Leakage hot area space-time clustering method and system based on inspection robot

Abstract

The invention provides a leakage hot area spatio-temporal clustering method and system based on an inspection robot, and relates to the technical field of pipeline detection, and the method comprises the steps: obtaining time sequence multi-source collection data of a pipe network through the inspection robot carrying a multi-modal sensor, carrying out the extraction of a time sequence and spatial feature relation, and constructing a spatio-temporal data grid; performing data feature association analysis based on the spatio-temporal data grids, performing multi-granularity aggregation on time and space according to association features, determining spatio-temporal relationship features among the grids, performing leakage hot area identification, performing visual conversion in the spatio-temporal data grids, and generating identification visual information. The technical problem that the pipe network leakage risk identification accuracy and reliability are not high due to the lack of comprehensive judgment capability for the pipe network state in the prior art is solved. The technical effects of comprehensively discriminating and analyzing the state of the pipe network and improving the accuracy and reliability of identifying the leakage hot area of the pipe network by introducing the multi-mode sensing data of the inspection robot are achieved.

Description

Technical Field

[0001] This application relates to the field of pipeline inspection technology, specifically to a spatiotemporal clustering method and system for leakage hot zones based on an inspection robot. Background Technology

[0002] With the continuous expansion of urban areas and the high density of underground infrastructure, water supply, heating, and gas pipeline systems are characterized by complex structures, wide distribution, and variable operating environments. During long-term operation, these networks are prone to leakage and rupture due to factors such as material aging, external loads, changes in the underground environment, and sudden external damage. Existing methods for detecting pipeline leakage mainly include manual inspections, fixed sensor monitoring, and single-sensor inspection equipment. These methods lack comprehensive judgment capabilities regarding the network's condition, resulting in significant errors in identifying and locating leakage risks, and are prone to missed or false alarms.

[0003] In summary, existing technologies lack the ability to comprehensively assess the status of pipeline networks, resulting in technical problems such as low accuracy and reliability in identifying pipeline leakage risks. Summary of the Invention

[0004] The purpose of this application is to provide a spatiotemporal clustering method and system for leakage hot zones based on inspection robots, in order to solve the technical problem that the existing technology lacks the ability to comprehensively judge the status of the pipeline network, resulting in low accuracy and reliability of pipeline network leakage risk identification.

[0005] To achieve the above objectives, this application provides a spatiotemporal clustering method and system for leakage hot zones based on inspection robots.

[0006] Firstly, this application provides a spatiotemporal clustering method for leakage hotspots based on an inspection robot. The method includes: acquiring temporal multi-source data of a pipeline network, including spatial label information, using an inspection robot equipped with multimodal sensors; extracting temporal and spatial feature relationships from the temporal multi-source data to construct a spatiotemporal data grid; performing data feature association analysis based on the spatiotemporal data grid, aggregating time and space at multiple granularities according to the association features, and determining the spatiotemporal relationship features between each grid; identifying leakage hotspots based on the spatiotemporal relationship features between each grid, and performing visualization transformation in the spatiotemporal data grid to generate identification visualization information.

[0007] Optionally, the multimodal sensor includes at least an infrared thermal imaging sensor, an acoustic data detection sensor, and a visible light image acquisition device. Based on the operating acquisition frequency of the inspection robot and the positioning device, the timestamp and acquisition location of the multi-source acquisition data are determined. The multi-source acquisition data is then arranged in chronological order according to the timestamp and integrated with data from the same source to obtain chronologically ordered multi-source acquisition data. The spatial mapping relationship of the multi-source acquisition data is determined by aligning the acquisition location with the pipeline topology data, generating spatial label information, including spatial location and corresponding pipeline topology positioning, and adding the spatial label information to the chronologically ordered multi-source acquisition data.

[0008] Optionally, based on the pipeline network density and analysis granularity requirements, the pipeline network topology data is physically spatially discretized to construct a spatial grid. Each spatial grid cell corresponds to a spatial size, spatial coordinates, and the set of pipeline segments it covers, as well as topological attributes. Based on the analysis time scale, the continuous time axis is divided into time windows, each with a unified start and end timestamp. The time windows are established as continuous, fixed, or continuous with variable length. Each spatial grid cell is combined with a time window to establish a spatiotemporal cube cell. All spatiotemporal cube cells constitute the spatiotemporal data grid.

[0009] Optionally, a data feature traversal is performed on each grid using a set leakage threshold to identify abnormal grids and abnormal data features; based on the abnormal data features, the abnormal grids are compared for temporal and spatial correlation to obtain correlation time features and correlation spatial features; the correlation time features or correlation spatial features are aggregated at multiple granularities to obtain spatiotemporal relationship features between each grid.

[0010] Optionally, a temporal proximity threshold and a spatial proximity threshold are set; with the similarity of the abnormal data features as the target and the temporal proximity threshold and spatial proximity threshold as constraints, a similar and adjacent grid search is performed to obtain a spatiotemporally related grid; the correlation analysis of the abnormal data features of the spatiotemporally related grid is performed to obtain the associated temporal features and associated spatial features.

[0011] Optionally, multiple temporal granularities and multiple spatial granularities are set. The multiple temporal granularities include time aggregation windows of different granularity levels, and the multiple spatial granularities include spatial aggregation intervals of different spatial recognition ranges. The associated temporal features are aggregated according to the multiple temporal granularities to obtain temporal relationship features. The associated spatial features are aggregated according to the multiple spatial granularities to obtain spatial relationship features.

[0012] Optionally, the grid relationships are divided according to multiple granularities to establish a multi-granularity associated grid, where the associated grid is a related grid whose relationship with the abnormal data features reaches a certain degree of correlation, including grid pairs and grid clusters; for the multi-granularity associated grid, the feature relationships of the abnormal data features are aggregated based on the time duration, spatial coverage, and feature intensity trend changes to obtain the spatiotemporal relationship features; wherein, the earliest and latest timestamps of the associated grids are extracted, and the time duration is determined based on the difference between the earliest and latest timestamps; the corresponding abnormal pipeline is located based on the minimum bounding rectangle of the associated grid to determine the spatial coverage; the maximum abnormal data is sorted based on the timestamps to obtain the abnormal data feature change trend, which is used as the feature intensity trend change.

[0013] Optionally, based on the spatiotemporal relationship characteristics, leakage time-series pattern analysis is performed at multiple preset spatiotemporal analysis granularities to obtain independent suspected leakage areas at each granularity, and the leakage identification results at each granularity are determined. Based on the hierarchical relationship of spatiotemporal granularity from fine to coarse, feature fusion and trend inference are performed between levels on the multi-granularity leakage identification results to analyze the consistency, evolution, and diffusion of suspected areas at different granularities, determine the leakage trend probability of abnormal representations, and generate leakage trend probability identification results. Based on the hot zone judgment rules, the leakage identification results and leakage trend probability identification results at each granularity are subjected to comprehensive clustering analysis in time and space dimensions to determine the leakage hot zone identification results. On the basis of the spatiotemporal data grid, the identified leakage hot zones and their corresponding multi-granularity spatiotemporal relationship characteristics are fused and visualized to generate identification visualization information containing hot zone spatial distribution, risk level, and spatiotemporal evolution trend.

[0014] Optionally, a pipeline leakage-related federated grid is constructed, wherein the related federated grid includes a traffic grid, a meteorological grid, and an underground space complexity grid; the spatiotemporal data grid is overlaid and correlated with the related federated grid to identify leakage hotspot risk impact information, the leakage hotspot risk impact information is visualized and displayed, and early warning reminder information is sent to the interactive ports of the related federated grid.

[0015] Secondly, this application also provides a spatiotemporal clustering system for leakage hotspots based on an inspection robot. The system includes: a data acquisition module, used to acquire temporal multi-source collected data of the pipeline network through an inspection robot equipped with multimodal sensors, including spatial label information; a spatiotemporal data grid construction module, used to extract temporal and spatial feature relationships from the temporal multi-source collected data to construct a spatiotemporal data grid; a spatiotemporal relationship feature determination module, used to perform data feature association analysis based on the spatiotemporal data grid, aggregate time and space at multiple granularities according to the association features, and determine the spatiotemporal relationship features between each grid; and a leakage hotspot identification module, used to identify leakage hotspots based on the spatiotemporal relationship features between each grid, perform visualization transformation in the spatiotemporal data grid, and generate identification visualization information.

[0016] One or more technical solutions provided in this application have at least the following technical effects or advantages: By employing an inspection robot equipped with multimodal sensors, temporal multi-source data of the pipeline network is acquired, including spatial label information. Temporal and spatial feature relationships are extracted from this data to construct a spatiotemporal data grid. Based on this grid, data feature association analysis is performed, and time and space are aggregated at multiple granularities according to the association features to determine the spatiotemporal relationship characteristics between each grid. Leakage hotspots are identified based on these spatiotemporal relationship characteristics, and the data is then visualized within the spatiotemporal data grid to generate visual identification information. This achieves the technical effect of improving the accuracy and reliability of pipeline network leakage hotspot identification by introducing multimodal sensing data from the inspection robot for comprehensive discriminative analysis of the pipeline network status.

[0017] The above description is merely an overview of the technical solution of this application. To enable a clearer understanding of the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0019] Figure 1This is a flowchart illustrating the spatiotemporal clustering method for leakage hot zones based on an inspection robot, as proposed in this application.

[0020] Figure 2 This is a schematic diagram of the structure of a spatiotemporal clustering system for leakage hot zones based on an inspection robot, as proposed in this application.

[0021] Figure labeling: Data acquisition module 11, spatiotemporal data grid construction module 12, spatiotemporal relationship feature determination module 13, leakage hot zone identification module 14. Detailed Implementation

[0022] This application provides a spatiotemporal clustering method and system for leakage hotspots based on inspection robots, solving the technical problem that existing technologies lack comprehensive judgment capabilities on pipeline network conditions, leading to low accuracy and reliability in identifying pipeline network leakage risks. It achieves the technical effect of improving the accuracy and reliability of pipeline network leakage hotspot identification by introducing multimodal perception data from inspection robots to perform comprehensive judgment and analysis of pipeline network conditions.

[0023] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0024] Example 1, please refer to the appendix. Figure 1 This application provides a spatiotemporal clustering method for leakage hotspots based on inspection robots, wherein the spatiotemporal clustering method for leakage hotspots based on inspection robots specifically includes: By using inspection robots equipped with multimodal sensors, time-series multi-source data of the pipeline network can be acquired, including spatial tag information.

[0025] Furthermore, by using an inspection robot equipped with multimodal sensors, temporal multi-source acquisition data of the pipeline network is obtained, including: the multimodal sensors including at least an infrared thermal imaging sensor, an acoustic data detection sensor, and a visible light image acquisition device; determining the timestamp and acquisition location of the multi-source acquisition data based on the operating acquisition frequency of the inspection robot and the positioning device; arranging the multi-source acquisition data in chronological order according to the timestamp and integrating data from the same source to obtain temporal multi-source acquisition data; aligning the acquisition location with the pipeline network topology data according to the position coordinates to determine the spatial mapping relationship of the multi-source acquisition data, generating spatial label information, including spatial location and corresponding pipeline topology positioning, and adding the spatial label information to the temporal multi-source acquisition data.

[0026] Specifically, inspection robots equipped with multimodal sensors are used to inspect pipeline networks. These multimodal sensors include at least infrared thermal imaging sensors, acoustic data detection sensors, and visible light image acquisition devices. Infrared thermal imaging sensors, such as infrared thermal imagers, are used to acquire temperature distribution information of the pipeline and its surrounding environment, reflecting abnormal thermal characteristics caused by media leakage. Acoustic data detection sensors, such as ultrasonic detectors, are used to collect acoustic signals generated during pipeline operation, capturing abnormal acoustic characteristics caused by leakage. For example, in a city's hot water pipelines, the surface temperature is typically maintained between 40℃ and 60℃. When a media leak occurs, the temperature around the leak point will change abnormally. If hot water leaks, the local temperature may rise to 70℃-90℃, depending on the amount of leakage, environmental heat dissipation conditions, etc. These abnormal thermal characteristics can be clearly captured by infrared thermal imagers. During normal operation, the acoustic signal frequency generated by water flow in the pipeline is usually low, generally between 100Hz and 500Hz, mainly caused by water turbulence, pipeline vibration, or steady-state flow noise. However, the frequency of the sound of water leakage in the pipeline is higher, possibly reaching 800Hz to 1500Hz. Visible light image acquisition equipment, such as high-precision cameras, is used to acquire visual image information of the pipeline and its surrounding structure to help identify visual anomalies such as cracks and water seepage marks.

[0027] The inspection robot moves along a predetermined inspection path at a pre-set operating frequency. Using its onboard positioning equipment, such as GPS, it provides a unified timestamp and corresponding collection location for each collection of multi-source data. The timestamp reflects the specific moment of data collection, and the collection location reflects the physical location of the data within the pipeline network. The collected multi-source data is arranged chronologically according to the timestamps, and data from the same source within the same time or collection period are integrated to form temporally continuous, sequential multi-source data. The collection locations are aligned with the pipeline network topology data, which is pre-constructed based on urban planning drawings, actual underground pipe survey records, and past construction documents, and includes at least information such as the pipeline network layout and connection relationships. By aligning coordinates, the location coordinates of multi-source acquired data are mapped to specific pipelines, pipe segments, or nodes, determining the spatial mapping relationship of multi-source acquired data in the pipeline network structure. Spatial label information containing spatial location and corresponding pipeline topology location is generated and added to the time-series multi-source acquired data. The spatial location refers to specific geographic coordinates, and the corresponding pipeline topology location refers to the specific location in the pipeline network topology structure.

[0028] By using multimodal sensors mounted on the inspection robot, it is possible to accurately acquire time-series multi-source data and its spatial label information, and obtain the status information of the pipeline network from different angles. This reduces the limitations that may exist with a single sensor and improves the accurate identification and effective management of pipeline network leakage hotspots. Pipeline network leakage hotspots refer to specific areas in the water supply network where the amount of water leakage is relatively concentrated and the degree of leakage is high.

[0029] Temporal and spatial feature relationships are extracted from the temporal multi-source acquired data to construct a spatiotemporal data grid.

[0030] Furthermore, the temporal and spatial feature relationships of the time-series multi-source acquired data are extracted to construct a spatiotemporal data grid. This includes: physically discretizing the pipeline topology data into spatial regions based on pipeline density and analysis granularity requirements to construct a spatial grid, wherein each spatial grid unit corresponds to spatial dimensions, spatial coordinates, and the set of pipeline segments it covers, as well as topological attributes; dividing the continuous time axis into time windows based on the analysis time scale, each window having a unified start and end timestamp, wherein the time window is continuous, fixed, or continuous with variable length; and combining each spatial grid unit with a time window to establish a spatiotemporal cube unit, the set of all spatiotemporal cube units constituting the spatiotemporal data grid.

[0031] Specifically, based on the pipeline density in the actual physical space and the required granularity of subsequent analysis, the physical region corresponding to the pipeline topology data is spatially discretized to construct a spatial grid. Pipeline density reflects the density of pipeline distribution per unit area, measured by the length of pipeline per square kilometer. Pipeline density = total pipeline length / area. For example, if a central area has a total area of ​​50 square kilometers and a total pipeline length of 1200 kilometers, the pipeline density for that area is 1200 / 50 = 24 kilometers per square kilometer. The granularity of analysis reflects the level of detail in the pipeline analysis. Each constructed spatial grid unit includes corresponding spatial dimensions, spatial coordinates, the set of covered pipeline segments, and topological attributes. Spatial dimensions refer to the length and width of the corresponding spatial grid unit in geographic space; spatial coordinates are specific geographical location information; the set of covered pipeline segments refers to all pipeline segments within the corresponding spatial grid; and topological attributes include the connection relationships and directions of the pipeline segments.

[0032] Based on the time scale set for leakage analysis, the continuous time axis is divided into multiple time windows. Each time window has a unified start and end timestamp, clearly defining the time range it covers. The time scale refers to the analytical precision of the pipeline network data in the time dimension, such as analysis in hours or days. Time windows can be continuous and of fixed length, or they can be continuously variable to adapt to the extraction of dynamic features at different stages. For example, when the time scale is set to the hour level, the continuous time axis is divided into multiple time windows at hourly intervals. For 24-hour leakage monitoring of a city's water supply network, starting from 00:00, a time window is divided every hour, forming a total of 24 time windows. To adapt to the dynamic changes in the pipeline network's operating status, when using continuously variable-length time windows, the time window length can be set to 15 minutes during peak water usage periods, such as 7:00-9:00 AM and 6:00-8:00 PM. During other time periods, the time window length can be extended to 1 hour to accurately extract dynamic features at different stages.

[0033] Each spatial network unit and time window is combined to form a spatiotemporal cube unit that simultaneously possesses spatial region and temporal range constraints. This spatiotemporal cube unit reflects the temporal characteristics of multi-source acquired data within its corresponding spatial region and time window. All spatiotemporal cube units are aggregated to form a spatiotemporal data grid, which comprehensively reflects the status information of the pipeline network within its corresponding spatial region and time range.

[0034] By constructing a spatiotemporal data grid, time-series multi-source collected data is organized and integrated in an orderly manner in the spatiotemporal dimension, providing clear and standardized data support for data feature correlation analysis and leakage hot zone identification, which helps to discover potential problems and anomalies in the pipeline network in a timely and accurate manner.

[0035] Based on the spatiotemporal data grid, data feature association analysis is performed, and time and space are aggregated at multiple granularities according to the association features to determine the spatiotemporal relationship features between each grid.

[0036] Furthermore, based on the spatiotemporal data grid, data feature association analysis is performed, and time and space are aggregated at multiple granularities according to the association features to determine the spatiotemporal relationship features between each grid. This includes: using a set leakage threshold to traverse the data features of each grid to identify abnormal grids and abnormal data features; comparing the temporal and spatial correlations of the abnormal grids based on the abnormal data features to obtain associated temporal features and associated spatial features; and performing multi-granular feature aggregation on the associated temporal features or associated spatial features to obtain the spatiotemporal relationship features between each grid.

[0037] Specifically, within each grid of the spatiotemporal data grid, multi-source data features are compared and traversed using pre-defined leakage thresholds. These leakage thresholds are determined based on normal network operation parameters and past leakage case data, serving as a standard range for determining whether leakage has occurred. This range can be dynamically adjusted; for example, a temperature threshold of 40℃-60℃, an acoustic threshold of 100Hz-500Hz, and an image threshold determined based on pipe image features from past leakage cases, such as the damaged area, are used to determine whether network image data is abnormal to aid in leakage identification. The image threshold is set to 2cm² for the damaged area on the pipe surface. By comparing multi-source data features within each grid, such as temperature, acoustics, and images, with the corresponding leakage thresholds, grids exhibiting anomalies can be quickly identified. Grids with values ​​greater than or less than the corresponding leakage threshold are marked as abnormal networks, and the abnormal data features corresponding to these abnormal grids are clearly identified. For instance, if the temperature within a grid exceeds the corresponding threshold, the grid can be determined to be abnormal, with the abnormal data feature being temperature anomaly.

[0038] Based on anomalous data features, correlation comparisons are performed on anomalous grids in both temporal and spatial dimensions. Temporal correlation refers to the similarity of anomalous features in terms of continuity, repetition, or evolutionary trends within different time windows. Spatial correlation refers to the similarity of anomalous features in terms of distribution location and feature morphology among adjacent or topologically related spatial grids. Correlation analysis yields correlated temporal and spatial features. To avoid the randomness of analysis at a single time or spatial scale, multi-granularity feature aggregation is performed on correlated temporal and spatial features. Multi-granularity refers to summarizing and enhancing correlation relationships across different time spans or spatial ranges. For example, at the temporal granularity level, correlated temporal features from multiple time windows within a day can be summarized to analyze their overall trend. At the spatial granularity level, correlated spatial features from multiple adjacent anomalous grids can be merged to determine spatial correlation patterns over a larger area. Through multi-granularity feature aggregation, the spatiotemporal relationship features between each grid can be comprehensively obtained.

[0039] By extracting spatiotemporal features and performing correlation analysis on time-series multi-source acquisition data, we can comprehensively consider the temporal and spatial variation patterns of leakage, more accurately and comprehensively identify leakage hotspots, avoid missed detections and false detections, and improve the accuracy and reliability of leakage hotspot detection.

[0040] Furthermore, based on the abnormal data features, the abnormal grids are compared for temporal and spatial correlation to obtain associated temporal and spatial features, including: setting temporal proximity thresholds and spatial proximity thresholds; using the similarity of the abnormal data features as the target and the temporal and spatial proximity thresholds as constraints, similar and adjacent grids are searched to obtain spatiotemporally correlated grids; and the correlation analysis of the abnormal data features of the spatiotemporally correlated grids is performed to obtain the associated temporal and spatial features.

[0041] Specifically, a temporal proximity threshold and a spatial proximity threshold are first set. The temporal proximity threshold defines the maximum time interval at which two anomalous grids are considered adjacent on the time axis, while the spatial proximity threshold defines the maximum spatial range at which two anomalous grids are considered adjacent in terms of spatial topology or physical distance. For example, the temporal proximity threshold is set to a time difference of 7 days between the occurrences of two anomalous grids, and the spatial proximity threshold is set to a geographical distance of 50 meters between two anomalous grids. Only anomalous grids whose geographical distance is less than the spatial proximity threshold and whose time difference is less than the temporal proximity threshold are considered to be in the same spatiotemporal neighborhood and have spatiotemporal correlation.

[0042] Using the similarity of anomalous data features as the search target, and setting temporal and spatial proximity thresholds as constraints, a search for similar and adjacent grids is performed on each anomalous grid within the spatiotemporal data grid. During the search, the anomalous data feature similarity is calculated based on anomalous features extracted from infrared thermal imaging, acoustic signals, or visible light images. This can be calculated by determining the Euclidean distance or cosine similarity of relevant indicators, such as the Euclidean distance of temperature deviation, anomalous sound wave patterns, or anomalous image textures. Anomalous data feature similarities exceeding the similarity threshold are used as the search target, and combined with the constraints, a comprehensive search is performed to ensure that only grids with similar feature shapes and spatiotemporal proximity are considered associated grids. After the search is complete, the resulting set of adjacent and similar anomalous grids is defined as the spatiotemporally associated grid.

[0043] Temporal and spatial correlation analysis of anomalous data features is performed on spatiotemporally correlated grids. Time series are constructed for anomalous data within each grid according to their timestamps, such as the changes in infrared temperature anomalies and acoustic signal anomalies over time. The temporal continuity of the anomalous data is analyzed to determine whether the anomaly occurs within multiple consecutive time windows. The duration of the time is calculated by subtracting the earliest timestamp from the latest timestamp of each window. Time series analysis methods, such as the sliding window method, are used to segment and statistically analyze the time series. The mean anomaly intensity is calculated for each window to smooth short-term fluctuations and capture overall trends. By comparing the mean changes of consecutive windows, patterns of increase, decrease, or periodic fluctuation in anomaly intensity over time are identified, thus forming correlated temporal features describing the temporal evolution of anomalous events, including anomaly duration, peak occurrence time, and intensity change trends. Meanwhile, by using statistical methods or spatial clustering algorithms, such as calculating the minimum bounding rectangle of all anomalous grids to determine the coverage area, or by using grid clustering methods to group spatially similar and characteristic anomalous grids into a cluster, the spatial coverage area, distribution density and local aggregation degree of spatiotemporally associated grids can be obtained by statistically analyzing the number, density and relative position distribution of grids in each cluster, thus forming associated spatial features.

[0044] By setting temporal and spatial proximity thresholds, the spatiotemporal range related to anomalous grids can be precisely focused, avoiding interference from irrelevant data and improving the efficiency and accuracy of the analysis. Searching based on the similarity of anomalous data features can identify grids with similar leakage characteristics. Through correlation analysis of spatiotemporally associated grids, correlation time and spatial features are obtained, improving the accuracy and reliability of identifying and determining pipeline network leakage hotspots, thereby effectively enhancing the safety and reliability of pipeline network operation.

[0045] Furthermore, multi-granularity feature aggregation is performed on the associated temporal features or associated spatial features, including: setting multiple temporal granularities and multiple spatial granularities, wherein the multiple temporal granularities include temporal aggregation windows of different granularity levels, and the multiple spatial granularities include spatial aggregation intervals of different spatial recognition ranges; performing feature temporal relationship aggregation on the associated temporal features according to the multiple temporal granularities to obtain temporal relationship features; and performing spatial association feature aggregation on the associated spatial features according to the multiple spatial granularities to obtain spatial relationship features.

[0046] Specifically, multiple temporal and spatial granularities are defined. The multiple temporal granularities include time aggregation windows of different granularity levels, such as hours, days, and weeks. Short-term windows are used to capture instantaneous abnormal fluctuations, medium-term windows reflect phased changes, and long-term windows analyze continuous trends. The multiple spatial granularities include aggregation intervals with different spatial identification ranges, such as grid pairs and grid clusters. Local grid neighborhoods are used to capture small-scale clusters, and regional ranges are used to identify anomalies in local areas of the pipeline network.

[0047] Based on the defined multi-time granularity, the temporal relationships of associated time features are aggregated at each granularity. For each time aggregation window at each time granularity level, the associated time features within that window are integrated and analyzed. For example, under an hourly time aggregation window, the average, maximum, and minimum values ​​of associated time features within each hour are statistically analyzed, or their trends and durations over time are analyzed to obtain temporal relationship features. Simultaneously, associated spatial features are aggregated according to different spatial granularities. By statistically analyzing the density of abnormal grid distribution, spatial coverage, and clustering degree within each spatial interval, spatial relationship features reflecting the spatial nature of abnormal events are obtained.

[0048] By setting multiple time and spatial granularities, a comprehensive and detailed analysis of associated time and spatial features can be performed at different scales. This avoids information omissions and biases that may result from single-granularity analysis, improves the accuracy and reliability of leak hotspot identification, and thus effectively enhances the safety and reliability of pipeline network operation and reduces losses caused by leakage.

[0049] Furthermore, multi-granularity feature aggregation is performed on the associated temporal or spatial features to obtain the spatiotemporal relationship features between each grid. This includes: dividing the grid relationships according to multi-granularity to establish multi-granularity associated grids, where associated grids are related grids whose abnormal data features have reached a certain degree of association, including grid pairs and grid clusters; for the multi-granularity associated grids, feature relationship aggregation is performed on the temporal duration span, spatial coverage, and feature intensity trend changes of the abnormal data features to obtain the spatiotemporal relationship features; wherein, the earliest and latest timestamps of the associated grids are extracted, and the time duration span is determined based on the difference between the earliest and latest timestamps; the corresponding abnormal pipeline is located based on the minimum bounding rectangle of the associated grid to determine the spatial coverage; and the largest abnormal data is sorted based on the timestamps to obtain the abnormal data feature change trend, which is used as the feature intensity trend change.

[0050] Specifically, based on the set multi-temporal granularity and multi-spatial granularity, the grid relationships are divided to establish a multi-granularity associated grid. The associated grid refers to the related grids that have reached the degree of correlation in time and space in the relationship of abnormal data features. It includes grid pairs and grid clusters. The grid pair consists of two highly correlated abnormal grids, and the grid cluster consists of multiple related abnormal grids, which are used to reflect the aggregation pattern of abnormal events at the local or regional scale.

[0051] Feature relationship aggregation is performed on multiple granularity-level association networks. This includes aggregating the temporal duration, spatial coverage, and trend changes in feature intensity of anomalous features. In the temporal dimension, the earliest and latest timestamps of anomalous data features in the associated grids are extracted, and the difference between them is calculated to obtain the temporal duration of the anomalous event, characterizing its persistence over time. For example, in a grid pair, if the earliest occurrence of an acoustic anomaly is at 8:00 AM and the latest is at 10:00 AM, then its temporal duration is 2 hours. In the spatial dimension, the spatial coverage of the anomalous event is determined by calculating the minimum bounding rectangle of the associated grid and locating the covered anomalous pipe, reflecting the degree of spatial diffusion of the anomalous event within the pipe network. For example, if a grid cluster composed of multiple adjacent grids has its minimum bounding rectangle covering a specific pipe area in the network, then this pipe area represents the spatial coverage of the associated grid. Simultaneously, the maximum anomalous data within the associated grids is sorted based on timestamps, and the trend of anomalous intensity over time is analyzed to obtain the feature intensity trend, reflecting the strength and evolution trend of the anomalous event.

[0052] After completing the aggregation of the above three dimensions, the duration of time, spatial coverage and trend changes of feature intensity are uniformly encoded into spatiotemporal relationship feature vectors or data structures of associated grids, forming spatiotemporal relationship features, which serve as a quantitative representation of the spatiotemporal distribution, persistence and change patterns of abnormal events.

[0053] By dividing the grid relationships into multiple granularities and establishing associated grids, we can conduct detailed analysis of pipeline leakage at different scales. By aggregating the characteristic relationships of associated grids based on their temporal span, spatial coverage, and trend changes in characteristic intensity, we can comprehensively and accurately grasp the dynamic changes and mutual influences of leakage in the spatiotemporal dimensions, thereby improving the accuracy and reliability of leakage hotspot identification.

[0054] Based on the spatiotemporal relationship characteristics between the grids, leakage hotspots are identified, and then visualized in the spatiotemporal data grid to generate identification visualization information.

[0055] Furthermore, based on the spatiotemporal relationship characteristics between the grids, leakage hotspots are identified, and visualization transformation is performed on the spatiotemporal data grid to generate identification visualization information. This includes: based on the spatiotemporal relationship characteristics, leakage time-series pattern analysis is performed at multiple preset spatiotemporal analysis granularities to obtain independent suspected leakage areas at each granularity, and the leakage identification results at each granularity are determined; based on the hierarchical relationship of spatiotemporal granularity from fine to coarse, feature fusion and trend inference are performed on the multi-granularity leakage identification results, analyzing the consistency, evolution, and diffusion of suspected areas at different granularities, determining the leakage trend probability of abnormal representations, and generating leakage trend probability identification results; based on hotspot judgment rules, the leakage identification results at each granularity and the leakage trend probability identification results are subjected to comprehensive clustering analysis in time and space dimensions to determine leakage hotspot identification results; based on the spatiotemporal data grid, the identified leakage hotspots and their corresponding multi-granularity spatiotemporal relationship characteristics are fused and visualized to generate identification visualization information containing hotspot spatial distribution, risk level, and spatiotemporal evolution trend.

[0056] Specifically, based on spatiotemporal relationship characteristics, leakage time-series pattern analysis is performed at multiple preset spatiotemporal analysis granularities. That is, according to preset multi-temporal and multi-spatial granularities, the spatiotemporal data grid is divided into different time windows and spatial identification intervals. For anomalies within each time window, their temporal persistence, spatial coverage, and feature intensity trend changes are calculated, and multiple indicators are normalized using Min-Max normalization. The normalized indicators are then weighted or comprehensively scored to quantify the risk level of the anomaly at that granularity. Risk level = weight 1 × temporal persistence + weight 2 × spatial coverage + weight 3 × feature intensity trend change. Weights 1, 2, and 3 are the weighting coefficients for temporal persistence, spatial coverage, and feature intensity trend change, respectively, and their sum is 1. These weights can be dynamically set according to actual needs and historical data analysis results. Based on set thresholds or scoring criteria, high-risk areas are marked as suspected leakage areas. For example, a risk threshold of 0.7 is set, and areas with a risk level greater than or equal to this threshold are considered high-risk areas. Ultimately, independent granular-level leakage identification results are formed for each time window and spatial interval combination, reflecting the potential leakage distribution of abnormal events at different time and spatial scales. For instance, by normalizing the three indicators of time persistence, spatial coverage, and feature intensity trend change, a time persistence value of 0.8, a spatial coverage value of 0.67, and a feature intensity change value of 0.7 are obtained. Simultaneously, based on actual needs and historical data analysis results, the weighting coefficients for time persistence, spatial coverage, and feature intensity trend change are set to 0.4, 0.3, and 0.3, respectively. Then, by calculating the abnormal time risk level = 0.4 × 0.8 + 0.3 × 0.67 + 0.3 × 0.7 = 0.713, which is greater than 0.7. Therefore, this area is marked as a suspected leakage area under this time window and spatial interval combination.

[0057] Based on a hierarchical relationship of spatiotemporal granularity from fine to coarse, the multi-granularity identification results are fused and trend inferred across levels. By comparing the consistency, evolution, and diffusion of the same anomalous region at different granularities, the probability of leakage trends in its anomalous characterization is calculated, forming a quantitative trend probability identification result to describe the future leakage possibility and evolution trend of the region. Consistency analysis is used to determine whether the anomalous region exhibits similar feature patterns at different granularities. This can be achieved by calculating consistency indices, such as cosine similarity or Pearson correlation coefficient, to determine the similarity between feature vectors at different granularities. The similarity coefficient is used as the consistency analysis result. If the similarity coefficient is higher than a preset threshold, such as 0.7, the region is considered to be consistent. Evolutionary analysis focuses on the changes in the features of the anomalous region at different time granularities. Using time sequence as a benchmark, the dynamic changes of anomalous data features at different granularities are analyzed. For example, the evolutionary index is obtained by calculating the ratio of the change amplitude of anomalous feature values ​​at different time points to the time interval. If the evolutionary index exceeds a certain threshold, such as 10%, it indicates that the anomalous features in the region exhibit an evolutionary trend. Diffusion analysis primarily examines the impact of anomaly regions on surrounding areas at different spatial granularities. It determines whether the anomaly has a tendency to spread to the surrounding areas by analyzing the occurrence of anomalous data characteristics in adjacent grids or larger grid clusters. For example, a diffusion index is calculated based on the ratio of the number of surrounding grids with anomalies to the total number of grids. When the diffusion index exceeds a preset threshold, such as 0.3, it indicates that the anomaly region has diffusion potential. Combining the results of consistency, evolution, and diffusion analyses, a weighted average method is used to calculate the leakage trend probability of the anomaly region. Different indicators are assigned corresponding weights; for example, consistency is weighted at 0.4, evolution at 0.3, and diffusion at 0.3. These weights can be adjusted based on actual needs and experience. The final leakage trend probability value is obtained by multiplying each indicator value by its corresponding weight and summing the results. The value ranges from 0 to 1. The closer the leakage trend probability value is to 1, the greater the likelihood of future leakage in the region and the more obvious the evolution trend; conversely, the closer it is to 0, the lower the likelihood of leakage.

[0058] Based on predefined hotspot judgment rules, such as comprehensively considering factors like the frequency of occurrence, degree of anomaly, and trend probability of suspected areas, the DBSCAN clustering algorithm is used. By setting the neighborhood radius and minimum number of points, the leakage identification results at each granularity are comprehensively clustered with the trend probability identification results in both time and space dimensions, grouping suspected areas with similar characteristics into one category. Based on the clustering results, the leakage hotspot identification results are determined, including multi-level risk hotspots, such as high-risk, medium-risk, and low-risk hotspots. Among them, high-risk hotspots are clusters of suspected areas that frequently appear at multiple time points, have a high degree of anomaly, and a high trend probability.

[0059] Based on the spatiotemporal data grid, and according to the identified leakage hotspots and their corresponding multi-granularity spatiotemporal relationship characteristics, information such as the spatial location, risk level, and spatiotemporal evolution trend of the leakage hotspots is presented in an intuitive way using graphics and colors, generating identification visualization information that includes the spatial distribution, risk level, and spatiotemporal evolution trend of the hotspots. For example, different colors are used to represent different risk levels: red represents high-risk hotspots, yellow represents medium-risk hotspots, and green represents low-risk hotspots, and dynamic graphics are used to show the diffusion process of leakage at different times.

[0060] By transforming the extracted and aggregated spatiotemporal relationship features into actionable hot zone identification results and visualization information, a closed-loop mapping from abnormal grid data to actual pipeline leakage hot zones is achieved. This not only improves the accuracy and reliability of leakage hot zone identification and location, but also provides an intuitive reference for operation and maintenance decisions based on risk level and evolution trend, effectively reducing pipeline leakage losses.

[0061] Furthermore, the method also includes: constructing a pipeline leakage-related federated grid, wherein the related federated grid includes a traffic grid, a meteorological grid, and an underground space complexity grid; performing overlay correlation analysis on the spatiotemporal data grid and the related federated grid to identify leakage hotspot risk impact information, visualizing the leakage hotspot risk impact information, and sending early warning reminder information to the interactive ports of the related federated grid.

[0062] Specifically, a federated grid related to pipeline leakage risk is constructed. This federated grid refers to a set of multi-source risk grids that represent different urban operational elements in a gridded manner, using a spatial coordinate system consistent with the spatiotemporal data grid of the pipeline network and an alignable grid scale. These grids include traffic grids, meteorological grids, and underground spatial complexity grids. Specifically, a traffic grid is constructed based on urban road and traffic operation data, with each traffic grid unit representing the road grade, traffic flow, and traffic sensitivity within its corresponding area. A meteorological grid is constructed based on meteorological monitoring data, with each meteorological grid unit representing environmental factors that may affect pipeline stability, such as rainfall and temperature / humidity variations. An underground spatial complexity grid is constructed based on underground space exploration and pipeline distribution data to reflect underground environmental characteristics such as underground pipeline density, soil structure, and cavity risk.

[0063] After the federated grid is constructed, the spatiotemporal data grid of the pipeline network is overlaid and correlated with the traffic grid, meteorological grid, and underground space complexity grid. The identified leakage hotspots are used as the core analysis objects. Spatially, their coverage is mapped to the corresponding federated grid cells, and temporally, their occurrence time windows are aligned. This allows the extraction of the corresponding risk attribute information of leakage hotspots in different federated grids. By analyzing the degree of overlap between leakage hotspots and traffic-sensitive areas, areas with extreme weather conditions, or areas with high underground space complexity, the potential superimposed risk impact information can be identified, such as pipeline rupture causing traffic disruptions or ground subsidence.

[0064] The identified leakage hotspots are overlaid with risk impact information for visualization. This visualization, presented within a unified spatiotemporal grid interface, uses heat maps, risk classification markers, or dynamic layers to show the overlay relationship between leakage hotspots and risk factors such as transportation, meteorology, and underground spaces. For example, different colored symbols mark the location of leakage hotspots, with color intensity indicating the degree of risk; arrows indicate possible directions of risk propagation; and bar charts or pie charts display the probability of occurrence and impact proportion of different types of risks. Simultaneously, through predefined federated grid interaction ports, corresponding early warning information is sent to relevant business systems or management platforms, such as sending road impact warnings to traffic management systems or ground subsidence risk alerts to urban safety management systems, thereby achieving cross-system collaborative response and coordinated handling.

[0065] By constructing a federated grid related to pipeline leakage and overlaying it with the spatiotemporal data grid of the pipeline network for correlation analysis, the leakage risk has been expanded from a single pipeline network dimension to a multi-element dimension of the city. The pipeline leakage is comprehensively assessed in conjunction with factors such as traffic operation, meteorological environment and underground space safety. This not only improves the comprehensiveness and foresight of leakage risk identification, but also supports cross-departmental joint early warning and collaborative decision-making, significantly enhancing the operational safety and emergency response capabilities of urban infrastructure.

[0066] Example 2: Based on the same inventive concept as the spatiotemporal clustering method for leakage hot zones based on inspection robots in Example 1, this application also provides a spatiotemporal clustering system for leakage hot zones based on inspection robots. Please refer to the appendix. Figure 2 The aforementioned spatiotemporal clustering system for leakage hot zones based on inspection robots includes: The data acquisition module 11 is used to acquire temporal multi-source collected data of the pipeline network through an inspection robot equipped with multimodal sensors, including spatial label information; the spatiotemporal data grid construction module 12 is used to extract temporal and spatial feature relationships from the temporal multi-source collected data and construct a spatiotemporal data grid; the spatiotemporal relationship feature determination module 13 is used to perform data feature association analysis based on the spatiotemporal data grid, aggregate time and space at multiple granularities according to the association features, and determine the spatiotemporal relationship features between each grid; the leakage hotspot identification module 14 is used to identify leakage hotspots based on the spatiotemporal relationship features between each grid, perform visualization transformation in the spatiotemporal data grid, and generate identification visualization information.

[0067] Furthermore, the data acquisition module 11 is also used for: the multimodal sensor including at least an infrared thermal imaging sensor, an acoustic data detection sensor, and a visible light image acquisition device; determining the timestamp and acquisition location of the multi-source acquisition data based on the operating acquisition frequency of the inspection robot and the positioning device; arranging the multi-source acquisition data in chronological order according to the timestamp and integrating data from the same source to obtain chronologically sequenced multi-source acquisition data; aligning the acquisition location with the pipeline topology data according to the position coordinates to determine the spatial mapping relationship of the multi-source acquisition data, generating spatial label information, including spatial location and corresponding pipeline topology positioning, and adding the spatial label information to the chronologically sequenced multi-source acquisition data.

[0068] Furthermore, the spatiotemporal data grid construction module 12 is also used to: physically discretize the pipeline topology data according to the pipeline density and analysis granularity requirements to construct a spatial grid, wherein each spatial grid unit corresponds to spatial size, spatial coordinates, and the set of pipeline segments and topological attributes it covers; divide the continuous time axis into time windows according to the analysis time scale, each window having a unified start and end timestamp, wherein the time window is continuous, fixed or continuous and variable in length; combine each spatial grid unit with the time window to establish a spatiotemporal cube unit, and the set of all spatiotemporal cube units constitutes the spatiotemporal data grid.

[0069] Furthermore, the spatiotemporal relationship feature determination module 13 is also used to: perform data feature traversal on each grid using a set leakage threshold to identify abnormal grids and abnormal data features; perform temporal correlation and spatial correlation comparison on the abnormal grids based on the abnormal data features to obtain correlation time features and correlation spatial features; and perform multi-granular feature aggregation on the correlation time features or correlation spatial features to obtain spatiotemporal relationship features between each grid.

[0070] Furthermore, the spatiotemporal relationship feature determination module 13 is also used to: set a time proximity threshold and a spatial proximity threshold; perform a similar and adjacent grid search with the similarity of the abnormal data features as the target and the time proximity threshold and spatial proximity threshold as constraints to obtain a spatiotemporally associated grid; and perform an abnormal data feature correlation analysis on the spatiotemporally associated grid to obtain the associated time feature and associated spatial feature.

[0071] Furthermore, the spatiotemporal relationship feature determination module 13 is also used to: set multiple temporal granularities and multiple spatial granularities, wherein the multiple temporal granularities include time aggregation windows of different granularity levels, and the multiple spatial granularities include spatial aggregation intervals of different spatial recognition ranges; perform feature temporal relationship aggregation of the associated temporal features at each granularity according to the multiple temporal granularities to obtain temporal relationship features; and perform spatial association feature aggregation of the associated spatial features according to the multiple spatial granularities to obtain spatial relationship features.

[0072] Furthermore, the spatiotemporal relationship feature determination module 13 is also used for: dividing the grid relationship according to multiple granularities to establish a multi-granularity associated grid, wherein the associated grid is a related grid whose relationship of abnormal data features reaches the degree of association, including grid pairs and grid clusters; aggregating the feature relationships of the abnormal data features in the multi-granularity associated grid by considering the time duration, spatial coverage, and feature intensity trend changes to obtain the spatiotemporal relationship features; wherein the earliest and latest timestamps of the associated grids are extracted, and the time duration is determined based on the difference between the earliest and latest timestamps; the corresponding abnormal pipe is located based on the minimum bounding rectangle of the associated grid to determine the spatial coverage; and the maximum abnormal data is sorted based on the timestamps to obtain the abnormal data feature change trend, which is used as the feature intensity trend change.

[0073] Furthermore, the leakage hotspot identification module 14 is also used for: performing leakage time-series pattern analysis at multiple preset spatiotemporal analysis granularities based on the spatiotemporal relationship characteristics, obtaining independent suspected leakage areas at each granularity, and determining the leakage identification results at each granularity; performing feature fusion and trend inference between levels based on the hierarchical relationship of spatiotemporal granularity from fine to coarse, analyzing the consistency, evolution, and diffusion of suspected areas at different granularities, determining the leakage trend probability of abnormal representations, and generating leakage trend probability identification results; performing comprehensive clustering analysis of the leakage identification results and leakage trend probability identification results at each granularity based on the hotspot judgment rules, and determining the leakage hotspot identification results in the time and space dimensions; and, based on the spatiotemporal data grid, performing fusion and visualization transformation on the identified leakage hotspots and their corresponding multi-granularity spatiotemporal relationship characteristics to generate identification visualization information containing hotspot spatial distribution, risk level, and spatiotemporal evolution trend.

[0074] Furthermore, the aforementioned spatiotemporal clustering system for leakage hotspots based on inspection robots is also used to: construct a pipeline leakage-related federated grid, wherein the related federated grid includes a traffic grid, a meteorological grid, and an underground space complexity grid; perform overlay correlation analysis on the spatiotemporal data grid and the related federated grid to identify the superimposed risk impact information of leakage hotspots, visualize the risk impact information of the superimposed risk impact information of leakage hotspots, and send early warning reminder information to the interactive ports of the related federated grids.

[0075] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0076] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A spatiotemporal clustering method for leakage hot zones based on inspection robots, characterized in that, include: By using inspection robots equipped with multimodal sensors, time-series multi-source data of the pipeline network can be acquired, including spatial tag information; Temporal and spatial feature relationships are extracted from the temporal multi-source acquired data to construct a spatiotemporal data grid; Based on the spatiotemporal data grid, data feature association analysis is performed, and time and space are aggregated at multiple granularities according to the association features to determine the spatiotemporal relationship features between each grid. Based on the spatiotemporal relationship characteristics between the grids, leakage hotspots are identified, and then visualized in the spatiotemporal data grid to generate identification visualization information.

2. The spatiotemporal clustering method for leakage hot zones based on inspection robots according to claim 1, characterized in that, By using inspection robots equipped with multimodal sensors, time-series multi-source data of the pipeline network is acquired, including: The multimodal sensor includes at least an infrared thermal imaging sensor, an acoustic data detection sensor, and a visible light image acquisition device. Based on the operating acquisition frequency of the inspection robot and the positioning device, the timestamp and acquisition location of the multi-source acquisition data are determined. The multi-source collected data is arranged in time sequence according to timestamps and the data from the same source is integrated to obtain time-series multi-source collected data; Aligning the acquisition locations with the pipeline topology data to determine the spatial mapping relationship of the multi-source acquisition data, generating spatial label information including spatial location and corresponding pipeline topology location, and adding the spatial label information to the time-series multi-source acquisition data.

3. The spatiotemporal clustering method for leakage hot zones based on inspection robots according to claim 2, characterized in that, Extracting temporal and spatial feature relationships from the aforementioned temporal multi-source acquired data and constructing a spatiotemporal data grid includes: Based on the pipeline density and analysis granularity requirements, the pipeline topology data is physically spatially discretized to construct a spatial grid, where each spatial grid unit corresponds to spatial size, spatial coordinates, and the set of pipeline segments and topological attributes it covers. Based on the time scale of the analysis, the continuous time axis is divided into time windows, each with a unified start and end timestamp, and time windows are established, wherein the time windows are continuous, fixed or continuous, or of variable length. Each spatial grid cell is combined with a time window to create a spatiotemporal cube cell, and the collection of all spatiotemporal cube cells constitutes the spatiotemporal data grid.

4. The spatiotemporal clustering method for leakage hot zones based on inspection robots according to claim 1, characterized in that, Based on the spatiotemporal data grid, data feature association analysis is performed, and time and space are aggregated at multiple granularities according to the association features to determine the spatiotemporal relationship features between each grid, including: By setting a leakage threshold, data features are traversed for each grid to identify abnormal grids and abnormal data features. Based on the abnormal data characteristics, the abnormal grids are compared in terms of temporal and spatial correlation to obtain correlation time features and correlation spatial features; Multi-granularity feature aggregation is performed on the associated temporal or spatial features to obtain the spatiotemporal relationship features between each grid.

5. The spatiotemporal clustering method for leakage hot zones based on inspection robots according to claim 4, characterized in that, Based on the aforementioned abnormal data characteristics, the abnormal grids are compared for temporal and spatial correlation to obtain correlation temporal and spatial features, including: Set time proximity thresholds and spatial proximity thresholds; Using the similarity of the abnormal data features as the target and the temporal proximity threshold and spatial proximity threshold as constraints, a similar and adjacent grid search is performed to obtain a spatiotemporally related grid. A correlation analysis of abnormal data features is performed on the spatiotemporal correlation grid to obtain the correlation time features and correlation spatial features.

6. The spatiotemporal clustering method for leakage hot zones based on inspection robots according to claim 5, characterized in that, Multi-granularity feature aggregation is performed on the associated temporal features or associated spatial features, including: Multiple temporal granularities and multiple spatial granularities are set. The multiple temporal granularities include temporal aggregation windows with different granularity levels, and the multiple spatial granularities include spatial aggregation intervals with different spatial recognition ranges. Based on the multi-time granularity, the associated time features are aggregated to obtain time relationship features; Spatial correlation features are aggregated based on the multi-spatial granularity to obtain spatial relationship features.

7. The spatiotemporal clustering method for leakage hot zones based on inspection robots according to claim 6, characterized in that, Multi-granularity feature aggregation is performed on the associated temporal or spatial features to obtain spatiotemporal relationship features between grids, including: The grid relationships are divided according to multiple granularities, and a multi-granularity associated grid is established. The associated grid is a related grid whose relationship with the abnormal data characteristics reaches the degree of association, including grid pairs and grid clusters. For the multi-granularity associated grid, feature relationships are aggregated based on the temporal duration, spatial coverage, and trend changes in feature intensity of abnormal data features to obtain the spatiotemporal relationship features; Specifically, the earliest and latest timestamps of the associated grids are extracted, and the time span is determined based on the difference between the earliest and latest timestamps; the corresponding abnormal pipes are located based on the minimum bounding rectangle of the associated grids to determine the spatial coverage; the largest abnormal data are sorted based on the timestamps to obtain the abnormal data feature change trend, which is used as the feature intensity trend change.

8. The spatiotemporal clustering method for leakage hot zones based on inspection robots according to claim 7, characterized in that, Based on the spatiotemporal relationship characteristics between the grids, leakage hotspots are identified, and a visualization transformation is performed on the spatiotemporal data grid to generate identification visualization information, including: Based on the spatiotemporal relationship characteristics, leakage time sequence pattern analysis is performed at multiple preset spatiotemporal analysis granularities to obtain independent suspected leakage regions at each granularity and determine the leakage identification results at each granularity. Based on the hierarchical relationship of spatiotemporal granularity from fine to coarse, feature fusion and trend inference are performed between levels on the multi-granularity leakage identification results. The consistency, evolution and diffusion of suspected areas under different granularities are analyzed, the leakage trend probability of abnormal characterization is determined, and the leakage trend probability identification result is generated. Based on the hot zone judgment rule, the leakage identification results and leakage trend probability identification results of each granularity are subjected to comprehensive cluster analysis in time and space dimensions to determine the leakage hot zone identification results. Based on the spatiotemporal data grid, the identified leakage hotspots and their corresponding multi-granular spatiotemporal relationship features are fused and visualized to generate identification visualization information that includes the spatial distribution of hotspots, risk levels, and spatiotemporal evolution trends.

9. The spatiotemporal clustering method for leakage hot zones based on inspection robots according to claim 1, characterized in that, Also includes: Construct a federal grid related to pipeline leakage, which includes a traffic grid, a meteorological grid, and an underground space complexity grid; The spatiotemporal data grid is overlaid and correlated with the relevant federated grid to identify the risk impact information of the superimposed leakage hot zone. The risk impact information of the superimposed leakage hot zone is visualized and displayed, and early warning reminder information is sent to the interaction port of the relevant federated grid.

10. A spatiotemporal clustering system for leakage hot zones based on an inspection robot, characterized in that, For implementing the spatiotemporal clustering method for leakage hot zones based on inspection robots according to any one of claims 1 to 9, the spatiotemporal clustering system for leakage hot zones based on inspection robots comprises: The data acquisition module is used to acquire time-series multi-source data of the pipeline network through an inspection robot equipped with multimodal sensors, including spatial tag information; The spatiotemporal data grid construction module is used to extract the temporal and spatial feature relationships from the temporal multi-source acquired data and construct a spatiotemporal data grid. The spatiotemporal relationship feature determination module is used to perform data feature association analysis based on the spatiotemporal data grid, and to aggregate time and space at multiple granularities according to the association features to determine the spatiotemporal relationship features between each grid. The leakage hotspot identification module is used to identify leakage hotspots based on the spatiotemporal relationship characteristics between the grids, perform visualization transformation on the spatiotemporal data grid, and generate identification visualization information.

Images