Data model design method, design device, electronic equipment and storage medium

Abstract

The application provides a data model design method, a design device, electronic equipment and a storage medium, and belongs to the field of data processing. The data model design method comprises the following steps: obtaining investigation results of a city rail vehicle-mounted autonomous sensing system, wherein the investigation results comprise business demand investigation results and business data investigation results; determining a sensing data conceptual model based on the business demand investigation results, data sources of multi-source heterogeneous data and data interfaces of the city rail vehicle-mounted autonomous sensing system; determining a sensing data physical model based on the sensing data conceptual model, the data interfaces of the city rail vehicle-mounted autonomous sensing system, the multi-source heterogeneous data and medium information of a storage medium adopted; and constructing a data model for a city rail vehicle-mounted autonomous sensing data closed loop based on the sensing data physical model. Through the design of a unified, standard and normative data model, data analysis and data mining are performed on the city rail vehicle-mounted autonomous sensing system, so that the detection capability and the diagnosis capability of the environmental relationship are improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a data model design method, design device, electronic device, and storage medium. Background Technology

[0002] Smart city rail transit systems are equipped with onboard autonomous sensing systems.

[0003] Currently, the detection capabilities and environmental relationship diagnostic capabilities of urban rail vehicle-mounted autonomous sensing systems are relatively weak. This is because different sensing applications within these systems perceive different objects, and the data structures obtained from these applications also differ. Sensing data from different sources and with varying structures cannot fully analyze and leverage the detection capabilities and environmental relationship diagnostic capabilities of urban rail vehicle-mounted autonomous sensing systems.

[0004] Therefore, how to integrate and manage data from different sources and with different structures in a unified manner, and improve the detection capabilities and environmental relationship diagnostic capabilities of urban rail vehicle-mounted autonomous sensing systems, is an urgent problem to be solved. Summary of the Invention

[0005] This invention provides a data model design method, design device, electronic device, and storage medium to address the shortcomings of weak detection capabilities and environmental relationship diagnostic capabilities in existing urban rail vehicle-mounted autonomous sensing systems, thereby improving the detection capabilities and environmental relationship diagnostic capabilities of urban rail vehicle-mounted autonomous sensing systems.

[0006] This invention provides a data model design method, comprising the following steps.

[0007] Obtain the survey results of the urban rail vehicle-mounted autonomous perception system. The survey results include the business demand survey results and the business data survey results. The business data survey results include multi-source heterogeneous data generated during the operation of the urban rail vehicle-mounted autonomous perception system. Based on the business requirements survey results, the data sources of the multi-source heterogeneous data, and the data interface of the urban rail vehicle-mounted autonomous perception system, a conceptual model of perception data is determined. The physical model of the sensing data is determined based on the conceptual model of the sensing data, the data interface of the urban rail vehicle-mounted autonomous sensing system, the multi-source heterogeneous data, and the media information of the storage medium used. Based on the aforementioned physical model of sensing data, a data model for closed-loop autonomous sensing data of urban rail vehicles is constructed.

[0008] According to a data model design method provided by the present invention, the perceived data concept model includes data entities, data analysis topics, and a common concept data model layer; The conceptual model for determining sensing data based on the business requirements survey results, the data sources of the multi-source heterogeneous data, and the data interface of the urban rail vehicle-mounted autonomous sensing system includes: The data entity is determined based on the data source of the multi-source heterogeneous data, and the data entity includes one or more of the following: positioning system, signal system, sensing system, and electronic map; The data analysis topic is determined based on the results of the business needs survey. Based on the data interface, the common concept data model layer that associates the data analysis topic and the data entity is determined.

[0009] According to a data model design method provided by the present invention, the common concept data model layer is further used to perform one or more of the following processes: Alignment of sensing data collected by the sensing system with signal log data in the signal system, alignment of positioning log data in the positioning system with signal data output by the signal system, and matching of sensing data with pose data.

[0010] According to a data model design method provided by the present invention, the data analysis topic includes one or more of the following: Positioning error analysis, sign recognition analysis, traffic signal recognition analysis, obstacle recognition analysis, and perception data annotation.

[0011] According to a data model design method provided by the present invention, the perceptual data annotation includes: Add one or more of the following tags to your point cloud and image data: Time period, season, location, gradient, curve, turnout, platform, transponder, signal, signage.

[0012] According to a data model design method provided by the present invention, the perceived data physical model includes a raw layer, a unified data warehouse layer, and an application layer; The determination of the physical model of the sensing data based on the conceptual model of the sensing data, the data interface of the urban rail vehicle-mounted autonomous sensing system, the multi-source heterogeneous data, and the media information of the storage medium used includes: The original layer, the unified data warehouse layer, and the application layer are deployed based on the media information. The original layer is configured based on the data interface of the urban rail vehicle-mounted autonomous sensing system and the multi-source heterogeneous data. Configure the unified data warehouse layer based on the aforementioned conceptual model of perceived data.

[0013] According to a data model design method provided by the present invention, the unified data warehouse layer includes a spatial dimension table, a train location detail table, and a first detail data segment table, wherein the first detail data segment table includes detail data segments of each section through which the train sets pass. Receive update operations on the spatial dimension table; In response to the update operation, the first detailed data fragment table is updated based on the train location details table and the updated spatial dimension table.

[0014] The present invention also provides a data model design apparatus, comprising the following modules: The acquisition module is used to acquire the survey results of the urban rail vehicle-mounted autonomous perception system. The survey results include business demand survey results and business data survey results. The business data survey results include multi-source heterogeneous data generated during the operation of the urban rail vehicle-mounted autonomous perception system. The first determining module is used to determine the conceptual model of the sensing data based on the results of the business demand survey, the data sources of the multi-source heterogeneous data, and the data interface of the urban rail vehicle-mounted autonomous sensing system. The second determining module is used to determine the physical model of the sensing data based on the conceptual model of the sensing data, the data interface of the urban rail vehicle-mounted autonomous sensing system, the multi-source heterogeneous data, and the media information of the storage medium used. The construction module is used to build a data model for the closed loop of autonomous sensing data on urban rail vehicles based on the physical model of the sensing data.

[0015] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the data model design method as described above.

[0016] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the data model design method as described above.

[0017] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the data model design method as described above.

[0018] This invention provides a data model design method, design device, electronic device, and storage medium. By conducting relevant research on urban rail vehicle-mounted autonomous sensing systems, a unified, standardized, and regulated data model is designed for data analysis and data mining of urban rail vehicle-mounted autonomous sensing systems, thereby improving the detection capabilities and environmental relationship diagnostic capabilities of urban rail vehicle-mounted autonomous sensing systems. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this invention 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 some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is one of the flowcharts illustrating the data model design method provided by this invention.

[0021] Figure 2 This is a flowchart illustrating the process of determining the conceptual model of sensing data based on the results of business needs surveys, data sources of multi-source heterogeneous data, and data interfaces of the urban rail vehicle-mounted autonomous sensing system, as provided by this invention.

[0022] Figure 3 This is a schematic diagram of the conceptual model of perceptual data provided by the present invention.

[0023] Figure 4 This is a schematic diagram of the physical model of sensing data provided by the present invention.

[0024] Figure 5 This is a schematic diagram of the data model design device provided by the present invention.

[0025] Figure 6 This is a schematic diagram of the structure of the electronic device provided by the present invention.

[0026] Figure label: 501: Acquisition module; 502: First determination module; 503: Second determination module; 504: Construction module; 610: Processor; 620: Communication interface; 630: Memory; 640: Communication bus. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0028] To address the technical challenge mentioned in the background section regarding the difficulty in effectively integrating and analyzing data from diverse sources and varying structures in smart city rail transit systems, which hinders the diagnosis of the relationship between the system's detection capabilities and the surrounding environment and ultimately limits the overall performance improvement of the system, this invention provides a data model design method.

[0029] The execution entity of this method can be a computing cluster consisting of one or more servers, a cloud computing platform, or any other electronic device with data processing and storage capabilities. This electronic device performs all or part of the steps of the method of the present invention by executing computer program instructions stored thereon.

[0030] The following is combined Figures 1 to 6 The present invention describes a data model design method, design apparatus, electronic device, and storage medium.

[0031] Figure 1 This is one of the flowcharts illustrating the data model design method provided by this invention, such as... Figure 1 As shown, the method includes the following: Step 101: Obtain the survey results of the urban rail vehicle-mounted autonomous sensing system. The survey results include the business requirements survey results and the business data survey results. The business data survey results include multi-source heterogeneous data generated during the operation of the urban rail vehicle-mounted autonomous sensing system.

[0032] In some embodiments, the survey results of the urban rail vehicle-mounted autonomous perception system can be obtained by receiving survey documents and data interface descriptions manually entered by technicians through a human-machine interface, or by automatically scanning and parsing metadata information in electronic documents, configuration files, or databases under a specified path.

[0033] The business requirements survey results summarize the technical pain points, business objectives, and analytical needs faced by the autonomous sensing system in actual operation. For example, technicians need to analyze the drift patterns of positioning accuracy under different weather conditions (such as rain, snow, and fog) and different route environments (such as tunnels, slopes, and curves); they need to diagnose whether the decline in the recognition rate of traffic signals or signs is related to camera imaging quality, changes in lighting, or material aging; and they need to quantitatively analyze the specific situations of missed and false alarms in obstacle detection systems under adverse weather conditions.

[0034] The business data survey results are an inventory of the data assets required to support business needs. It details the various multi-source heterogeneous data generated during the operation of the urban rail vehicle-mounted autonomous sensing system. "Multi-source" means the data originates from multiple independent systems both inside and outside the urban rail vehicle-mounted autonomous sensing system. For example, data may come from the Autonomous Train Positioning System (ITE) used for positioning, the Intelligent Train Detection System for obstacle detection, the signaling system providing ground truth, LiDAR and cameras collecting environmental information, and electronic maps providing basic line information. "Heterogeneous" refers to the significant differences in format, structure, temporal granularity, and semantics among these data from different sources. For example, ITE logs may be structured text files, point cloud data is a large collection of unordered 3D coordinate points, image data is unstructured binary files, and signal data may come from a real-time database.

[0035] Step 102: Based on the results of the business needs survey, the data sources of multi-source heterogeneous data, and the data interface of the urban rail vehicle-mounted autonomous sensing system, determine the conceptual model of sensing data.

[0036] The conceptual model of perceived data is a macroscopic, abstract model that is independent of specific technical implementations. It is a conceptual model used to guide the design of the physical model of perceived data and ensure the accuracy of the physical model of perceived data.

[0037] Step 103: Determine the physical model of the sensing data based on the conceptual model of the sensing data, the data interface of the urban rail vehicle-mounted autonomous sensing system, multi-source heterogeneous data, and the media information of the storage medium used.

[0038] The media information can refer to a specific database product or data warehouse platform, such as MaxCompute, or other relational databases (such as PostgreSQL), NoSQL databases, or big data platforms.

[0039] In some embodiments, media information includes its storage model (row-based or column-based), computing model (single-machine or distributed), whether it supports partitioning, and index type. This attribute information influences the design decisions of the physical model. Its powerful computing capabilities and column-based storage characteristics can be leveraged to clean, transform, and correlate heterogeneous data, ultimately forming a unified, standardized wide table structure. This physically transforms data with different structures into data with a unified structure, enabling jointly queried and analyzed data that was previously impossible to analyze.

[0040] For example, if the storage medium is a big data platform that supports columnar storage and partitioning, the physical model can be designed to build wider tables to reduce join operations during queries, and partition by time or business identifiers to improve data management and query efficiency.

[0041] In addition, the data interface of the urban rail vehicle-mounted autonomous sensing system and multi-source heterogeneous data were referenced in the process of determining the physical model of the sensing data to ensure that the physical model of the sensing data is compatible with the data interface of the urban rail vehicle-mounted autonomous sensing system and is suitable for processing multi-source heterogeneous data.

[0042] Step 104: Construct a data model for closed-loop autonomous sensing data of urban rail vehicles based on the physical model of sensing data.

[0043] In this embodiment, a unified, standardized, and regulated data model is designed by conducting relevant research on the urban rail vehicle-mounted autonomous sensing system. This model is used for data analysis and data mining of the urban rail vehicle-mounted autonomous sensing system, thereby improving the detection capability and environmental relationship diagnosis capability of the urban rail vehicle-mounted autonomous sensing system.

[0044] The business requirements survey results were obtained through research, the conceptual model of the sensing data was determined based on the business requirements survey results, the physical model of the sensing data was constructed based on the conceptual model of the sensing data, and the data model for the closed loop of autonomous sensing data for urban rail vehicles was determined with reference to the physical model of the sensing data. As can be seen from the above, the data model for the closed loop of autonomous sensing data for urban rail vehicles actually references the business requirements survey results. The business requirements survey results are a summary of the technical pain points, business objectives, and analytical needs faced in actual operation. Therefore, the data model for the closed loop of autonomous sensing data for urban rail vehicles can, to a certain extent, represent the technical pain points, business objectives, and analytical needs faced in actual operation. Thus, the data model for the closed loop of autonomous sensing data for urban rail vehicles can effectively improve the detection capability and environmental relationship diagnostic capability of the autonomous sensing system for urban rail vehicles.

[0045] By taking four steps—obtaining comprehensive survey results, constructing a conceptual model for top-level design, determining a physical model to unify heterogeneous data, and finally constructing a workable data closed-loop model—the detection capabilities and environmental relationship diagnostic capabilities of the urban rail vehicle-mounted autonomous sensing system are effectively improved.

[0046] For example, through this unified and integrated data model, technicians can easily perform complex correlation queries. For instance, they can quickly filter out all cases of signal recognition failures occurring in curved sections during nighttime rainy conditions and retrieve relevant images, point clouds, and log data. This greatly enhances the diagnostic capabilities for environmental relationships.

[0047] In some embodiments, the perceptual data conceptual model includes data entities, data analysis topics, and a common conceptual data model layer.

[0048] like Figure 2As shown, based on the results of business needs surveys, the data sources of multi-source heterogeneous data, and the data interface of the urban rail vehicle-mounted autonomous sensing system, a conceptual model of sensing data is determined, including: Step 201: Determine the data entity based on the data source of multi-source heterogeneous data.

[0049] Data entities include one or more of the following: positioning systems, signaling systems, sensing systems, and electronic maps.

[0050] The goal of this step is to extract the core data source categories, or data entities, from a complex pool of data sources, working from the bottom up. These data entities are logical abstractions and classifications of the various business systems or modules that provide the data. This process enables the integration and processing of multi-source heterogeneous data, thereby empowering the urban rail vehicle's autonomous sensing system with data capabilities.

[0051] Specifically, a positioning system represents a system that provides information on the spatial position and attitude of a train, such as ITE (Simultaneous Localization and Mapping) technology. The data it generates primarily includes high-frequency train pose logs and positioning error information.

[0052] A signaling system represents a system that provides ground truth information, such as transponders, track circuits, or communication-based train control systems. The data it generates can include the train's precise location (e.g., via transponder messages), the status of the preceding signal lights, switch status, and so on.

[0053] A perception system represents a collection of sensors that interact directly with the external environment, along with their initial processing units, such as onboard LiDAR, long- and short-focus cameras, and millimeter-wave radar. The data it generates primarily consists of raw or pre-processed point cloud data, image data, and obstacle lists.

[0054] Electronic maps represent systems that provide static geographic information about railway lines. They contain data on the line's topology, such as line codes, directions, link definitions, and the precise locations of fixed facilities like ramps, curves, switches, and platforms.

[0055] Step 202: Determine the data analysis topic based on the results of the business needs survey.

[0056] The goal of this step is to identify, from a top-down perspective, the core areas requiring analysis and exploration—the data analysis themes. These themes stem directly from business needs surveys, translating macro-level business objectives into specific data analysis domains. By defining these data analysis themes, the specific business scenarios that the data model needs to support are clarified, making the model design more targeted.

[0057] Step 203: Determine the common concept data model layer of related data analysis topics and data entities based on the data interface.

[0058] This step is central to building the conceptual model, aiming to establish a pivotal layer—the common conceptual data model layer. This layer's role is to effectively connect the bottom-up identified data entities with the top-down defined data analysis themes by analyzing the potential relationships between different data entities at the data interface level.

[0059] By constructing such a common conceptual data model layer, it logically connects the originally isolated data entities, enabling cross-entity data to serve specific analytical topics at the upper level.

[0060] In this embodiment, the conceptual model of perceived data is concretized into data entities, data analysis topics, and a common conceptual data model layer, and its construction is defined in three sub-steps, making the design process of the conceptual model more structured and systematic. This design ensures that the final data model accurately reflects the complex relationship between business needs and data sources, thereby effectively improving diagnostic capabilities.

[0061] In some embodiments, the data interface includes information such as data timestamps, encoding identifiers, and coordinate systems that can be used for correlation. Using this interface information, the logic for data fusion and correlation is designed at the conceptual level. For example: To support the positioning error analysis topic, the common concept data model layer defines a logic that associates the pose data generated by the positioning system entity with the true position data of the transponder provided by the signal system entity through timestamps.

[0062] To support the signal recognition and analysis topic, the common concept data model layer defines a logic that associates the image data generated by the sensing system entity with the train position data of the positioning system entity and the signal status data of the signal system entity through timestamps and spatial location (pose).

[0063] In some embodiments, the public concept data model layer is also used to perform one or more of the following processes: Alignment of sensing data collected by the sensing system with signal log data in the signal system, alignment of positioning log data in the positioning system with signal data output by the signal system, and matching of sensing data with pose data.

[0064] Typically, when engineers analyze specific problems, they need to manually align data from different sources using timestamps to reproduce the problem scenario before they can conduct further on-site analysis. This process is highly manual and time-consuming.

[0065] In this embodiment, by pre-defining these alignment and matching logics, it is ensured that the subsequent physical model design and data development can accurately achieve the fusion of cross-system data. Only by completing these basic data alignment and matching can higher-level analysis topics (such as positioning error analysis, recognition rate analysis, etc.) be supported, and ultimately the detection capability and environmental relationship diagnosis capability of the urban rail vehicle-mounted autonomous sensing system be improved.

[0066] Specifically, this involves aligning the sensor data collected by the sensing system with the signal log data from the signal system. When analyzing a case of signal recognition failure, it's necessary to see not only the image captured by the camera (from the sensing system) but also the actual light color displayed on the signal (from the signal log data) to determine whether it was a false alarm or a missed alarm. However, in its original state, these two sets of data are completely independent. To address this need, the common concept data model layer logically defines an alignment process. The core idea of ​​this process is to use timestamps as a common association key. The image data stream with timestamps generated by the sensing system is matched temporally with the signal status log stream with timestamps generated by the signal system. The goal is to find the signal status record that is closest to the timestamp of each frame of the image (or within a certain time window), thereby logically binding the image with the signal light color at that time.

[0067] The alignment of positioning log data in the positioning system with the signal data output by the signaling system also relies on timestamps for correlation. Specifically, it matches the train pose (coordinates and attitude) output by the positioning system at a certain point in time with the event recorded by the signaling system at almost the same time when the train passes a transponder. Through this alignment, the deviation between the position output by the positioning system and the true position when the train passes the transponder (the true point) can be calculated, thereby quantifying the positioning error.

[0068] Regarding the matching of sensing data and pose data, timestamps are still used as the key link to precisely match each frame of data (point cloud frame or image frame) generated by the sensing system (such as LiDAR, camera) with the train pose data generated by the positioning system at the same moment. Logically, this is equivalent to giving each piece of sensing data a spatiotemporal stamp, assigning it precise coordinates in the track environment.

[0069] In some embodiments, the data analysis topics include one or more of the following: Positioning error analysis, sign recognition analysis, traffic signal recognition analysis, obstacle recognition analysis, and perception data annotation.

[0070] In this embodiment, the data analysis topic and the business needs survey results are in one-to-one correspondence. Specifically, the business needs survey results include one or more of the following questions: Positioning accuracy, signal recognition, sign recognition, obstacle detection, and multi-source data fusion.

[0071] Regarding positioning accuracy, the vehicle-mounted autonomous positioning system uses SLAM technology for positioning. During the positioning process, the positioning accuracy of SLAM is greatly affected by the dynamic changes of the surrounding environment, and this impact is difficult to predict accurately through algorithm models. Therefore, to achieve more accurate positioning, it is necessary to collect positioning data generated by the system during the actual operation of the train in various environments (sunny weather, strong winds, rain and snow, tunnels, uphill and downhill, etc.). Then, by statistically analyzing this large amount of data, the positioning error is evaluated, the specific factors affecting the positioning error are identified, and the algorithm of the vehicle-mounted autonomous positioning system is adjusted to improve the positioning accuracy of the system.

[0072] Based on this, a positioning error analysis was designed to quantitatively evaluate the positioning accuracy of the vehicle-mounted autonomous positioning system and to delve into the relevant factors that cause the error.

[0073] For example, positioning error analysis includes: 1) Error analysis based on key location segments: Analyze the positioning data of trains when passing through certain special or critical sections (such as near signs, platform areas, switch areas, and curve areas), and calculate the deviation from the preset geographical location (true value). Through statistical analysis, it can be determined whether specific types of geographical environments will systematically affect positioning accuracy.

[0074] 2) Transponder-based pose deviation analysis: The transponder is a point on the track with known absolute coordinates. This analysis calculates the real-time positioning error by comparing the pose data output by the positioning system when the train passes the transponder with the precise position of the transponder.

[0075] Regarding traffic signal recognition in the business requirements survey results: Traffic signal recognition performance is highly dependent on camera image quality, which in turn depends on the camera manufacturer (different manufacturers produce cameras with varying image quality). Even for the same model from the same manufacturer, image quality can be limited by the installation angle. Furthermore, the image quality of a mounted camera varies under different environments (such as lighting, fog, rain, and snow). These realities may lead to less than ideal versatility and stability of the system under different conditions. Currently, the lack of a comprehensive dataset covering various conditions limits the system's adaptability to traffic signal recognition in different scenarios. To improve the adaptability of traffic signal recognition in diverse environments, it is necessary to collect data on traffic signal recognition in various scenarios, classify and analyze this data, identify influencing factors, and pinpoint problems, thereby improving the system's traffic signal recognition capabilities.

[0076] Based on this, a signal recognition analysis is designed to evaluate and diagnose the system's performance in recognizing signals and their light color states.

[0077] Specifically, the signal recognition analysis includes statistically analyzing the recognition accuracy of signal light colors (red, yellow, green, etc.) under different distances, lighting conditions, and weather backgrounds for different camera types (such as telephoto cameras and short-focus cameras).

[0078] For example, technicians can use the model built on this topic to quickly query all cases where telephoto cameras misidentified yellow traffic lights as red during the evening hours and retrieve relevant images for in-depth analysis.

[0079] Regarding the sign recognition aspect of the business requirements survey, the performance of traffic signal recognition is highly dependent on the camera's image quality. However, camera image quality is highly dependent on the camera manufacturer (different manufacturers produce cameras with varying image quality). Even for the same model from the same manufacturer, image quality can be limited by the installation angle. Furthermore, the image quality of an installed camera varies under different environments (such as lighting conditions, fog, rain, and snow). These realities may lead to less than ideal versatility and stability of the system under different conditions. Currently, the lack of a comprehensive dataset covering various conditions limits the system's adaptability to traffic signal recognition in different scenarios. To improve the adaptability of traffic signal recognition in diverse environments, it is necessary to collect data on traffic signal recognition in various scenarios, classify and analyze this data, identify influencing factors, and pinpoint problems, thereby improving the system's traffic signal recognition capabilities.

[0080] Based on this, a sign recognition analysis was designed to evaluate and diagnose the system's recognition performance for various types of signs (including coded and uncoded signs) along the track.

[0081] Specifically, statistical analysis is conducted to assess the system's success rate and failure rate in recognizing signs under different time periods (day / night), weather conditions (sunny / rainy / snowy), and train speeds. Further analysis of failed recognition cases is possible; for example, by correlating image data, it can be determined whether the failure was due to aging or contamination of the sign itself, or excessively strong or weak ambient light. This helps diagnose the causes of declining recognition capabilities and guides improvements to the recognition algorithm or the sign material.

[0082] Regarding obstacle detection in the business requirements survey, the current ITE and train intelligent detection systems experience a decline in recognition and obstacle detection capabilities under adverse weather conditions (rain, snow, heavy fog, low visibility, etc.). This can lead to missed and false alarms during obstacle detection. There is insufficient data for quantitative analysis of the relationship between the system's detection capability and the surrounding environment, and a lack of appropriate data analysis and mining methods. To improve the system's detection and obstacle detection capabilities, more data needs to be collected and statistically analyzed to determine the impact of the environment on the system.

[0083] Based on this, an obstacle recognition analysis was designed to evaluate the system's ability to detect obstacles ahead of the track in complex environments, with a focus on missed detections (obstacles exist but are not detected) and false alarms (no obstacles are detected but false alarms are triggered).

[0084] Specifically, the false alarm rate and false alarm rate of obstacle detection are statistically analyzed under different weather conditions (such as rain, snow, and fog leading to low visibility). By comparing the system's detection results with manually labeled ground truth values ​​or cross-validation data from other sensors, the impact of environmental factors on obstacle detection capabilities is quantified, thereby providing data support for robust optimization of the algorithm.

[0085] Regarding the multi-source data fusion aspect of the business needs survey, the current video and point cloud data only have limited timestamp information, lacking ground truth information such as location, speed, and light color. To retrieve data based on location, light color, or other criteria, the data needs to be manually tagged, a process highly dependent on manual execution.

[0086] Based on this, we designed a sensory data annotation system to automatically and batch-add rich, structured contextual information labels to massive amounts of unstructured raw sensory data (such as point cloud frames and images).

[0087] This embodiment clearly demonstrates how the data model translates business requirements into actionable analytical domains by listing specific data analysis topics. It is through the support of these topics that the model helps technical personnel gain a deeper understanding of the relationship between system performance and various complex factors, thereby effectively guiding system optimization and iteration, ultimately achieving the goal of improving the detection capabilities and environmental relationship diagnosis capabilities of urban rail vehicle-mounted autonomous sensing systems.

[0088] In some embodiments, such as Figure 3 As shown, the perceptual data conceptual model includes data sources (i.e., data entities in this invention), analysis topics (i.e., data analysis topics in this invention), and a public conceptual model layer (i.e., public conceptual data model layer in this invention).

[0089] The data sources include the ITE system, signaling system, sensing system, electronic map data, and other data. Specifically, the ITE system includes safety board data and signal light board data; the signaling system includes train position information and signal information; the sensing system includes point cloud information and image information; the electronic map data includes sections, platforms, switches, curves, gradients, tunnels, coding plates, signs, signals, and transponders; and other data includes time periods, seasons, weather, and system clock differences.

[0090] The analysis topics include positioning error analysis, sign recognition analysis, traffic signal recognition analysis, obstacle recognition analysis, and perception data annotation. Positioning error analysis includes vehicle passage through key areas and vehicle passage past transponders; sign recognition includes coded sign recognition analysis and non-coded sign recognition analysis; traffic signal recognition includes long-focus camera recognition analysis and short-focus camera recognition analysis; obstacle recognition includes obstacle identification analysis; and perception data annotation includes point cloud data labeling and image data labeling.

[0091] The common concept model layer includes data on the train passing through key sections, detailed data under clock alignment, and fused perception and ITE data. Specifically, the data on the train passing through key sections includes data on the front and rear of the train passing through key sections. The detailed data under clock alignment includes detailed data on the safety panel and the lighting panel. The fused perception and ITE data includes point cloud data and safety panel data fusion, and image data and lighting panel data fusion.

[0092] In some embodiments, the perceptual data annotation includes: Add one or more of the following tags to your point cloud and image data: Time period, season, location, gradient, curve, turnout, platform, transponder, signal, signage.

[0093] In this embodiment, each frame of image or point cloud data is automatically labeled with tags such as time period (early morning / noon / evening), season, location (tunnel / open-air), and route features (slope / curve / turnout) based on its matched pose data and time information. This automated labeling greatly improves the usability of the data, enabling technicians to perform efficient data retrieval based on these tags. For example, it can quickly filter out all point cloud data collected in winter, at tunnel exits, and on downhill sections, providing massive datasets for algorithm development and testing in specific scenarios.

[0094] In some embodiments, the perceived data physical model includes a raw layer, a unified data warehouse layer, and an application layer.

[0095] The physical model of the sensing data is determined based on the conceptual model of the sensing data, the data interface of the urban rail vehicle-mounted autonomous sensing system, multi-source heterogeneous data, and the media information of the storage medium used, including: Deployment of the primary layer, unified data warehouse layer, and application layer based on media information; The original layer is based on the data interface and multi-source heterogeneous data configuration of the urban rail vehicle-mounted autonomous sensing system. Configure a unified data warehouse layer based on the conceptual model of perceived data.

[0096] Among them, such as Figure 4 As shown, the raw data layer (ODS), unified data warehouse layer (DW), and application layer (ADS) are distributed according to the principle of horizontal layering. The unified data warehouse layer (DW) includes the detailed data layer (DWD), the summary data layer (DWS), and the dimensional data layer (DIM). The raw data layer (ODS) mainly parses, aggregates, and integrates raw log (ITE, sensing, signal) data into MAXCOMPUTE, providing data sources for the unified data warehouse layer (DW). The unified data warehouse layer (DW) organizes data according to business themes and business processes, defines consistency indicators and dimensions, and each business theme is independently constructed with unified specifications to form a unified and standardized standard business data system. The application layer (ADS) is oriented towards the needs of final business applications, obtains data from the unified data warehouse layer, and processes business-specific data to meet the specific needs of the business.

[0097] Table 1 shows the data structures of the raw layer ODS, the unified data warehouse layer DW, and the application layer ADS.

[0098] Table 1

[0099] In some embodiments, the unified data warehouse layer includes a spatial dimension table, a train location details table, and a first detailed data segment table, the first detailed data segment table including detailed data segments of the train passing through each section.

[0100] Receive update operations on the spatial dimension table; In response to the update operation, the first detailed data fragment table is updated based on the train location details table and the updated spatial dimension table.

[0101] Since the most basic unit of a subway line is composed of links and offsets, when a train passes through a specific area, this area can be divided into links and offsets, such as a platform area (including the starting link and offset, intermediate link and offset, and ending link and offset). A corresponding dimension table is then designed, containing all links corresponding to the platform area, as well as the starting and ending offsets of each link. In the signal log (train position data), the train's end (front or rear) will have corresponding head or tail link and offset during operation. Once the train's position table (link and offset) enters the platform area (containing all link and offset information in the platform dimension table), the corresponding train position data is tagged with platform information (generating a corresponding information table of the sections the train has passed through).

[0102] Based on this, a spatial dimension table, a train location detail table, and a first detail data segment table are constructed. The first detail data segment table includes detail data segments of the train passing through each section, which is also the corresponding information table of the train passing through each section.

[0103] As shown in Table 1, the spatial dimension table includes one or more of the following: Section dimension table, platform dimension table, turnout dimension table, curve dimension table, gradient dimension table, signage dimension table, signal dimension table, transponder dimension table.

[0104] In this embodiment, based on the characteristics of onboard autonomous perception services, the train will pass through key areas that are of concern to technicians during operation. These key areas include one or more of the following: passing signs (coded signs, uncoded signs), passing signals (long-focus cameras, short-focus cameras), passing transponders, passing platforms, passing switches, passing curves, and passing slopes.

[0105] Based on this, we construct dimension tables for the aforementioned key areas, including interval dimension tables, platform dimension tables, turnout dimension tables, curve dimension tables, gradient dimension tables, signage dimension tables, signal dimension tables, and transponder dimension tables.

[0106] Table 2 shows the data structure of the platform dimension table.

[0107] Table 2

[0108] Table 3 shows the data structure of the train location details table.

[0109] Table 3

[0110] Table 4 shows the data structure of the first detailed data segment table.

[0111] Table 4

[0112] As shown in Tables 2, 3 and 4, in the signal log (train position data), the train end (head or tail) will have a corresponding head or tail link+offset during operation. Once the train position table end (link+offset) enters the platform area (all link+offset information contained in the platform dimension table), the corresponding train position data is tagged with platform information (generating the corresponding train end through each section information table).

[0113] Based on this, when adding or removing key sections of interest, it is only necessary to add or remove the corresponding dimension tables, and add or remove the corresponding section dimension label information in the information table of each section passed by the train (front or rear). At the same time, the algorithm logic for generating the relevant dimension labels of the train passing through this section when adding a section of interest is consistent with the algorithm logic for adding a platform. Thus, the plug-and-play design pattern is used in the autonomous perception data model.

[0114] In some embodiments, the data model design method includes: The data model design apparatus provided by the present invention is described below. The data model design apparatus described below and the data model design method described above can be referred to in correspondence.

[0115] In one embodiment, such as Figure 5 As shown, a data model design apparatus is proposed, comprising: The acquisition module 501 is used to acquire the survey results of the urban rail vehicle-mounted autonomous perception system. The survey results include the business demand survey results and the business data survey results. The business data survey results include multi-source heterogeneous data generated during the operation of the urban rail vehicle-mounted autonomous perception system. The first determining module 502 is used to determine the conceptual model of sensing data based on the results of business needs surveys, the data sources of multi-source heterogeneous data, and the data interface of the urban rail vehicle-mounted autonomous sensing system. The second determining module 503 is used to determine the physical model of the sensing data based on the conceptual model of the sensing data, the data interface of the urban rail vehicle-mounted autonomous sensing system, multi-source heterogeneous data, and the media information of the storage medium used. Module 504 is used to build a data model for constructing a closed loop of autonomous sensing data for urban rail vehicles based on the physical model of sensing data.

[0116] Figure 6 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 6 As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630, and a communication bus 640, wherein the processor 610, the communication interface 620, and the memory 630 communicate with each other through the communication bus 640. The processor 610 can call logical instructions in the memory 630 to execute a data model design method, which includes: obtaining the survey results of the urban rail vehicle-mounted autonomous sensing system, including business requirement survey results and business data survey results, including multi-source heterogeneous data generated during the operation of the urban rail vehicle-mounted autonomous sensing system; determining a conceptual model of sensing data based on the business requirement survey results, the data sources of the multi-source heterogeneous data, and the data interface of the urban rail vehicle-mounted autonomous sensing system; determining a physical model of sensing data based on the conceptual model of sensing data, the data interface of the urban rail vehicle-mounted autonomous sensing system, the multi-source heterogeneous data, and the media information of the storage medium used; and constructing a data model for the closed loop of urban rail vehicle-mounted autonomous sensing data based on the physical model of sensing data.

[0117] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0118] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the data model design method provided by the above methods. The method includes: obtaining the survey results of the urban rail vehicle-mounted autonomous sensing system, the survey results including business requirement survey results and business data survey results, the business data survey results including multi-source heterogeneous data generated during the operation of the urban rail vehicle-mounted autonomous sensing system; determining a conceptual model of sensing data based on the business requirement survey results, the data sources of the multi-source heterogeneous data, and the data interface of the urban rail vehicle-mounted autonomous sensing system; determining a physical model of sensing data based on the conceptual model of sensing data, the data interface of the urban rail vehicle-mounted autonomous sensing system, the multi-source heterogeneous data, and the media information of the storage medium used; and constructing a data model for the closed loop of urban rail vehicle-mounted autonomous sensing data based on the physical model of sensing data.

[0119] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the data model design method provided by the above methods. This method includes: acquiring survey results of an urban rail vehicle-mounted autonomous sensing system, the survey results including business requirement survey results and business data survey results, the business data survey results including multi-source heterogeneous data generated during the operation of the urban rail vehicle-mounted autonomous sensing system; determining a conceptual model of sensing data based on the business requirement survey results, the data sources of the multi-source heterogeneous data, and the data interface of the urban rail vehicle-mounted autonomous sensing system; determining a physical model of sensing data based on the conceptual model of sensing data, the data interface of the urban rail vehicle-mounted autonomous sensing system, the multi-source heterogeneous data, and the media information of the storage medium used; and constructing a data model for a closed-loop system of urban rail vehicle-mounted autonomous sensing data based on the physical model of sensing data.

[0120] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0121] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0122] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A data model design method, characterized in that, include: Obtain the survey results of the urban rail vehicle-mounted autonomous perception system. The survey results include the business demand survey results and the business data survey results. The business data survey results include multi-source heterogeneous data generated during the operation of the urban rail vehicle-mounted autonomous perception system. Based on the business requirements survey results, the data sources of the multi-source heterogeneous data, and the data interface of the urban rail vehicle-mounted autonomous perception system, a conceptual model of perception data is determined. The physical model of the sensing data is determined based on the conceptual model of the sensing data, the data interface of the urban rail vehicle-mounted autonomous sensing system, the multi-source heterogeneous data, and the media information of the storage medium used. Based on the aforementioned physical model of sensing data, a data model for closed-loop autonomous sensing data of urban rail vehicles is constructed.

2. The data model design method according to claim 1, characterized in that, The perceptual data conceptual model includes data entities, data analysis topics, and a common conceptual data model layer; The conceptual model for determining sensing data based on the business requirements survey results, the data sources of the multi-source heterogeneous data, and the data interface of the urban rail vehicle-mounted autonomous sensing system includes: The data entity is determined based on the data source of the multi-source heterogeneous data, and the data entity includes one or more of the following: positioning system, signal system, sensing system, and electronic map; The data analysis topic is determined based on the results of the business needs survey. Based on the data interface, the common concept data model layer that associates the data analysis topic and the data entity is determined.

3. The data model design method according to claim 2, characterized in that, The public concept data model layer is also used to perform one or more of the following processes: Alignment of sensing data collected by the sensing system with signal log data in the signal system, alignment of positioning log data in the positioning system with signal data output by the signal system, and matching of sensing data with pose data.

4. The data model design method according to claim 2, characterized in that, The data analysis topics include one or more of the following: Positioning error analysis, sign recognition analysis, traffic signal recognition analysis, obstacle recognition analysis, and perception data annotation.

5. The data model design method according to claim 4, characterized in that, The sensory data annotation includes: Add one or more of the following tags to your point cloud and image data: Time period, season, location, gradient, curve, turnout, platform, transponder, signal, signage.

6. The data model design method according to any one of claims 1 to 5, characterized in that, The physical model of the perceived data includes a raw layer, a unified data warehouse layer, and an application layer; The determination of the physical model of the sensing data based on the conceptual model of the sensing data, the data interface of the urban rail vehicle-mounted autonomous sensing system, the multi-source heterogeneous data, and the media information of the storage medium used includes: The original layer, the unified data warehouse layer, and the application layer are deployed based on the media information. The original layer is configured based on the data interface of the urban rail vehicle-mounted autonomous sensing system and the multi-source heterogeneous data. Configure the unified data warehouse layer based on the aforementioned conceptual model of perceived data.

7. The data model design method according to claim 6, characterized in that, The unified data warehouse layer includes a spatial dimension table, a train location detail table, and a first detail data segment table, the first detail data segment table including detail data segments of each section through which the train sets pass. Receive update operations on the spatial dimension table; In response to the update operation, the first detailed data fragment table is updated based on the train location details table and the updated spatial dimension table.

8. A data model design apparatus, characterized in that, include: The acquisition module is used to acquire the survey results of the urban rail vehicle-mounted autonomous perception system. The survey results include business demand survey results and business data survey results. The business data survey results include multi-source heterogeneous data generated during the operation of the urban rail vehicle-mounted autonomous perception system. The first determining module is used to determine the conceptual model of the sensing data based on the results of the business demand survey, the data sources of the multi-source heterogeneous data, and the data interface of the urban rail vehicle-mounted autonomous sensing system. The second determining module is used to determine the physical model of the sensing data based on the conceptual model of the sensing data, the data interface of the urban rail vehicle-mounted autonomous sensing system, the multi-source heterogeneous data, and the media information of the storage medium used. The construction module is used to build a data model for the closed loop of autonomous sensing data on urban rail vehicles based on the physical model of the sensing data.

9. 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 data model design method as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the data model design method as described in any one of claims 1 to 7.

Images