Rail robot inspection methods, systems, equipment, and media

Abstract

This application provides a method, system, equipment, and medium for track robot inspection, belonging to the field of intelligent inspection. The method includes: determining a target inspection mode from multiple inspection modes based on the safety level of the power equipment; the data collection order for different categories of data in the multiple inspection modes differs; collecting data from the power equipment in the target inspection mode to obtain multiple types of monitoring data; and inputting the multiple types of monitoring data into a pre-determined inspection model to determine the equipment status of the power equipment, including normal or abnormal status. The track robot inspection method, system, equipment, and medium provided in this application can more accurately grasp the status of power equipment, ensuring the accuracy and effectiveness of inspection.

Description

Technical Field

[0001] This application belongs to the field of intelligent inspection technology, and more specifically, it relates to a method, system, equipment, and medium for inspection of a track robot. Background Technology

[0002] In recent years, with the rapid development of technology, unmanned substations have been gradually promoted, which has greatly increased the workload of substation inspections, making it more difficult to guarantee the on-site and timely nature of manual inspections.

[0003] To address the aforementioned issues, track-mounted robots have emerged to replace manual labor in routine inspections of power equipment, significantly improving the efficiency and quality of power equipment inspections while reducing the labor intensity and risks for personnel. However, how track-mounted robots can effectively inspect power equipment remains a pressing problem to be solved. Summary of the Invention

[0004] The purpose of this application is to provide a method, system, equipment, and medium for track robot inspection, so as to enable track robots to inspect power equipment.

[0005] A first aspect of this application provides a method for inspecting a tracked robot, comprising:

[0006] The target inspection mode is determined from multiple inspection modes based on the safety level of the power equipment. The data collection order of different types of data collected by the multiple inspection modes is different. Different types of data include power data and image data. The types of power data collected by different inspection modes are different, and the resolution of the image data collected by different inspection modes is different.

[0007] Data is collected from power equipment based on the target inspection mode to obtain various types of monitoring data;

[0008] Multiple types of monitoring data are input into a pre-defined inspection model to determine the equipment status of power equipment, which includes normal or abnormal status.

[0009] A second aspect of this application provides a track robot inspection system, comprising:

[0010] The inspection mode determination module is used to determine the target inspection mode from multiple inspection modes based on the safety level of the power equipment. The data collection order of different types of data collected by multiple inspection modes is different. Different types of data include power data and image data. The types of power data collected by different inspection modes are different, and the resolution of the image data collected by different inspection modes is different.

[0011] The monitoring data acquisition module is used to collect data from power equipment based on the target inspection mode and obtain various types of monitoring data.

[0012] The equipment status determination module is used to input various types of monitoring data into a pre-defined inspection model to determine the equipment status of power equipment, which includes normal or abnormal status.

[0013] A third aspect of this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the above-described track robot inspection method.

[0014] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described track robot inspection method.

[0015] The beneficial effects of the track robot inspection method, system, equipment, and medium provided in this application embodiment are as follows:

[0016] In this application embodiment, different safety levels of power equipment correspond to different inspection modes. The order in which power data and image data are collected differs between different inspection modes, and the types of power equipment collected and the resolution of the collected image data also differ. In this application embodiment, the target inspection mode is determined by the safety level of the power equipment. Data can then be collected according to the order of power data and image data collected by the target inspection mode, as well as the types of power data and the resolution of the image data to be collected. Based on the collected data, the equipment status of the power equipment during the inspection process can be determined through a pre-determined inspection model, thereby realizing the inspection work of the power equipment by the track robot. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A schematic flowchart of a track robot inspection method provided in an embodiment of this application;

[0019] Figure 2 This is a flowchart illustrating the first inspection mode in an embodiment of this application.

[0020] Figure 3This is a flowchart illustrating the second inspection mode in an embodiment of this application.

[0021] Figure 4 This is a flowchart illustrating the third inspection mode in an embodiment of this application.

[0022] Figure 5 This is a flowchart illustrating the fourth inspection mode in an embodiment of this application.

[0023] Figure 6 This is a structural block diagram of a track robot inspection system provided in one embodiment of this application;

[0024] Figure 7 This is a schematic block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0025] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0026] To make the objectives, technical solutions, and advantages of this application clearer, the following description will be provided in conjunction with the accompanying drawings and specific embodiments.

[0027] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating a method for inspecting a tracked robot according to an embodiment of this application. The method may include:

[0028] S101: Based on the safety level of the power equipment, the target inspection mode is determined from multiple inspection modes. The data collection order for different categories of data collected by multiple inspection modes is different.

[0029] In the embodiments of this application, different types of data include power data and image data. The types of power data collected by different inspection modes are different, and the resolution of the image data collected by different inspection modes is different.

[0030] In this embodiment, after determining the safety level of the power equipment, the power equipment is then divided into multiple safety levels according to the relationship between the safety level and each preset level, such as Level 1 safety level, Level 2 safety level, Level 3 safety level and Special level safety level.

[0031] For example, in response to the power equipment's safety level being less than a first preset level but greater than a second preset level, corresponding to a first-level safety level, the target inspection mode is determined to be the first inspection mode; in response to the power equipment's safety level being less than or equal to the second preset level but greater than a third preset level, corresponding to a second-level safety level, the target inspection mode is determined to be the second inspection mode; in response to the power equipment's safety level being less than or equal to the third preset level, corresponding to a third-level safety level, the target inspection mode is determined to be the third inspection mode; in response to the power equipment's safety level being greater than or equal to the first preset level, corresponding to a special-level safety level, the target inspection mode is determined to be the fourth inspection mode. As another example, if a power equipment 1 has a safety level of 6, a first preset level of 10, a second preset level of 5, and a third preset level of 1, then the safety level of power equipment 1 (level 6) is less than the first preset level (level 10) and greater than the second preset level (level 5). Therefore, the safety level corresponding to power equipment 1 is the first-level safety level, and its corresponding inspection mode is the first inspection mode.

[0032] In this embodiment, different security levels are set, the core of which lies in the differences in the importance of power equipment, the classification of risk consequences, the optimal allocation of resources, and the corresponding limited and clear needs. Different power equipment plays different roles in the power grid, and the impact of core equipment failure is far greater than that of auxiliary equipment, requiring differentiated protection. At the same time, failure may lead to large-scale power outages, local anomalies, or only affect the inspection system, so it is necessary to match the prevention and control efforts according to the consequences in order to avoid waste of resources and clarify the inspection mode.

[0033] In this embodiment, the highest safety level can involve core power grid equipment, where a fault would lead to a large-scale power outage or a major safety accident; the first level of safety can involve important power transmission / distribution equipment, where a fault would affect local power supply; the second level of safety can involve auxiliary production equipment, where a fault would affect the local operation of the equipment but would not directly interrupt the power supply; and the third level of safety can involve peripheral auxiliary equipment, where a fault would only affect the auxiliary equipment itself without causing other damage.

[0034] For example, top-level safety equipment can be main transformers and ultra-high voltage circuit breakers in hub substations. For instance, the main transformers of hub substations with a voltage level of 220kV or above can be classified as top-level safety equipment, and their failures would cause large-scale power outages. Level 1 safety equipment can be switchgear and cable joints in regional substations. For instance, the switchgear in hub substations with a voltage level of 110kV or above can be classified as level 1 safety equipment, and its failures would affect local power supply. Level 2 safety equipment can be ventilation systems in power distribution rooms and low-voltage control cabinets. For instance, the ventilation systems in power distribution rooms are level 2 safety equipment, and their failures would only affect equipment heat dissipation. Level 3 safety equipment can be environmental sensors and indicator lights. For instance, indicator lights in equipment areas are level 3 safety equipment, and their failures would only affect visual cues and would not change the actual operating status of the equipment.

[0035] In this embodiment, the target inspection modes include: a first inspection mode, a second inspection mode, a third inspection mode, and a fourth inspection mode, with different inspection modes corresponding to different security levels. Specifically, the first inspection mode corresponds to the first security level, the second inspection mode corresponds to the second security level, the third inspection mode corresponds to the third security level, and the fourth inspection mode corresponds to the highest security level.

[0036] For example, the first inspection mode is an inspection mode that collects power data and image data of power equipment sequentially; the second inspection mode is an inspection mode that collects image data and power data of power equipment sequentially; and the third and fourth inspection modes are inspection modes that collect image data and power data of power equipment simultaneously.

[0037] For example, the first inspection mode collects more types of power data from the power equipment than the second inspection mode collects. The image data resolution of the power equipment collected by the first inspection mode is lower than that of the second inspection mode. The image data resolution of the power equipment collected by the third inspection mode is the same as that of the first inspection mode, and the types of power data collected are the same as those of the second inspection mode. The types of power data collected by the fourth inspection mode are the same as those of the first inspection mode, and the image data resolution of the collected data is the same as that of the second inspection mode.

[0038] S102: Data collection is performed on power equipment based on the target inspection mode to obtain various types of monitoring data.

[0039] In this embodiment, the various types of monitoring data include: image data and power data.

[0040] For example, image data may include: oil temperature gauges, hydraulic gauges, various oil level gauges, switch action count counters, and the appearance of equipment and facilities; power data may include at least one of: partial discharge data, winding temperature, insulation resistance, current, contact temperature, bus voltage, motor temperature, and ambient temperature.

[0041] For example, the multi-type monitoring data of the first inspection mode may include: first collecting power data such as partial discharge data, winding temperature, insulation resistance, current, contact temperature, bus voltage, motor temperature and ambient temperature of the power equipment, and then collecting low-resolution image data of the power equipment; the multi-type monitoring data of the second inspection mode may include: first collecting high-resolution image data of the power equipment, and then collecting power data such as current, contact temperature, bus voltage, motor temperature and ambient temperature of the power equipment; the multi-type monitoring data of the third inspection mode may include: simultaneously collecting low-resolution image data of the power equipment and power data such as current, contact temperature, bus voltage, motor temperature and ambient temperature of the power equipment; the multi-type monitoring data of the fourth inspection mode may include: simultaneously collecting power data such as partial discharge data, winding temperature, insulation resistance, current, contact temperature, bus voltage, motor temperature and ambient temperature of the power equipment and high-resolution image data of the power equipment.

[0042] S103: Input multiple types of monitoring data into a pre-determined inspection model to determine the equipment status of power equipment, including normal or abnormal status.

[0043] In this embodiment, the inspection model may include: an image data recognition model and a time-series data recognition model.

[0044] For example, image data can intuitively reflect explicit issues such as the appearance, structure, display values, or physical state of equipment. Furthermore, image data recognition models can adapt to image data from different devices through repeated learning, and can still stably output the equipment status (e.g., the probability of anomalies) of power equipment even under complex environments such as varying lighting and weather conditions, compensating for missed detections caused by visual fatigue during manual inspections. Power data has a temporal sequence; time-series data recognition models can extract features such as periodicity and abrupt changes by analyzing fluctuation patterns in continuous time series. Moreover, the characteristics of time-series data are more quantitative, objectively reflecting the deeper state of the equipment, thus allowing the determination of the equipment status (e.g., the probability of anomalies). In other words, image data recognition models can be used to identify image data collected from power equipment, and by analyzing this image data, the equipment status of the power equipment can be obtained; time-series data recognition models can be used to identify power data collected from power equipment, and by analyzing this power data, the equipment status of the power equipment can be determined.

[0045] For example, if both image data and power data indicate that the power equipment is in an abnormal state, then the power equipment is determined to be in an abnormal state. If both image data and power data indicate that the power equipment is in a normal state, then the power equipment is determined to be in a normal state. If both image data and power data indicate that the power equipment is in an abnormal state and the power equipment is in a normal state (e.g., the image data indicates that the power equipment is in an abnormal state, while the power data indicates that the power equipment is in a normal state), then the power equipment state can be determined based on the first probability of abnormality (the probability that the power equipment is in an abnormal state obtained from the power data) and the second probability of abnormality (the probability that the power equipment is in an abnormal state obtained from the image data).

[0046] As can be seen from the above, in this application embodiment, different safety levels of power equipment correspond to different inspection modes. The order of collecting power data and image data is different in different inspection modes, and the types of power equipment collected by different inspection modes are also different, as are the resolutions of the collected image data. In this application embodiment, the target inspection mode is determined by the safety level of the power equipment. Thus, data can be collected according to the order of collecting power data and image data corresponding to the target inspection mode, as well as the types of power data and the resolution of the image data to be collected. Based on the collected data, the equipment status of the power equipment during the inspection process can be determined by a pre-determined inspection model, thereby realizing the inspection work of the track robot on the power equipment.

[0047] In one embodiment of this application, multiple inspection modes include a first inspection mode and a second inspection mode. The first inspection mode sequentially collects power data and image data from the power equipment, while the second inspection mode sequentially collects image data and power data from the power equipment. The first inspection mode collects more types of power data than the second inspection mode, and the resolution of the image data collected by the first inspection mode is lower than that of the second inspection mode. In this embodiment, the image data may include: oil temperature gauges, hydraulic gauges, various oil level gauges, switch action counters, and the appearance of the equipment and facilities; the power data may include at least one of: partial discharge data, winding temperature, insulation resistance, current, contact temperature, bus voltage, motor temperature, and ambient temperature.

[0048] In this embodiment of the application, determining a target inspection mode from multiple inspection modes based on the safety level of the power equipment includes: determining the target inspection mode as the first inspection mode in response to the power equipment's safety level being less than a first preset level and greater than a second preset level; and determining the target inspection mode as the second inspection mode in response to the power equipment's safety level being less than or equal to the second preset level and greater than a third preset level; wherein the first preset level is greater than the second preset level, and the second preset level is greater than the third preset level.

[0049] As described in the above embodiments, the safety level of power equipment can be divided into four safety levels to determine the safety level corresponding to each power equipment, and then the inspection mode corresponding to each power equipment can be determined according to the safety level corresponding to each power equipment.

[0050] For example, based on different security levels, corresponding inspection modes are assigned. For power equipment in the first inspection mode, the impact of a fault is greater than that in the second inspection mode. Therefore, more power data of the power equipment needs to be acquired and assigned a higher weight to it in order to more accurately determine whether the power equipment in the first inspection mode is faulty. Conversely, for power equipment in the second inspection mode, rich power data is not required, but high image data is required. Therefore, high-resolution image data needs to be acquired for power equipment in the second inspection mode and assigned a higher weight to it in order to accurately determine whether the power equipment in the second inspection mode is faulty.

[0051] In this embodiment, the safety levels are divided according to a first preset level, a second preset level, and a third preset level. The safety levels include: Level 1 safety level, Level 2 safety level, Level 3 safety level, and Special Level safety level. The first preset level corresponds to the threshold between the Special Level safety level and the Level 1 safety level, and must be higher than the upper limit of the Level 1 safety level to ensure that Special Level equipment (such as the main transformer of a hub substation) is accurately identified due to the most serious risk consequences and matched with the highest standard fourth inspection mode. The second preset level is the threshold between the Level 1 safety level and the Level 2 safety level, and lies between the lower limit of the Level 1 safety level and the upper limit of the Level 2 safety level. This allows Level 1 equipment (such as switchgear of a regional substation) of the Level 1 safety level to correspond to the first inspection mode, prioritizing the monitoring of core functions. The third preset level is the threshold between the Level 2 safety level and the Level 3 safety level, and is lower than the lower limit of the Level 2 safety level and higher than the upper limit of the Level 3 safety level. This allows Level 2 equipment (such as ventilation systems) of the Level 2 safety level to adapt to the second inspection mode, balancing monitoring accuracy and resource investment.

[0052] For example, in response to the power equipment's safety level being less than a first preset level and greater than a second preset level, the power equipment's safety level is designated as Level 1, and the determined target inspection mode is designated as the first inspection mode; in response to the power equipment's safety level being less than or equal to the second preset level and greater than a third preset level, the power equipment's safety level is designated as Level 2, and the determined target inspection mode is designated as the second inspection mode; in response to the power equipment's safety level being less than or equal to the third preset level, the power equipment's safety level is designated as Level 3, and the determined target inspection mode is designated as the third inspection mode; in response to the power equipment's safety level being greater than or equal to the first preset level, the power equipment's safety level is designated as the highest safety level, and the determined target inspection mode is designated as the fourth inspection mode.

[0053] As can be seen from the above, the embodiments of this application, by differentiating the types of power data collected and image resolutions, enable the first inspection mode to focus on in-depth analysis of power data from high-impact equipment, and the second inspection mode to focus on image detail recognition of low-impact equipment, thereby achieving targeted optimization of precise resource allocation and equipment status judgment. Simultaneously, the three preset levels form a gradient through quantified risk levels, achieving precise matching between safety levels and inspection modes, providing a basis for subsequent accurate fault location and inspection.

[0054] In one embodiment of this application, the pre-determined inspection model includes an image data recognition model and a time-series data recognition model. A flowchart of the first inspection mode is shown below. Figure 2 As shown in this embodiment, in response to the target inspection mode being the first inspection mode, multiple types of monitoring data are input into a pre-determined inspection model to determine the equipment status of the power equipment. This includes: inputting the power data of the power equipment collected according to the first inspection mode into a time-series data recognition model to obtain a first time-series recognition result; the first time-series recognition result includes: power data status; if the power data status is abnormal, then inputting the power abnormality location label and the image data of the power equipment collected according to the first inspection mode into an image data recognition model to obtain a first image recognition result; the power abnormality location label is output by the time-series data recognition model when the power data status is abnormal; and determining the equipment status of the power equipment based on the first time-series recognition result and the first image recognition result.

[0055] In this embodiment, if the target inspection mode is the first inspection mode, the obtained multi-type monitoring data are input into a pre-determined inspection model to determine the equipment status of the power equipment. The power data of the power equipment collected in the first inspection mode is input into a pre-trained time-series data recognition model to obtain a first time-series recognition result, wherein the first time-series recognition result includes the power data status.

[0056] In this embodiment, the power data status can be characterized by the abnormal probability of power equipment in the first inspection mode.

[0057] For example, if the power equipment is in a normal state as indicated by the power data status, the image data of the power equipment collected according to the first inspection mode is input into a pre-trained image data recognition model to obtain the first image recognition result. Further, weights are assigned to the first time-series recognition result and the first image recognition result respectively to comprehensively determine the equipment status of the power equipment in the first inspection mode. For example, a higher weight is assigned to the first time-series recognition result and a lower weight is assigned to the first image recognition result. The equipment status of the power equipment in the first inspection mode is determined by combining the two.

[0058] Assuming that the probability of the power equipment being abnormal is 0.3 in the first time series result obtained from the power data (such as voltage, current, and temperature time series data) of the power equipment in the first inspection mode, and the probability of the power equipment being abnormal is 0.9 in the first image recognition result, then a higher weight (such as 0.7) is assigned to the first time series recognition result, and a lower weight (such as 0.3) is assigned to the first image recognition result. After combining the results, the probability of the power equipment being abnormal is judged to be 0.48. Therefore, the probability of the power equipment being abnormal is less than the preset threshold (0.7), which means that the power equipment is in a normal state.

[0059] For example, if the power data status is abnormal, the power abnormality location label and the image data of the power equipment collected according to the first inspection mode are input into the pre-trained image data recognition model to obtain the first image recognition result. The power abnormality location label is output by the time series data recognition model when the power data status is abnormal. The first time series recognition result is assigned a higher weight, and the first image recognition result is assigned a lower weight. The equipment status of the power equipment in the first inspection mode is determined by combining the two.

[0060] In cases where the power data status is abnormal, the first time-series identification result is directly derived from the core operating parameters of the power equipment (such as the characteristics of voltage, current, power and other changes over time). These data can essentially reflect the operating mechanism and root cause of the fault of the equipment, and are the fundamental basis for judging the abnormality of the equipment. The abnormal status has a clear direction and is decisive.

[0061] For example, suppose that the probability of the power equipment being abnormal is 0.9 in the first time series result obtained from the power data (such as voltage, current, and temperature time series data) of the power equipment in the first inspection mode, and the probability of the power equipment being abnormal is 0.3 in the first image recognition result. In this case, a higher weight (such as 0.7) is assigned to the first time series recognition result, and a lower weight (such as 0.3) is assigned to the first image recognition result. After combining the results, the probability of the power equipment being abnormal is judged to be 0.72. The probability of the power equipment being abnormal is greater than the preset threshold (0.7), which means that the power equipment is in an abnormal state.

[0062] Specifically, the power anomaly location tag includes: the type of abnormal component in the power anomaly, the spatial coordinates of the abnormal component, and the anomaly type corresponding to the abnormal component. In this embodiment, the power anomaly location tag and image data of the power equipment collected according to the first inspection mode are input into an image data recognition model to obtain a first image recognition result, including: determining the target abnormal component and each component associated with the target abnormal component based on the type of abnormal component in the power anomaly and the spatial coordinates of the abnormal component; determining the anomaly probability corresponding to each associated component based on the anomaly type corresponding to the abnormal component; performing semantic recognition on the image data of the power equipment to obtain the image region corresponding to each component in the power equipment; and determining the first target location corresponding to the target abnormal component. The system identifies the target image region and the second target image regions corresponding to each component with an anomaly probability greater than a preset probability threshold. It then extracts global and local texture features for each of the first and second target image regions. Next, it determines the third target image regions corresponding to each component with an anomaly probability not greater than a preset probability threshold, and extracts corresponding standard image features for each third target image region. These standard image features include morphological geometric features and position and pose features. Finally, it identifies the fourth target image regions corresponding to each component that is not associated with the target anomaly component. Contour features are extracted for each fourth target image region. Anomaly recognition is then performed based on global and local texture features, standard image features, and contour features to obtain the first image recognition result. In this embodiment, semantic recognition can be performed on the image data of the power equipment first to obtain the image regions corresponding to each component in the power equipment. Then, based on the abnormal component type of the power anomaly and the spatial coordinates of the abnormal component, the target abnormal component and the components associated with the target abnormal component can be determined. Alternatively, the target abnormal component and the components associated with the target abnormal component can be determined first based on the abnormal component type of the power anomaly and the spatial coordinates of the abnormal component, and then semantic recognition can be performed on the image data of the power equipment to obtain the image regions corresponding to each component in the power equipment. Of course, these can be performed simultaneously. In this embodiment, the specific execution order is not limited, and any possible execution order is within the protection scope of this embodiment.

[0063] In this embodiment, the power anomaly location tag is an identifier used to mark information related to power anomalies. It includes the type of abnormal component (such as the specific type of component that is abnormal, such as transformers or cable heads), the spatial coordinates of the abnormal component (which can accurately locate the abnormal component in the power system), and the type of anomaly (such as short circuit, leakage, overheating, etc.).

[0064] In this embodiment of the application, based on the abnormal component type of the power anomaly and the spatial coordinates of the abnormal component, the target abnormal component can be accurately determined. The target abnormal component refers to the component in the power equipment that is abnormal, obtained by detecting the power data of the power equipment through a time-series data identification model.

[0065] Furthermore, after obtaining the target abnormal component, based on the relationships between components in the power equipment, the various components associated with the target abnormal component can be determined. In the embodiments of this application, the various components associated with the target abnormal component can be various components in the power equipment connected to the target abnormal component, or various components in the power equipment that have a functional relationship with the target abnormal component. After determining the various components associated with the target abnormal component, since the abnormality type corresponding to the target abnormal component is different, the abnormality probability corresponding to each connected component is different, or the abnormality probability of each component that has a functional relationship with it is different.

[0066] For example, suppose the target abnormal component is a circuit breaker in a substation, and its abnormality type is short circuit. Components connected to the circuit breaker include disconnect switches, current transformers, and cables, while functionally related components include relay protection devices and busbars. When the abnormality type is short circuit, the disconnect switches and current transformers directly connected to the circuit breaker are more likely to be impacted by the short-circuit current due to their proximity to the fault point, with abnormal probabilities of 80% and 70%, respectively; while the cables, due to their slightly greater distance, have an abnormal probability of 40%. From a functional perspective, relay protection devices need to operate quickly during a short circuit, as excessive short-circuit current may cause them to malfunction or be damaged, resulting in an abnormal probability of 60%; although busbars are affected by short circuits, they usually have protective measures in place, resulting in an abnormal probability of 30%. If the circuit breaker's fault type changes to mechanical jamming (non-electrical fault), the connected disconnecting switch, lacking direct electrical impact, has a fault probability reduced to 20%; the relay protection device with functional relationship, not needing to handle short-circuit current, has a fault probability reduced to 10%, while the fault probability of the operating mechanism linkage (related component) associated with mechanical operation rises to 50%. This demonstrates that the fault probability of related components varies significantly depending on the fault type of the target faulty component.

[0067] In this embodiment of the application, semantic recognition is performed on the image data of the power equipment to obtain the image regions corresponding to each component of the power equipment. Specifically, this may include: preprocessing the image data of the power equipment, for example, adjusting the image size of the power equipment image (1920×1080) to 512×512 (maintaining the aspect ratio), normalizing it to [0,1], and outputting a 7-channel probability map based on the preprocessed image (each pixel corresponds to 7 classes of probability), taking the class with the highest probability for each pixel to obtain a semantic segmentation mask (512×512), and then resizing the mask back to the original image size (1920×1080) to obtain the image regions of each component.

[0068] As can be seen from the above embodiments, the abnormal concepts corresponding to each component and the image regions corresponding to each component can be obtained through the above embodiments, so as to determine the first target image region corresponding to the target abnormal component and the second target image region corresponding to each component with an abnormal probability greater than a preset probability threshold from the image data of the power equipment. Furthermore, global texture features and local texture features corresponding to each image region are extracted from the first target image region and each second target image.

[0069] Specifically, global texture features reflect the overall texture distribution of a region (such as roughness and regularity), and are extracted using the Gray-Level Co-occurrence Matrix (GLCM) and Global LBP (Local Binary Pattern). The GLCM reflects the texture's coarseness and directionality by statistically analyzing the co-occurrence frequencies of pixels with gray values ​​i and j at specific distances and angles. Global LBP generates binary codes by comparing the gray values ​​of a pixel with its neighboring pixels; the global statistical code distribution reflects the overall texture pattern. Specifically, local texture features focus on key details within a region (such as local wear and cracks), and are extracted using local LBP and SIFT feature point descriptors. In this embodiment, the local LBP extraction method may include: dividing the region into small windows, extracting LBP features for each window, and preserving local texture variations. The SIFT feature point descriptor extraction method may include: detecting key points (such as edges and corners) in the image and generating a 128-dimensional vector to describe the local texture and shape. This embodiment describes the overall texture pattern of transformer components using global texture features (GLCM, global LBP) and captures subtle changes using local texture features (local LBP, SIFT), achieving refined characterization of the texture features of the first target (such as the oil tank) and the second target (such as the heat sink and bushing). These features can be directly used for subsequent component status classification (such as normal / abnormal) or defect location tasks.

[0070] Furthermore, for each third target image region (the region corresponding to each component with an anomaly probability not greater than a preset probability threshold), each third target image region can be cropped from the image data of the power equipment. Each third target image region is preprocessed (binarization and contour extraction). Binarization: OTSU threshold segmentation (automatic threshold determination) is used to separate the target from the background (insulators are light-colored insulating materials, and terminals are dark metal). Contour extraction: Canny edge detection (threshold [50, 150]) is used to obtain the target edge contour and remove burrs (morphological closing operation, kernel size 3×3) to extract the morphological geometric features and position and pose features of the preprocessed image region to obtain the standard image features corresponding to each third target image region.

[0071] This embodiment focuses on the third target area (insulators and terminals), describing its shape and size specifications through morphological geometric features (such as roundness and number of corner points), and reflecting its spatial installation status through position and orientation features (such as center distance and rotation angle).

[0072] Furthermore, for each fourth target image region (the image region corresponding to each component that is not associated with the target abnormal component), the contour features corresponding to each fourth target image region are extracted. Taking the sleeve as an example, which is not associated with the target abnormal component, the method for extracting the contour features corresponding to the fourth target image region containing the sleeve is described. Specifically:

[0073] Step 1: Region preprocessing (ensuring contour accuracy), that is, firstly, based on the semantic segmentation results, crop the ROI region of the sleeve; then perform binarization and noise reduction processing. For example, since there may be reflections on the surface of the sleeve, convert the image into a binary image (the sleeve outline is white and the background is black). Remove the pseudo contours caused by surface stains through erosion operation (3×3 rectangular kernel), and connect the broken contour edges through dilation operation (3×3 rectangular kernel) to ensure contour continuity.

[0074] Step 2: Contour extraction and optimization, which involves extracting the outermost contour and retaining the longest contour (excluding small-area noise contours) to obtain the closed contour point set of the sleeve (contour, in the format of an N×1×2 pixel coordinate array, where N is the number of contour points). Then, the Douglas-Peucker algorithm is used to simplify the contour, retaining key structural points, thereby reducing the original 1500 contour points to 300, while retaining key features such as inflection points and curvature change points.

[0075] As can be seen from the above, the embodiments of this application, through detailed analysis of power anomaly location tags and combined with multi-dimensional feature extraction and analysis of image data, can accurately identify abnormal conditions of power equipment. This not only improves the accuracy and comprehensiveness of anomaly identification, but also clarifies the status of abnormal components and their related components, providing accurate and reliable basis for the maintenance and repair of power systems. It also helps to promptly identify potential faults and ensure the safe and stable operation of power systems.

[0076] Further, after obtaining global texture features, local texture features, standard image features, and contour features, anomaly recognition is performed based on these features to obtain a first image recognition result. In this embodiment, anomaly recognition based on global texture features, local texture features, standard image features, and contour features to obtain the first image recognition result includes: determining the anomaly probability corresponding to each image region based on the anomaly probability corresponding to the target anomaly component and each associated component; determining the weight corresponding to each image region based on the anomaly probability corresponding to each image region; and performing anomaly recognition based on the global texture features, local texture features, standard image features, contour features, and the weights corresponding to each image region to obtain the first image recognition result.

[0077] In this embodiment, the anomaly probability corresponding to each image region refers to the probability value of an anomaly in the area occupied by the target abnormal component and its associated components in the image. The weight corresponding to each image region is determined based on the anomaly probability of each image region and is used to represent the importance of the feature of that region in the anomaly identification process. The higher the anomaly probability, the greater the corresponding weight. Anomaly identification based on multiple features and weights comprehensively considers global texture features, local texture features, standard image features, contour features, and the weight of each image region. It can determine whether there is an anomaly in the power equipment through certain algorithms or models.

[0078] For example, the target anomalous component can be a "cable connector," and associated components can be a "connected cable" and an "insulating bracket," with anomalous probabilities of 60% and 30%, respectively. Correspondingly, the anomalous probability of the first target image region of the "cable connector" can be 100%, the anomalous probability of the second target image region of the "connected cable" can be 60%, the anomalous probability of the third target image region of the "insulating bracket" can be 30%, and the anomalous probability of the fourth target image region of the "distant distribution cabinet," which is unrelated to the target anomalous component, can be 0%. Based on these anomalous probabilities, the weights of each image region are determined. Assuming a positive correlation between anomalous probability and weight, the weights of the first target image region can be set to 0.4, the second target image region to 0.3, the third target image region to 0.2, and the fourth target image region to 0.1. Then, the extracted global and local texture features of the "cable connector," the global and local texture features of the "connected cable," the standard image features of the "insulating bracket," and the contour features of the "distant distribution cabinet" are multiplied by their respective weights and input into the anomaly recognition model. The model integrates these weighted features for calculation and analysis, and finally obtains the first image recognition result.

[0079] As can be seen from the above, the embodiments of this application, by determining the anomaly probability of each image region and assigning corresponding weights, enable the focus to be placed on regions with high anomaly probabilities during the anomaly identification process, while not ignoring other regions, thus improving the targeting and accuracy of anomaly identification. This multi-feature fusion identification method combined with weights can more efficiently utilize the effective information in image data and reduce the interference of irrelevant information, thereby providing more reliable results for power equipment inspection, helping to formulate more accurate maintenance plans, and reducing the risk of power system failures.

[0080] As can be seen from the above, for power equipment in the first inspection mode, the embodiments of this application prioritize the power data recognition results as the core and the image data recognition results as the auxiliary verification, and assign corresponding weights, which can accurately focus on the abnormal power state affecting the equipment and improve the accuracy and pertinence of fault judgment.

[0081] In one embodiment of this application, the flowchart of the second inspection mode is as follows: Figure 3As shown, the pre-determined inspection model includes an image data recognition model and a time-series data recognition model. In response to the target inspection mode being the second inspection mode, multiple types of monitoring data are input into the pre-determined inspection model to determine the equipment status of the power equipment. This includes: inputting image data of the power equipment collected according to the second inspection mode into the image data recognition model to obtain a second image recognition result; the second image recognition result includes: image data status; if the image data status is abnormal, then inputting the image abnormality location label and the power data of the power equipment collected according to the second inspection mode into the time-series data recognition model to obtain a second time-series recognition result; the image abnormality location label is output by the image data recognition model when the image data status is abnormal; and determining the equipment status of the power equipment based on the second image recognition result and the second time-series recognition result.

[0082] In this embodiment, if the target inspection mode is the second inspection mode, the obtained multi-type monitoring data are input into a pre-determined inspection model to determine the equipment status of the power equipment, which is characterized by the anomaly probability of the power equipment. The image data of the power equipment collected in the second inspection mode is input into a pre-trained image data recognition model to obtain a second image recognition result, wherein the second image recognition result includes the image data status, which is characterized by the anomaly probability of the power equipment.

[0083] For example, if the image data status is normal, the power data of the power equipment collected according to the second inspection mode will be input into the pre-trained time-series data recognition model to obtain the second time-series recognition result; a higher weight will be assigned to the second image recognition result and a lower weight will be assigned to the second time-series recognition result, and the equipment status of the power equipment in the second inspection mode will be determined by combining the two.

[0084] For example, if the image data status is abnormal, the image abnormality location label and the power data of the power equipment collected according to the second inspection mode are input into a pre-trained time-series data recognition model to obtain a second time-series recognition result. The image abnormality location label is output by the image data recognition model when the image data status is abnormal. A higher weight is assigned to the second image recognition result, and a lower weight is assigned to the second time-series recognition result. The equipment status of the power equipment under the second inspection mode is determined by combining the two. In this embodiment, the image abnormality location label and the power data of the power equipment collected according to the second inspection mode are input into the pre-trained time-series data recognition model for recognition, so that the model can focus on the components identified by the image abnormality location label when recognizing based on the power data, thereby obtaining the second time-series recognition result.

[0085] As can be seen from the above, for power equipment under the second inspection mode, the embodiments of this application take image data recognition results as the core, power data recognition results as auxiliary verification and assign corresponding weights, which can accurately focus on the abnormal image state affecting the equipment and improve the accuracy and efficiency of fault judgment.

[0086] In one embodiment of this application, the flowchart of the third inspection mode is as follows: Figure 4 As shown, the pre-determined inspection model includes an image data recognition model and a time-series data recognition model. Inputting multiple types of monitoring data into the pre-determined inspection model to determine the equipment status of the power equipment further includes: responding to the target inspection mode being a third inspection mode, inputting multiple types of monitoring data into the pre-determined inspection model to determine the equipment status of the power equipment, including inputting image data of the power equipment collected according to the third inspection mode into the image data recognition model to obtain a third image recognition result; simultaneously, inputting power data of the power equipment collected according to the third inspection mode into the time-series data recognition model to obtain a third time-series recognition result; wherein the third image recognition result includes image data status, and the third time-series recognition result includes power data status, wherein the resolution of the power equipment image data in the third inspection mode is the same as the resolution of the power equipment image data in the first inspection mode, and the data types of the power equipment are the same as the data types in the second inspection mode.

[0087] For example, the image data status and power data status in the third image recognition result and the third time-series recognition result are directly assigned average weights, and the equipment status of the power equipment is determined based on the final result.

[0088] In one embodiment of this application, the flowchart of the fourth inspection mode is as follows: Figure 5As shown, the pre-determined inspection model includes an image data recognition model and a time-series data recognition model. Inputting multiple types of monitoring data into the pre-determined inspection model to determine the equipment status of the power equipment further includes: responding to the target inspection mode being the fourth inspection mode, inputting multiple types of monitoring data into the pre-determined inspection model to determine the equipment status of the power equipment. This includes inputting image data of the power equipment collected according to the fourth inspection mode into the image data recognition model to obtain a fourth image recognition result; simultaneously, inputting power data of the power equipment collected according to the fourth inspection mode into the time-series data recognition model to obtain a fourth time-series recognition result. The fourth image recognition result includes image data status, and the fourth time-series recognition result includes power data status. The types of power equipment data in the fourth inspection mode are the same as those in the first inspection mode, and the resolution of the power equipment image data is the same as that in the second inspection mode. The image data status and power data status in the fourth image recognition result and the fourth time-series recognition result are directly assigned average weights, and the equipment status of the power equipment is determined based on the final result.

[0089] As can be seen from the above, for the third and fourth inspection modes, the embodiments of this application assign equal weight to the recognition results of image data and power data, thereby achieving comprehensive coverage of the status judgment of the two types of equipment. This not only ensures balanced monitoring of the basic status of the three-level equipment, but also ensures comprehensive evaluation of the electrical and image status of the special-level equipment, thus improving the comprehensiveness and reliability of the status judgment of equipment at different safety levels.

[0090] In one embodiment of this application, after determining the equipment status of the power equipment, the method further includes: when the equipment status of the power equipment is abnormal, inputting multiple types of monitoring data into a pre-determined graded early warning model to obtain the degree of abnormality of the power equipment, and performing graded early warning based on the degree of abnormality of the power equipment. The degree of abnormality includes minor abnormality, general abnormality, severe abnormality, and emergency abnormality; the graded early warning includes Level 1, Level 2, Level 3, and Level 4 early warnings, and the degree of abnormality and the graded early warning are positively correlated.

[0091] In this embodiment, abnormal states are divided into four levels of abnormality, which are positively correlated with the graded early warning.

[0092] For example, if the equipment has minor defects that do not affect operation, such as insulation resistance slightly higher than the standard value or minor stains on the equipment's appearance, these abnormalities are classified as minor abnormalities, corresponding to Level 1 warnings. If the equipment exhibits certain abnormalities that require attention and planned maintenance, such as winding temperatures occasionally exceeding the normal range or slight deviations in oil temperature gauge readings, these abnormalities are classified as general abnormalities, corresponding to Level 2 warnings. If the equipment abnormalities are significant and may affect operational safety, such as frequent exceedances of partial discharge data or continuous increases in contact temperature approaching critical values, these abnormalities are classified as serious abnormalities, corresponding to Level 3 warnings. If the equipment has a serious fault that requires immediate shutdown, such as short-circuit current, a sudden and irrecoverable drop in bus voltage, or obvious damage and oil leakage on the equipment's appearance, these abnormalities are classified as emergency abnormalities, corresponding to Level 4 warnings.

[0093] In this embodiment, the pre-determined graded early warning model is based on multiple types of monitoring data under abnormal power equipment conditions. By integrating and analyzing data characteristics and the degree of abnormality, the model automatically determines and outputs the corresponding early warning level according to the preset early warning level classification rules.

[0094] For example, for a Level 1 warning of minor anomalies, a blue "Minor Anomaly" label can be displayed on the device status page without any audible alert. Level 1 warnings are only recorded in the system background and pushed to the daily inspection list, without actively displaying a pop-up window. For a Level 2 warning of general anomalies, a yellow flashing border indicating "General Anomaly" can be displayed next to the device name without any audible alert. Only a low-frequency system beep is emitted when the warning is triggered, and a non-forced pop-up window appears on the maintenance personnel's system workbench, automatically shrinking to the notification bar after 3 seconds. For a Level 3 warning of severe anomalies, it can be prominently displayed at the top of the main monitoring interface. The system displays an orange animated icon with the red text "Serious Anomaly," accompanied by an intermittent medium-frequency beep every 10 seconds, and a pop-up window forcibly displays details of the anomaly, which can only be closed by the maintenance manager clicking "Read." For Level 4 emergency anomalies, a flashing red warning box covers the core area of ​​the monitoring interface and can display eye-catching text such as "Emergency Anomaly! Handle Immediately!" while emitting a continuous high-frequency alarm sound until manually closed. Simultaneously, the system pushes the highest priority information, including the anomaly location, real-time data, and other key information, to the maintenance manager and the emergency command center, and automatically triggers emergency plan prompts.

[0095] As can be seen from the above, this embodiment of the application classifies abnormal states into four levels and automatically classifies them using a hierarchical early warning model, achieving four levels of early warning. Differentiated alerts are implemented based on the severity and scope of the equipment abnormality, coupled with differentiated visual, auditory, and information push methods: minor abnormalities are only recorded, general abnormalities prompt planned maintenance, serious abnormalities emphasize safety risks, and emergency abnormalities mandate immediate handling. This hierarchical approach avoids minor issues disrupting normal work processes while ensuring timely attention and priority response to serious and emergency faults, allowing maintenance personnel to quickly grasp the priority of handling. This significantly improves the accuracy of power equipment abnormality response, fault handling efficiency, and equipment safety assurance level, providing a more reliable guarantee for safe equipment operation.

[0096] Corresponding to the track robot inspection method in the above embodiment, Figure 6 This is a structural block diagram of a track-mounted robot inspection system provided in one embodiment of this application. For ease of explanation, only the parts relevant to the embodiment of this application are shown. References Figure 6 The track robot inspection system 20 includes: an inspection mode determination module 21, a monitoring data acquisition module 22, and an equipment status determination module 23.

[0097] Among them, the inspection mode determination module 21 is used to determine the target inspection mode from multiple inspection modes based on the safety level of the power equipment. The data collection order of different types of data collected by multiple inspection modes is different. Different types of data include power data and image data. The types of power data collected by different inspection modes are different, and the resolution of the image data collected by different inspection modes is different.

[0098] The monitoring data acquisition module 22 is used to collect data from power equipment based on the target inspection mode and obtain multiple types of monitoring data.

[0099] The device status determination module 23 is used to input multiple types of monitoring data into a pre-determined inspection model to determine the device status of the power equipment, which includes normal or abnormal status.

[0100] In one embodiment of this application, the multiple inspection modes include a first inspection mode and a second inspection mode; the first inspection mode is an inspection mode that sequentially collects power data and image data of the power equipment, and the second inspection mode is an inspection mode that sequentially collects image data and power data of the power equipment.

[0101] Among them, the first inspection mode collects more types of power data from power equipment than the second inspection mode, and the resolution of the image data collected by the first inspection mode is lower than that of the image data collected by the second inspection mode.

[0102] The inspection mode determination module 21 determines the target inspection mode from multiple inspection modes based on the safety level of the power equipment. Specifically, it is used to: determine the target inspection mode as the first inspection mode in response to the power equipment's safety level being less than the first preset level and greater than the second preset level; and determine the target inspection mode as the second inspection mode in response to the power equipment's safety level being less than or equal to the second preset level and greater than the third preset level.

[0103] Among them, the first preset level is greater than the second preset level, and the second preset level is greater than the third preset level.

[0104] In one embodiment of this application, the device status determination module 23 is specifically used for: the pre-determined inspection model includes an image data recognition model and a time-series data recognition model;

[0105] In response to the target inspection mode being the first inspection mode, multiple types of monitoring data are input into a pre-defined inspection model to determine the equipment status of the power equipment, including:

[0106] The power data of the power equipment collected according to the first inspection mode is input into the time series data recognition model to obtain the first time series recognition result; the first time series recognition result includes: power data status;

[0107] If the power data status is abnormal, the power abnormality location tag and the image data of the power equipment collected according to the first inspection mode are input into the image data recognition model to obtain the first image recognition result; the power abnormality location tag is output by the time series data recognition model when the power data status is abnormal.

[0108] The equipment status of the power equipment is determined based on the first time sequence recognition result and the first image recognition result.

[0109] In one embodiment of this application, the device status determination module 23 is further configured to: the pre-determined inspection model includes an image data recognition model and a time-series data recognition model;

[0110] In response to the target inspection mode being the second inspection mode, multiple types of monitoring data are input into a pre-defined inspection model to determine the equipment status of the power equipment, including:

[0111] The image data of the power equipment collected according to the second inspection mode is input into the image data recognition model to obtain the second image recognition result; the second image recognition result includes: image data status;

[0112] If the image data status is abnormal, the abnormal image location label and the power data of the power equipment collected according to the second inspection mode are input into the time series data recognition model to obtain the second time series recognition result; the abnormal image location label is output by the image data recognition model when the image data status is abnormal.

[0113] The equipment status of the power equipment is determined based on the second image recognition result and the second time sequence recognition result.

[0114] In one embodiment of this application, the power anomaly location tag includes: the abnormal component type of the power anomaly, the spatial coordinates of the abnormal component, and the anomaly type corresponding to the abnormal component;

[0115] The device status determination module 23 inputs the power anomaly location tag and the image data of the power equipment collected according to the first inspection mode into the image data recognition model to obtain the first image recognition result, which is specifically used for:

[0116] Based on the abnormal component type of the power anomaly and the spatial coordinates of the abnormal component, the target abnormal component and the components associated with the target abnormal component are determined.

[0117] Based on the anomaly type corresponding to the abnormal component, determine the anomaly probability corresponding to each associated component;

[0118] Semantic recognition is performed on the image data of power equipment to obtain the image regions corresponding to each component of the power equipment.

[0119] Determine the first target image region corresponding to the abnormal target component and the second target image region corresponding to each component whose abnormal probability is greater than a preset probability threshold;

[0120] Extract global and local texture features corresponding to each image region of the first target image region and each second target image region;

[0121] The third target image region corresponding to each component with an abnormal probability not greater than a preset probability threshold is determined, and the standard image features corresponding to each third target image region are extracted. The standard image features include: morphological geometric features and position and pose features.

[0122] Identify the fourth target image regions corresponding to each component that is not associated with the target abnormal component;

[0123] Extract the corresponding contour features for each fourth target image region;

[0124] Anomaly recognition is performed based on global and local texture features, standard image features, and contour features to obtain the first image recognition result.

[0125] In one embodiment of this application, when the device status determination module 23 performs anomaly recognition based on global texture features, local texture features, standard image features, and contour features to obtain a first image recognition result, it is specifically used for:

[0126] Based on the target abnormal component and the abnormal probability corresponding to each associated component, the abnormal probability corresponding to each image region is determined.

[0127] Based on the anomaly probability corresponding to each image region, the weight corresponding to each image region is determined.

[0128] Anomaly identification is performed based on global and local texture features, standard image features and contour features, as well as the weights corresponding to each image region, to obtain the first image recognition result.

[0129] See Figure 7 , Figure 7 This is a schematic block diagram of an electronic device provided according to an embodiment of this application. Figure 7 The electronic device 300 in this embodiment may include one or more processors 301, one or more input devices 302, one or more output devices 303, and one or more memories 304. The processors 301, input devices 302, output devices 303, and memories 304 communicate with each other via a communication bus 305. The memories 304 store computer programs, including program instructions. The processors 301 execute the program instructions stored in the memories 304. Specifically, the processors 301 are configured to invoke the program instructions to perform the functions of the modules in the aforementioned device embodiments, for example... Figure 6 The functions of the inspection mode determination module 21, the monitoring data acquisition module 22, and the equipment status determination module 23.

[0130] It should be understood that, in the embodiments of this application, the processor 301 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0131] Input device 302 may include a touchpad, a fingerprint sensor (for collecting the user's fingerprint information and fingerprint orientation information), a microphone, etc., and output device 303 may include a display (LCD, etc.), a speaker, etc.

[0132] The memory 304 may include read-only memory and random access memory, and provides instructions and data to the processor 301. A portion of the memory 304 may also include non-volatile random access memory. For example, the memory 304 may also store device type information.

[0133] In specific implementations, the processor 301, input device 302, and output device 303 described in the embodiments of this application can execute the implementation method described in the track robot inspection method provided in the embodiments of this application, or they can execute the implementation method of the electronic device described in the embodiments of this application, which will not be repeated here.

[0134] In another embodiment of this application, a computer-readable storage medium is provided. This computer-readable storage medium stores a computer program, which includes program instructions. When executed by a processor, the program instructions implement all or part of the processes in the methods described above. Alternatively, the computer program can instruct related hardware to complete the process. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include any entity or device capable of carrying computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0135] The computer-readable storage medium can be an internal storage unit of the electronic device in any of the foregoing embodiments, such as a hard disk or memory of the electronic device. The computer-readable storage medium can also be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., provided on the electronic device. Furthermore, the computer-readable storage medium can include both internal and external storage units of the electronic device. The computer-readable storage medium is used to store computer programs and other programs and data required by the electronic device. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0136] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0137] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the electronic devices and units described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0138] In the several embodiments provided in this application, it should be understood that the disclosed electronic devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules, units, or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces or modules, or it may be an electrical, mechanical, or other form of connection.

[0139] The modules described as separate components may or may not be physically separate. Similarly, the components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of the embodiments of this application, depending on actual needs.

[0140] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0141] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for inspecting rail-mounted robots, characterized in that, include: The target inspection mode is determined from multiple inspection modes based on the safety level of the power equipment. The multiple inspection modes collect different types of data in different orders. The different types of data include power data and image data. The types of power data collected by different inspection modes are different, and the resolution of the image data collected by different inspection modes is different. The multiple inspection modes include a first inspection mode and a second inspection mode; The first inspection mode is an inspection mode that sequentially collects power data and image data of the power equipment, and the second inspection mode is an inspection mode that sequentially collects image data and power data of the power equipment; wherein, the first inspection mode collects more types of power data than the second inspection mode collects power data, and the resolution of the image data of the power equipment collected by the first inspection mode is lower than the resolution of the image data of the power equipment collected by the second inspection mode; The target inspection mode is determined from multiple inspection modes based on the safety level of the power equipment, including: In response to the power equipment's safety level being less than a first preset level and greater than a second preset level, the target inspection mode is determined to be the first inspection mode; in response to the power equipment's safety level being less than or equal to the second preset level and greater than a third preset level, the target inspection mode is determined to be the second inspection mode; wherein, the first preset level is greater than the second preset level, and the second preset level is greater than the third preset level; Data is collected from the power equipment based on the target inspection mode to obtain various types of monitoring data. The various types of monitoring data are input into a pre-determined inspection model to determine the equipment status of the power equipment, which includes normal or abnormal status.

2. The track robot inspection method as described in claim 1, characterized in that, The predetermined inspection model includes an image data recognition model and a time-series data recognition model; In response to the target inspection mode being the first inspection mode, the step of inputting the multiple types of monitoring data into a pre-determined inspection model to determine the equipment status of the power equipment includes: The power data of the power equipment collected according to the first inspection mode is input into the time series data recognition model to obtain the first time series recognition result; the first time series recognition result includes: power data status; If the power data status is abnormal, the power abnormality location tag and the image data of the power equipment collected according to the first inspection mode are input into the image data recognition model to obtain the first image recognition result; the power abnormality location tag is output by the time series data recognition model when the power data status is abnormal. The device status of the power equipment is determined based on the first time sequence recognition result and the first image recognition result.

3. The track robot inspection method as described in claim 1, characterized in that, The predetermined inspection model includes an image data recognition model and a time-series data recognition model; In response to the target inspection mode being the second inspection mode, the step of inputting the multiple types of monitoring data into a pre-determined inspection model to determine the equipment status of the power equipment includes: The image data of the power equipment collected according to the second inspection mode is input into the image data recognition model to obtain the second image recognition result; the second image recognition result includes: image data status; If the image data status is abnormal, the abnormal image location label and the power data of the power equipment collected according to the second inspection mode are input into the time series data recognition model to obtain the second time series recognition result; the abnormal image location label is output by the image data recognition model when the image data status is abnormal. The device status of the power equipment is determined based on the second image recognition result and the second time sequence recognition result.

4. The track robot inspection method as described in claim 1, characterized in that, After determining the equipment status of the power equipment, the process also includes: When the power equipment is in an abnormal state, the multi-type monitoring data is input into a pre-determined graded early warning model to obtain the degree of abnormality of the power equipment, and a graded early warning is given based on the degree of abnormality of the power equipment. The degree of abnormality includes minor abnormality, general abnormality, severe abnormality and emergency abnormality; the graded warning includes level one warning, level two warning, level three warning and level four warning, and the degree of abnormality and the graded warning are positively correlated.

5. The track robot inspection method as described in claim 2, characterized in that, The power anomaly location tag includes: the abnormal component type of the power anomaly, the spatial coordinates of the abnormal component, and the anomaly type corresponding to the abnormal component; The step of inputting the power anomaly location tag and the image data of the power equipment collected according to the first inspection mode into the image data recognition model to obtain the first image recognition result includes: Based on the abnormal component type of the power anomaly and the spatial coordinates of the corresponding abnormal component, the target abnormal component and each component associated with the target abnormal component are determined. Based on the anomaly type corresponding to the abnormal component, determine the anomaly probability corresponding to each of the associated components; Semantic recognition is performed on the image data of the power equipment to obtain the image regions corresponding to each component of the power equipment. Determine the first target image region corresponding to the target abnormal component and the second target image region corresponding to each component whose abnormal probability is greater than a preset probability threshold; Extract global and local texture features corresponding to each image region of the first target image region and each of the second target images; The third target image region corresponding to each component with an abnormal probability not greater than a preset probability threshold is determined, and standard image features corresponding to each third target image region are extracted. The standard image features include: morphological geometric features and position and pose features. Identify the fourth target image regions corresponding to each component that is not associated with the target abnormal component; Extract the corresponding contour features for each of the fourth target image regions; Anomaly recognition is performed based on the global and local texture features, the standard image features, and the contour features to obtain the first image recognition result.

6. The method according to claim 5, characterized in that, The anomaly identification based on the global texture features, local texture features, standard image features, and contour features to obtain a first image recognition result includes: Based on the target abnormal component and the abnormal probability corresponding to each of the associated components, the abnormal probability corresponding to each image region is determined. Based on the anomaly probability corresponding to each image region, the weight corresponding to each image region is determined. Based on the global and local texture features, the standard image features and the contour features, and the weights corresponding to each image region, anomaly identification is performed to obtain the first image recognition result.

7. A track robot inspection system, characterized in that, include: The inspection mode determination module is used to determine the target inspection mode from multiple inspection modes based on the safety level of the power equipment. The multiple inspection modes collect different types of data in different orders. The different types of data include power data and image data. The types of power data collected by different inspection modes are different, and the resolution of the image data collected by different inspection modes is different. The multiple inspection modes include a first inspection mode and a second inspection mode; The first inspection mode is an inspection mode that sequentially collects power data and image data of the power equipment, and the second inspection mode is an inspection mode that sequentially collects image data and power data of the power equipment; wherein, the first inspection mode collects more types of power data than the second inspection mode collects power data, and the resolution of the image data of the power equipment collected by the first inspection mode is lower than the resolution of the image data of the power equipment collected by the second inspection mode; The target inspection mode is determined from multiple inspection modes based on the safety level of the power equipment, including: In response to the power equipment's safety level being less than a first preset level and greater than a second preset level, the target inspection mode is determined to be the first inspection mode; in response to the power equipment's safety level being less than or equal to the second preset level and greater than a third preset level, the target inspection mode is determined to be the second inspection mode; wherein, the first preset level is greater than the second preset level, and the second preset level is greater than the third preset level; The monitoring data acquisition module is used to collect data from the power equipment based on the target inspection mode and obtain multiple types of monitoring data. The device status determination module is used to input the various types of monitoring data into a pre-determined inspection model to determine the device status of the power equipment, which includes normal status or abnormal status.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 6.

Images