustment. This intelligent linkage compensation mechanism ensures that when an anomaly occurs in one link, other related links can respond collaboratively, forming an overall optimized solution, rather than simple independent optimization. Furthermore, since the parameter adjustment process is implemented only through software algorithms, there is no need to modify the hardware structure of the monitoring device, greatly improving the system's flexibility, maintainability, and deployment efficiency, reducing operation and maintenance costs, and enabling the transmission line image monitoring system to maintain a continuously efficient and stable operating state in complex and changing operating environments, thus improving the overall reliability and adaptability of monitoring.

[0072] In some of the solutions mentioned above in this application, parameter adjustment is proposed to achieve collaborative compensation for defects in each link. However, in the process of implementation, there is a lack of comprehensive verification and analysis of the operating status after adjustment, and it is impossible to accurately locate the root cause of parameter deviation.

[0073] To address this, this application further proposes a coupled verification method based on a feature coupling model to check the operational status after parameter adjustment, analyze the correlation and impact of parameter adjustments at each stage, and pinpoint the parameter deviations and core causes that still exist after regulation. Specifically, this method includes: First, data is collected on the operational status of each component after parameter adjustment to obtain multi-dimensional operational data. This step aims to obtain the actual operational status of the system after parameter adjustment, providing a real-time and accurate data foundation for subsequent verification and analysis. Through various sensors and detectors built into the transmission line image monitoring device, adjusted image quality parameters, voltage and current data of the power supply module, packet loss rate and latency of the communication link, and other operational parameters are continuously collected. Alternatively, a dedicated post-adjustment data collection cycle can be initiated, centrally collecting operational data from all relevant subsystems—image acquisition, power supply, and communication—after parameter adjustment is completed, ensuring data integrity and synchronization.

[0074] Subsequently, the adjusted multi-dimensional operational data is input into the feature coupling model for single-stage effect verification and multi-stage correlation verification to determine whether there are any new abnormal feature values ​​or parameter correlation conflicts. This step utilizes the feature coupling model to comprehensively evaluate the adjustment effect. Specifically, the feature coupling model receives this new multi-dimensional operational data and uses its internally established feature mapping and weight calculation logic to regenerate multi-source feature coupling correlation results. By comparing these new correlation results with the expected optimization target or historical baseline, it can be evaluated whether the adjustment of a single stage has achieved the expected effect. At the same time, the model also analyzes the mutual influence between features of different stages, identifying whether there are any anomalies or conflicts in other stages caused by the optimization of one stage, such as increasing image resolution potentially leading to excessive power load or insufficient communication bandwidth.

[0075] Based on this, a feature-based source analysis method is employed. Based on the feature correlation relationships within the feature coupling model, the initial features and root causes of parameter deviations are traced. These root causes include at least environmental interference factors, parameter correlation conflict factors, and threshold adaptation deviation factors. The feature-based source analysis method aims to deeply uncover the underlying causes of parameter deviations. Using a pre-defined causal chain or correlation graph within the feature coupling model, the method traces the associated feature chain backwards from the detected parameter deviation. If image blurring is highly correlated with a sudden drop in ambient light intensity, the root cause may be environmental interference. If image blurring is highly correlated with voltage fluctuations in the power supply module, and these voltage fluctuations are related to excessive adjustments in image acquisition parameters, the root cause may be parameter correlation conflict. If image blurring is related to a long-term deviation of the image acquisition adaptation threshold from actual operating conditions, the root cause may be threshold adaptation deviation. Another approach is to integrate a rule-based inference engine or Bayesian network into the feature coupling model. When a parameter deviation is detected, this engine automatically infers the most likely root cause based on learned feature correlation weights and historical anomaly patterns.

[0076] Through the above technical solutions, this application ensures comprehensive verification and analysis of the operational status after parameter adjustment. By collecting multi-dimensional operational data after adjustment and inputting it into the feature coupling model for single-stage effect verification and multi-stage correlation verification, it can promptly detect whether there are new abnormal feature values ​​or parameter correlation conflicts in the adjusted operational status, avoiding the problem that optimization of a single stage may cause operational conflicts in other stages. Furthermore, by employing feature source tracing analysis, based on the feature correlation relationships established in the feature coupling model, it can accurately trace the initial features and root factors causing parameter deviations, such as environmental interference factors, parameter correlation conflict factors, or threshold adaptation deviation factors. This enables the system to fundamentally understand and solve problems, rather than merely making superficial adjustments, thereby providing accurate and powerful feedback for subsequent dynamic threshold self-learning model updates and feature coupling model optimization, improving the overall adaptability, stability, and reliability of the transmission line image monitoring system.

[0077] In some of the embodiments described above in this application, a dynamic threshold self-learning model is proposed to be constructed using the gradient descent algorithm to generate an adaptive threshold. In its implementation, a fixed learning rate may lead to low threshold adjustment efficiency, the threshold system may introduce errors due to lack of verification, and the lack of logging hinders subsequent optimization.

[0078] In response, this application further proposes that the learning rate of the gradient descent algorithm be dynamically adjusted based on historical optimization results. When the historical optimization deviation is small, the learning rate is reduced to ensure threshold stability; when the historical optimization deviation is large, the learning rate is increased to accelerate threshold adaptation. After the initial adaptation threshold system is constructed, it needs to be verified in a small-scale scenario before it can be applied to actual monitoring. After the threshold update process is triggered, the parameter changes and adaptation effects before and after the threshold adjustment are recorded synchronously to form a threshold adjustment log, providing data support for subsequent model optimization.

[0079] Specifically, the learning rate of the gradient descent algorithm determines the step size for updating model parameters in each iteration. Dynamically adjusting the learning rate aims to flexibly change the step size based on the model's performance during optimization, thereby improving convergence speed and stability. Adaptive learning rate algorithms, such as Adam, RMSprop, or Adagrad, can automatically adjust the learning rate of each parameter based on historical gradient information; or they can be adjusted through a preset strategy, multiplying the learning rate by a decay factor after a certain number of iterations or when the change in the loss function is less than a certain threshold. When historical optimization deviations are small, the learning rate is reduced to ensure threshold stability. This means that when the model optimization is close to convergence, reducing the learning rate avoids the model oscillating around the optimal solution, helping to converge more accurately to a local optimum. Alternatively, a deviation threshold can be set, and when the historical optimization deviation, such as the change in the loss function over several consecutive iterations or the improvement in validation set performance, is lower than this threshold, the learning rate is reduced proportionally. Another approach is to use a performance-based scheduler, which automatically reduces the learning rate when the performance on the validation set does not improve over N consecutive epochs. Conversely, when historical optimization bias is large, the learning rate is increased to accelerate threshold fitting. This allows the model to escape local optima more quickly in the early stages of optimization or when encountering large biases, thus accelerating the convergence process. The learning rate can be increased proportionally when historical optimization bias is detected to be higher than a certain baseline value, or a larger learning rate can be used in the early stages of optimization. Alternatively, a cyclical learning rate or learning rate warm-up strategy can be employed, gradually increasing the learning rate in the early stages of training to help the model find a good parameter space more quickly.

[0080] Furthermore, after the initial adaptive threshold system is constructed, it needs to be validated in a small-scale scenario. Before applying the newly constructed or significantly updated adaptive threshold system to actual monitoring, validation in a controlled, small-scale real or simulated scenario can identify potential problems and inaccuracies in advance, reducing the risks associated with direct deployment. A small number of typical and representative transmission line monitoring points should be selected, and the new threshold system should be deployed at these points for short-term trial operation, comparing its effectiveness with existing or manually set thresholds. Alternatively, a simulation environment can be constructed to simulate various operating conditions and abnormal situations of the transmission line, and the new threshold system can be applied to this simulation environment for testing, evaluating its performance under different scenarios. Only after successful validation can it be applied to actual monitoring, ensuring that only threshold systems that have undergone sufficient testing and proven their effectiveness and reliability can be officially put into use, thereby guaranteeing the accuracy and stability of transmission line monitoring. Set clear verification standards. For example, in small-scale scenario verification, the false alarm rate and false negative rate of the new threshold system should be lower than the preset threshold, and the accuracy of anomaly identification should reach a certain level. Alternatively, establish an approval process where operation and maintenance experts or system administrators review and confirm the verification results, and full deployment can only be carried out after approval.

[0081] Furthermore, after the threshold update process is triggered, the parameter changes and adaptation effects before and after the threshold adjustment are recorded synchronously, forming a threshold adjustment log. Recording historical information about threshold adjustments provides data for subsequent model optimization and problem troubleshooting, helping to understand the effectiveness of threshold adjustments and their impact on system performance. Each time a threshold is updated, information such as the threshold values ​​before and after the adjustment, the adjustment time, the reason for the update, and the short-term performance in the validation set or actual monitoring after the update is automatically stored in a database or log file. Alternatively, a log management module can be designed to record not only the numerical changes of the threshold itself, but also historical optimization data related to this adjustment, monitoring scenario requirements, and verification data of the system's operating status after the adjustment, forming structured log entries. This log data provides data support for subsequent model optimization. The accumulated threshold adjustment log is an important foundation for continuous model improvement and optimization. By analyzing these logs, the performance patterns of the model in different scenarios can be discovered, thereby optimizing the feature association logic of the feature coupling model or the update strategy of the dynamic threshold self-learning model. Regularly perform data mining and analysis on threshold adjustment logs to identify patterns that lead to frequent or ineffective threshold adjustments, thereby guiding model developers to improve the algorithm; or use the log data as part of the training set to train a meta-learning model that learns how to more effectively adjust the learning rate or trigger threshold updates.

[0082] Through the above technical solution, this application effectively solves the problem of low efficiency caused by the fixed learning rate in the dynamic threshold self-learning process of the gradient descent algorithm. Specifically, the learning rate is dynamically adjusted based on historical optimization results, enabling the model to converge quickly in the early stages of optimization and to converge stably when approaching the optimal solution, avoiding unnecessary oscillations and thus improving the efficiency and accuracy of threshold adaptation. Simultaneously, by conducting small-scale scenario verification after the initial adapted threshold system is constructed, errors and risks that may be introduced by directly applying an unverified threshold system to actual monitoring are effectively avoided, ensuring the reliability of the threshold system. Furthermore, parameter changes and adaptation effects are recorded synchronously after the threshold update process is triggered, forming a valuable threshold adjustment log. This provides a solid data foundation for the continuous optimization of the subsequent dynamic threshold self-learning model and feature coupling model, enabling the entire monitoring optimization method to continuously improve itself and adapt to the complex and ever-changing transmission line monitoring environment, enhancing the robustness and long-term operational stability of the system.

[0083] In some of the solutions mentioned above in this application, a cross-linkage compensation mechanism is proposed to synchronously and correlate the adjustment of image acquisition parameters, power supply output parameters, and communication transmission parameters. In this process, the specific implementation of parameter adjustment may lack clear scope definition and coordination guarantee mechanism, which may lead to the incompatibility of the operating status of each link after adjustment. There is a risk that optimization of a single link may cause new anomalies in other links. Inappropriate parameter adjustment magnitude or directional deviation may aggravate the conflict between links and reduce the stability and reliability of the overall monitoring system.

[0084] To address this, this application further proposes a specific implementation method for synchronously adjusting image acquisition parameters, power output parameters, and communication transmission parameters in response to the aforementioned abnormal feature values. The image acquisition parameter adjustment includes at least resolution adjustment, frame rate adjustment, exposure adjustment, and night vision mode switching; the power output parameter adjustment includes at least power mode switching, output current adjustment, and charging / discharging strategy adjustment; and the communication transmission parameter adjustment includes at least transmission link switching, data compression ratio adjustment, and transmission rate adaptation. The magnitude and direction of parameter adjustments at each stage are determined based on the correlation weight in the feature coupling correlation results. Stages with higher correlation weights exhibit stronger synergy in parameter adjustments, ensuring that the operating states of each stage are mutually compatible after adjustment, and preventing optimization of a single stage from causing new anomalies in other stages.

[0085] Specifically, adjusting image acquisition parameters is a key setting that directly affects the quality and amount of information in the monitored images. By adjusting these parameters, the image acquisition module can adapt to different ambient lighting conditions, target movement speeds, and requirements for image detail, thereby acquiring clear and effective monitoring images. Resolution adjustment is achieved by changing the number of pixels output by the image sensor, switching from high-definition (HD) mode to standard-definition (SD) mode to reduce data volume, or switching from SD mode to HD mode to obtain more detail. Frame rate adjustment adapts to the target movement speed by controlling the number of image frames captured per second, increasing the frame rate to reduce blur in fast-moving target scenes, or decreasing the frame rate in static scenes to save power and storage space. Exposure adjustment controls the image brightness by changing the exposure time or gain of the image sensor to adapt to different lighting conditions, reducing exposure in strong light and increasing exposure in low light. Night vision mode switching typically involves turning infrared illuminators on / off and switching image processing algorithms to acquire visible images in complete darkness or extremely low light environments.

[0086] Adjusting power output parameters is a crucial control variable for ensuring stable operation and extending the battery life of the monitoring device. By flexibly adjusting these parameters, energy distribution is optimized to ensure the power needs of each module under different operating conditions and to effectively manage battery health. Power supply mode switching involves switching between different modes such as mains power, backup battery power, and solar power to adapt to the availability of external power. Output current adjustment dynamically adjusts the output current of the power module according to current load demand, increasing the current when the image acquisition module is under high load and decreasing the current in standby mode to save energy. Charging and discharging strategy adjustment involves battery charging and discharging management, prioritizing external power when the battery is fully charged and switching to battery power when external power is insufficient. The charging current and cutoff voltage are adjusted according to battery health and ambient temperature to extend battery life.

[0087] Adjusting communication transmission parameters is crucial for ensuring timely and reliable transmission of monitoring data. Optimizing these parameters addresses the complexity of different network environments, balances data transmission efficiency and reliability, and prevents data loss or delays. Transmission link switching selects between different communication methods such as 4G / 5G cellular networks, Wi-Fi, and satellite communication to address differences in signal strength, bandwidth, and cost. Data compression ratio adjustment balances transmission bandwidth and image quality by changing the compression level of image or video data; it increases the compression ratio when network bandwidth is limited and decreases it when bandwidth is sufficient to retain more detail. Transmission rate adaptation dynamically adjusts the data transmission rate based on current network conditions and data volume requirements to avoid network congestion or transmission interruptions. It reduces the transmission rate when the network signal is weak and increases it when the network signal is strong to maximize transmission efficiency.

[0088] The magnitude and direction of parameter adjustments at each stage are determined based on the correlation weights in the feature coupling correlation results. This feature clarifies the decision-making basis for parameter adjustments. The feature coupling correlation results reflect the inherent strength of the connections between different monitoring stages, such as image acquisition, power supply, and communication, as well as their connections with environmental factors and equipment operating conditions. The correlation weights quantify these connections, guiding the system to prioritize stages with the greatest impact on the overall system when adjusting parameters, and ensuring the correctness of the adjustment direction. Correlation weights are learned from historical data during the training phase using machine learning models, such as neural networks and decision trees, or initialized using expert experience and preset rules. These weights are dynamically updated based on the analysis results of multi-dimensional operational data by the feature coupling model to reflect the correlation strength in real-time scenarios. When an abnormal feature value is identified, the system queries the correlation weights of other stages related to that abnormal feature value in the feature coupling correlation results. Stages with higher weights will receive higher adjustment priority or larger adjustment magnitudes to ensure effective compensation for anomalies and avoid negative impacts on other related stages.

[0089] The higher the correlation weight of a component, the stronger the synergy of parameter adjustments, emphasizing the collaborative adjustment mechanism based on correlation weight. A high correlation weight implies a close interdependence between two or more components; adjusting one component will affect others. Therefore, when adjusting parameters, the system prioritizes synchronous and coordinated adjustments to these highly correlated components to achieve overall optimization. If the correlation weight between image acquisition and power supply is high, such as high-resolution acquisition leading to a surge in power consumption, the system will simultaneously consider adjusting power supply parameters when image acquisition parameters need adjustment, ensuring that power supply capacity can support the new acquisition mode. When multiple components need adjustment simultaneously, the system prioritizes adjustment tasks based on correlation weight and designs a coordinated adjustment strategy to ensure that adjustments to key components drive synchronous optimization of other related components, forming an organic whole.

[0090] Ultimately, ensuring that the operational states of each stage are mutually compatible after adjustment, and avoiding new anomalies in other stages due to optimization of a single stage, is the ultimate goal and guarantee of the parameter adjustment mechanism. Through the aforementioned collaborative adjustment based on correlation weights, the system aims to eliminate the problems that may occur in the traditional independent optimization mode, namely, that optimization of one stage leads to new problems in another stage. It ensures the overall stability and reliability of the system. After parameter adjustment, the system will perform coupled verification on the adjusted operational state through a feature coupling model. If new anomalies or parameter deviations are found, the dynamic threshold self-learning model will be driven again for optimization, forming a closed-loop adaptive adjustment process.

[0091] Through the above technical solutions, this application clarifies the specific adjustment ranges of image acquisition parameters, power output parameters, and communication transmission parameters, and introduces correlation weights based on feature coupling correlation results to guide the magnitude and direction of parameter adjustments, effectively solving the problem of lack of clear definition and coordination guarantee in parameter adjustments in traditional methods. Specifically, the fine adjustment of image acquisition parameters enables the monitoring device to flexibly adapt to complex and changing ambient lighting and target characteristics, ensuring the acquisition of high-quality image data in different scenarios. The optimization of power output parameters ensures the stability of the monitoring device's energy supply and its endurance, providing reliable support for the continuous operation of each module. The adaptation of communication transmission parameters ensures that monitoring data can be transmitted back efficiently and reliably, avoiding data loss or delay due to changes in network conditions. More importantly, this application achieves deep cross-stage coordination by closely combining the magnitude and direction of parameter adjustments in each stage with the correlation weights in the feature coupling correlation results. When an anomaly occurs in a certain stage and adjustment is required, the system can identify other highly related stages based on the correlation weights and adjust the parameters of these related stages synchronously and in a coordinated manner. When communication transmission anomalies occur, the system not only adjusts communication parameters but also synchronously optimizes image acquisition parameters based on correlation weights to reduce data transmission volume and ensures priority power supply from the power supply module to the communication module. This avoids operational conflicts in other components that might arise from optimizing a single component. This collaborative adjustment mechanism based on correlation weights ensures the compatibility of the operational states of each component after adjustment, improving the overall stability, reliability, and adaptability to complex dynamic scenarios of the transmission line image monitoring system, and effectively reducing operation and maintenance costs and fault handling risks.

[0092] In some of the embodiments described above in this application, correlation verification and feature tracing analysis methods are proposed to analyze the correlation effects and positioning deviations after parameter adjustment. In the implementation process, the verification method may lack objective quantitative standards, and the tracing analysis may not be systematic enough, resulting in inaccurate positioning.

[0093] In response, this application further proposes that the correlation verification is implemented by calculating the coupling degree. When the coupling degree after adjusting the parameters of multiple links reaches the preset qualified standard, it is determined that the verification is passed; otherwise, it is determined that there is an abnormal correlation and triggers secondary control. The feature tracing analysis method is as follows: based on the multi-source feature coupling model, the correlation feature chain of abnormal feature values ​​is traced. By analyzing the change law and mutual influence of each feature in the correlation feature chain, the starting feature and core cause of the abnormality are located, providing accurate feedback basis for threshold iteration and model optimization.

[0094] Specifically, correlation verification is implemented using coupling degree calculation. Coupling degree calculation is a technique for quantifying the degree of interdependence or interaction strength between different components or parameters within a system. This calculation is implemented using various methods. Based on statistical methods, coupling degree is quantified by calculating the correlation coefficient or mutual information between parameter changes in different stages; high correlation or mutual information values ​​indicate strong coupling. Alternatively, based on a system dynamics model, a mathematical model describing the interaction of each stage is constructed, and the coupling degree is determined by analyzing the transfer function or sensitivity between variables in the model. Furthermore, a fuzzy comprehensive evaluation method is employed to comprehensively assess the coupling degree of each stage through expert scoring and fuzzy reasoning for factors that are difficult to quantify precisely. By introducing coupling degree calculation, an objective and quantitative standard is provided for evaluating the synergy and potential conflicts between stages after parameter adjustments, avoiding the bias of subjective judgment.

[0095] When the coupling degree after multi-stage parameter adjustments reaches the preset acceptable standard, the verification is considered successful; otherwise, it is considered to have an abnormal correlation and triggers secondary control. The preset acceptable standard is a threshold or range pre-set by the system to determine whether the coupling degree meets the requirements. This standard can be a single threshold, such as a coupling degree greater than 0.7 being considered acceptable, or it can be a range, such as a coupling degree between 0.6 and 0.8 being considered acceptable. This standard can be determined based on historical operational data analysis, expert experience, or simulation test results. When the calculated coupling degree is compared with this standard, if it does not meet the standard, the system will automatically invoke the preset secondary control strategy module. This module includes restarting the parameter optimization process, adjusting the priority of the control strategy, or issuing an alert to maintenance personnel and providing suggested adjustment solutions. This mechanism ensures the effectiveness of parameter adjustments and the stability of the system, enabling timely intervention and correction once potential abnormal correlations are detected.

[0096] Feature-based source tracing analysis specifically involves tracing the associated feature chains of anomalous feature values ​​based on a multi-source feature coupling model. An associated feature chain refers to a series of interconnected feature sequences or networks that, starting from an anomalous feature value, trace back to its source along the relationships between features within the multi-source feature coupling model. This tracing process is implemented using a graph traversal algorithm. The multi-source feature coupling model is represented as a directed graph, where nodes represent features and edges represent the relationships between features. When an anomalous feature value is detected, starting from that anomalous feature node, a depth-first search (DFS) or breadth-first search (BFS) algorithm is used to traverse back along the relationships, thereby constructing a complete associated feature chain. Another approach is based on a causal reasoning model, embedding causal relationships within the feature coupling model. When an anomaly occurs, reverse reasoning is performed using a causal graph or Bayesian network to identify the most likely direct and indirect causal features leading to the anomaly. This systematic tracing method can reveal the underlying causes behind anomalies, rather than merely remaining at the surface symptoms.

[0097] By analyzing the changing patterns and mutual influences of features in a chain of related features, the initial features and core causes of anomalies can be identified. Pattern analysis involves performing time-series analysis on the historical data of each feature in the chain to identify its trends, fluctuations, or abrupt changes before and after the anomaly occurs. For each feature in the chain, time-series analysis is performed, using moving averages, exponential smoothing, or more complex machine learning models to identify its trend or abrupt change before the anomaly. Mutual influence analysis assesses the causal strength and direction between features in the chain. Combining the established mapping relationships and weights in the feature coupling model, the causal strength and direction between features in the chain are evaluated. If a change in feature A leads to a change in feature B, and a change in feature B leads to an anomalous feature C, then feature A is likely the initial feature. By deeply analyzing the dynamic behavior of features and their interactions, the initial features and root causes of anomalies can be accurately located, providing a clear direction for subsequent optimization.

[0098] After identifying the initial features and core causes of anomalies, precise feedback is provided for threshold iteration and model optimization. This precise feedback means using the identified initial features and core causes as input information to guide further optimization of the dynamic threshold self-learning model and the feature coupling model. The identified core causes, as input, drive the dynamic threshold self-learning model to update its adaptive threshold system and recalculate or adjust the adaptive thresholds of relevant monitoring stages. Simultaneously, based on the identified initial features and core causes, the correlation logic and weights of relevant features in the feature coupling model are evaluated to determine if adjustments are needed, thereby improving the model's ability to identify and predict similar anomalies. This mechanism achieves closed-loop management of the optimization process, ensuring that the system can learn from each anomaly handling and continuously improve, enhancing its overall adaptability and robustness.

[0099] Through the above technical solutions, this application effectively solves the problems of inaccurate positioning caused by the lack of objective quantitative standards in correlation verification and insufficient systematic source analysis. Specifically, by using coupling degree calculation to verify the operating status after multi-stage parameter adjustments, an objective and quantitative evaluation standard is provided, avoiding the bias of subjective judgment and ensuring the reliability of verification results. When the coupling degree fails to meet the preset qualified standard, the system can automatically trigger secondary regulation to correct potential correlation anomalies in a timely manner, enhancing the system's responsiveness and stability. In addition, based on the multi-source feature coupling model, the system traces the associated feature chains of abnormal feature values ​​and deeply analyzes the changing patterns and mutual influences of each feature in the chain, enabling the systematic and accurate positioning of the initial features and core causes of anomalies. This precise source analysis provides clear and targeted feedback for the iterative update of the dynamic threshold self-learning model and the optimization of the feature coupling model, thereby improving the adaptation accuracy and overall operational reliability of the entire transmission line image monitoring optimization method, enabling it to maintain high efficiency and stable performance in complex and ever-changing monitoring scenarios.

[0100] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for optimizing image monitoring of power transmission lines, characterized in that, Includes the following steps: Construct and run a feature coupling model and a dynamic threshold self-learning model based on multi-source data perception. The feature coupling model receives multi-dimensional operational data during the transmission line monitoring process and outputs multi-source feature coupling correlation results. The dynamic threshold self-learning model receives the feature coupling correlation results, historical optimization data and monitoring scenario requirements, and outputs an adaptive threshold system for each monitoring link. The feature coupling correlation results and the adaptive threshold system are input into the cross-linkage linkage compensation mechanism. The cross-linkage linkage compensation mechanism performs the following operations to achieve collaborative optimization of monitoring parameters: real-time comparison of the matching status of the current multi-source operating data with the adaptive threshold system, identification of abnormal feature values ​​that exceed the threshold range, and synchronous correlation adjustment of image acquisition parameters, power supply output parameters and communication transmission parameters for the abnormal feature values, so as to achieve collaborative compensation for defects in each link. Based on the aforementioned feature coupling model, the operating status after parameter adjustment is coupled and verified to analyze the correlation and influence of parameter adjustment in each link, and to locate the parameter deviations and core causes that still exist after regulation. The parameter deviation and core cause are used as feedback to drive the dynamic threshold self-learning model to update the adaptive threshold system, while optimizing the feature association logic of the feature coupling model and improving the model adaptation accuracy. The optimized feature coupling model, dynamic threshold self-learning model, and cross-linkage compensation mechanism are integrated and deployed on the transmission line monitoring and operation platform. The integrated model is driven by real-time collected multi-dimensional operation data to achieve full-process early warning and data accumulation in the monitoring process. The multi-dimensional operational data includes at least environmental perception data, image acquisition data, power supply status data, communication status data, and equipment operating condition data. The adaptation threshold system covers at least image acquisition adaptation threshold, power supply stability threshold, and communication transmission qualification threshold. The historical optimization data includes at least historical parameter adjustment data, operational status verification data, deviation processing data, and early warning handling data.

2. The method for optimizing image monitoring of transmission lines according to claim 1, characterized in that: Build and run a feature coupling model based on multi-source data awareness, including: Multi-dimensional operational data is collected by the built-in sensing module of the power transmission line image monitoring device, and the multi-dimensional operational data is preprocessed to remove data noise and redundant information. Extract the core representational features of each dimension of the preprocessed data, establish the mapping relationship between features of different dimensions through feature correlation analysis, and construct a multi-source feature coupling model. The coupling features in the coupling model are weighted and determined based on the degree of influence of the features on the monitoring effect. The weights can be dynamically adjusted according to the actual monitoring scenario, and the multi-source feature coupling correlation results are output.

3. The method for optimizing image monitoring of transmission lines according to claim 1, characterized in that: Build and run a dynamic threshold self-learning model, including: Based on the gradient descent algorithm, the optimal fit range of each parameter in the historical optimization data is calculated iteratively to generate an initial fit threshold system. Establish a threshold update trigger mechanism to automatically trigger the threshold update process when the monitoring scenario changes or the accumulated deviation data reaches a preset level; Based on the feature coupling correlation results and the monitoring scenario requirements, the initial adaptation threshold system is dynamically adjusted to output an adaptation threshold system that is adapted to the current monitoring conditions.

4. The method for optimizing image monitoring of transmission lines according to claim 1, characterized in that: The feature coupling model receives multi-dimensional operational data during transmission line monitoring and outputs multi-source feature coupling correlation results. The dynamic threshold self-learning model receives the feature coupling correlation results, historical optimization data, and monitoring scenario requirements, and outputs an adapted threshold system, including: The preprocessed multi-dimensional operational data is input into the feature coupling model. Through feature mapping and weight calculation within the model, the multi-source feature coupling correlation result is output. The correlation result represents the intrinsic correlation between environmental features, equipment operation features, power supply features and communication features. The feature coupling and correlation results, historical optimization data, and monitoring scenario requirements are input into the dynamic threshold self-learning model. Through threshold iterative calculation and adaptation adjustment, an adaptation threshold system is output, which includes image acquisition adaptation threshold, power supply stability threshold, and communication transmission qualification threshold.

5. The method for optimizing image monitoring of transmission lines according to claim 1, characterized in that: Real-time comparison of the current multi-source operating data with the adapted threshold system to identify abnormal feature values ​​exceeding the threshold range, including: The feature values ​​of each dimension corresponding to the current multi-source running data are compared one by one with the corresponding threshold range in the adaptive threshold system; For each dimension's feature value, determine whether it falls within the corresponding threshold range. If the feature value exceeds the threshold range, mark it as an abnormal feature value and record the value, dimension, and occurrence time of the abnormal feature value. The marked abnormal feature values ​​are classified and summarized to form an abnormal feature list, which provides a basis for subsequent cross-linked linkage compensation and control.

6. The method for optimizing image monitoring of transmission lines according to claim 1, characterized in that: In response to the aforementioned abnormal feature values, the image acquisition parameters, power supply output parameters, and communication transmission parameters are adjusted synchronously and in conjunction, including: Based on the feature coupling correlation results, the correlation weight between the link to which the abnormal feature value belongs and other monitoring links is determined; Based on the priority of the associated weights, the operating parameters of the associated links are adjusted synchronously: if the communication transmission characteristic value is abnormal, the image acquisition parameters are optimized simultaneously to reduce the data transmission volume and ensure that the power supply module gives priority to the communication module; if the power supply characteristic value is abnormal, the power supply output parameters are adjusted simultaneously to optimize the power consumption parameters of image acquisition and communication transmission. The parameter adjustment process does not require any modification to the hardware structure of the monitoring device; parameter adaptation is achieved solely through software algorithms.

7. The method for optimizing image monitoring of transmission lines according to claim 1, characterized in that: Based on the aforementioned feature coupling model, the operational status after parameter adjustment is coupled and verified. The correlation and impact of parameter adjustments in each stage are analyzed to pinpoint the parameter deviations and core causes that still exist after regulation, including: Data is collected on the operational status of each stage after parameter adjustment to obtain multi-dimensional operational data after adjustment. The adjusted multi-dimensional operational data is input into the feature coupling model to verify the effect of a single link and the correlation of multiple links, and to determine whether there are any new abnormal feature values ​​or parameter correlation conflicts. The feature source analysis method is adopted to trace the initial features and fundamental factors that cause parameter deviations based on the feature correlation relationship in the feature coupling model. The fundamental factors include at least environmental interference factors, parameter correlation conflict factors, and threshold adaptation deviation factors.

8. The method for optimizing image monitoring of transmission lines according to claim 3, characterized in that: The learning rate of the gradient descent algorithm is dynamically adjusted based on historical optimization results. When the historical optimization deviation is small, the learning rate is reduced to ensure threshold stability; when the historical optimization deviation is large, the learning rate is increased to accelerate threshold adaptation. After the initial adaptation threshold system is constructed, it needs to be verified in a small-scale scenario before it can be applied to actual monitoring. Once the threshold update process is triggered, the parameter changes and adaptation effects before and after the threshold adjustment are recorded synchronously to form a threshold adjustment log, providing data support for subsequent model optimization.

9. The method for optimizing image monitoring of transmission lines according to claim 6, characterized in that: The image acquisition parameter adjustment includes at least resolution adjustment, frame rate adjustment, exposure adjustment, and night vision mode switching; the power output parameter adjustment includes at least power supply mode switching, output current adjustment, and charging / discharging strategy adjustment; the communication transmission parameter adjustment includes at least transmission link switching, data compression ratio adjustment, and transmission rate adaptation. The magnitude and direction of parameter adjustments in each stage are determined based on the correlation weights in the feature coupling correlation results. The higher the correlation weight of a stage, the stronger the synergy of parameter adjustments, ensuring that the operating states of each stage are compatible after adjustment, and avoiding new anomalies in other stages caused by optimization of a single stage.

10. The method for optimizing image monitoring of transmission lines according to claim 7, characterized in that: The correlation verification is implemented by coupling degree calculation. When the coupling degree after the adjustment of parameters of multiple links reaches the preset qualified standard, it is determined that the verification is passed; otherwise, it is determined that there is a correlation anomaly and triggers secondary control. The feature source tracing analysis method specifically involves tracing the associated feature chain of abnormal feature values ​​based on a multi-source feature coupling model. By analyzing the changing patterns and mutual influences of each feature in the associated feature chain, the starting feature and core cause of the abnormality can be located, providing accurate feedback for threshold iteration and model optimization.

Images