A multi-dimensional intelligent inspection system for a subway train in a train interior is fused with edge computing

Abstract

The application discloses a kind of fusion edge computing subway train in-car multidimensional intelligent inspection systems, including vehicle-mounted edge computing module, fixed perception module, inspection robot module, cloud platform management module and alarm module;The vehicle-mounted edge computing module is respectively connected with fixed perception module, inspection robot module, and communication connection is established, for receiving the perception data collected by both, carries out defect identification, feature extraction and data fusion, and the structured data after processing is uploaded to cloud platform management module;The data collected by the fixed perception module is transmitted to the vehicle-mounted edge computing module in real time;The inspection robot module is communicated with the vehicle-mounted edge computing module;The cloud platform management module is deployed in subway operation control center, and uses "cloud-edge-end" collaborative architecture;The alarm module is communicated with the vehicle-mounted edge computing module and cloud platform management module by vehicle-mounted local area network and wireless communication network, and realizes data interaction.

Description

Technical Field

[0001] This invention relates to the field of subway train operation and maintenance inspection technology, specifically a multi-dimensional intelligent inspection system for subway train interiors that integrates edge computing. Background Technology

[0002] With the continuous growth of urban rail transit operating mileage in my country, subways have become the core carrier of urban public transportation. As the core facilities that passengers directly come into contact with, the condition of subway train interior components directly affects the passenger's riding experience, travel safety, and the train's operational image. These components mainly include seats, handrails, door interior panels, walls, ceilings, seat covers, headrests, and handrail covers. They must meet the basic requirements of structural stability, reasonable function, and safety and reliability. Therefore, when defects occur, they need to be dealt with in a timely manner.

[0003] Currently, defect detection of interior components in subway trains mainly relies on traditional manual inspection, which has many insurmountable drawbacks: inspections cannot be carried out during passenger operation and can only be conducted manually during non-operational windows after the train is returned to the depot, resulting in untimely defect detection. Some minor defects may accumulate over time and develop into serious safety hazards, affecting operational safety. Manual inspection is affected by subjective factors such as the fatigue level, sense of responsibility, and professional level of the operators, resulting in a high rate of missed and false detections. In particular, it is difficult to achieve comprehensive coverage of defects in hidden parts and blind spots. Moreover, manual inspection is inefficient and cannot meet the maintenance needs of large-scale subway lines. At the same time, working late at night further exacerbates the risk of missed detections.

[0004] To address the shortcomings of manual inspections, some subway lines have attempted to deploy fixed monitoring equipment for in-car monitoring. However, existing monitoring systems mostly employ a centralized data processing model, which presents significant technical bottlenecks: Firstly, fixed cameras have blind spots, failing to cover all areas inside the carriage, especially hidden areas such as under seats, handrail connections, and corners, making it difficult to comprehensively detect defects in interior components. Secondly, high-definition monitoring video data is large in volume; if all of it is transmitted to the cloud for processing, it will cause excessive network bandwidth pressure and high data transmission latency, making it impossible to achieve real-time defect identification and alarms, and posing a risk of data leakage, which does not meet the data security requirements of rail transit.

[0005] Edge computing technology, as a new paradigm that brings AI computing power down to the source of data generation, can provide data processing and AI inference capabilities at the network edge, reducing response time to milliseconds, effectively alleviating network pressure, improving data privacy, and providing a solution for intelligent inspection of rail transit. Currently, edge computing has been initially applied in some rail transit scenarios, but an integrated intelligent inspection solution for detecting defects in subway train interior components has not yet been formed. Existing technologies either rely solely on fixed sensing devices, failing to compensate for blind spots, or deploy inspection robots only during off-peak hours, failing to achieve real-time monitoring during operation. Furthermore, they do not fully integrate the advantages of edge computing, resulting in inspection efficiency and detection accuracy that are insufficient to meet actual operation and maintenance needs.

[0006] In addition, existing interior component defect detection methods mostly target single-type defects and have not formed a multi-dimensional, all-scenario detection system. They cannot fully cover various issues such as appearance defects, size defects, and functional defects. Furthermore, they lack a collaborative mechanism for real-time monitoring during operation and comprehensive investigation during non-operational periods, making it difficult to achieve full lifecycle management of interior component defects.

[0007] Therefore, developing a multi-dimensional intelligent inspection system for subway trains that integrates edge computing and can cover all time periods, both during and outside of train operation, with high precision, has become a pressing technical problem to be solved in the current subway operation and maintenance field. Summary of the Invention

[0008] The purpose of this invention is to provide a multi-dimensional intelligent inspection system for subway train interiors that integrates edge computing, aiming to improve the problems of low efficiency, high rate of missed and false detection, incomplete coverage, poor real-time performance, and inability to meet the inspection needs of both running and non-running periods in the existing technology for subway train interior component defect detection.

[0009] This invention is implemented as follows:

[0010] A multi-dimensional intelligent inspection system for subway trains that integrates edge computing includes an onboard edge computing module, a fixed sensing module, an inspection robot module, a cloud platform management module, and an alarm module.

[0011] The vehicle-mounted edge computing module establishes communication connections with the fixed sensing module and the inspection robot module, respectively.

[0012] The fixed sensing modules are deployed in key locations inside the subway train carriages, including both ends of the carriage, the middle of the carriage, above the seats, and both sides of the doors.

[0013] The inspection robot module is deployed in the subway depot and is used during train return and non-operational periods.

[0014] The cloud platform management module is deployed in the subway operation control center and interacts with the on-board edge computing module through a wireless communication network.

[0015] The alarm module interacts with the vehicle edge computing module and cloud platform management module through the vehicle local area network and wireless communication network.

[0016] Preferably, the vehicle-mounted edge computing module is deployed in each subway train carriage, employing a lightweight edge computing server and incorporating a pruned and quantized optimized AI defect detection model. This AI defect detection model supports real-time local data processing, model inference, and result caching, eliminating the need to transmit all raw data to the cloud. The vehicle-mounted edge computing module also features built-in network interruption recovery and caching capabilities, allowing it to locally cache alarm events and critical data during temporary network outages, automatically synchronizing them to the cloud platform once the network is restored. Furthermore, the vehicle-mounted edge computing module incorporates a local rule engine, enabling independent analysis and alarm generation based on preset rules even when disconnected from the cloud.

[0017] Preferably, the fixed sensing module includes a high-definition infrared camera, a vibration sensor, and a sound sensor for all-weather real-time monitoring during train passenger operation. The high-definition infrared camera adopts a modular design, supports adaptive zoom and region of interest analysis, and only performs full-resolution acquisition of key areas of interior components, while other areas are analyzed at low resolution. The high-definition infrared camera is used to acquire image data of interior components inside the carriage, capturing surface defects such as scratches, damage, color differences, and stains, as well as structural defects such as loose armrests and detached interior panels. The vibration sensor is deployed at the connection points of armrests and seats to detect the installation firmness of interior components and capture vibration signals corresponding to looseness and abnormal noises. The sound sensor is used to collect abnormal sound signals generated by loose or damaged interior components to assist in defect identification.

[0018] Preferably, the inspection robot module includes an autonomous navigation unit, a near-field perception unit, a robot body, and a local control unit. The autonomous navigation unit adopts a fusion navigation technology of LiDAR and visual SLAM, pre-loads a map of the vehicle interior, supports autonomous obstacle avoidance, path planning, and precise positioning, and can autonomously travel along a preset inspection route within the vehicle interior to achieve full coverage inspection without blind spots. The near-field perception unit consists of a high-definition industrial camera, a laser rangefinder, and a supplementary light. The high-definition industrial camera can capture details of interior parts at close range and detect defects in the blind spots of the fixed perception module. The laser rangefinder is used to measure the dimensional parameters of assembly gaps and installation deviations of interior parts. The supplementary light is used to provide supplementary lighting for the high-definition industrial camera when there is insufficient light. The inspection robot module is equipped with an embedded AI computing box, which can perform preliminary defect identification in real time during the inspection process and only transmit the defect results and key images back.

[0019] Preferably, the robot body is provided with a control box, and the autonomous navigation unit and the local control unit are both integrated inside the control box; the robot body is provided with a robotic arm, and the output end of the robotic arm is provided with a mounting frame, and the high-definition industrial camera, laser rangefinder, and supplementary light of the near-field sensing unit are all mounted on the mounting frame.

[0020] Preferably, the robot body has wheels at its bottom corners, the control box has a connection slot on its side, and the upper surface of the control box has multiple control panels; the bottom of the robotic arm has a base with multiple fixing holes, and the robotic arm is fixed to the robot body by bolts passing through the fixing holes; the bottom side of the robotic arm has a connection line with a connection plug at its end, and the connection plug is connected to the connection slot; the output end of the robotic arm has a mounting base with multiple mounting slots.

[0021] Preferably, the mounting frame has a mounting post at the top center, a mounting plate at the top of the mounting post, and multiple mounting holes on the mounting plate; the mounting frame has a locking rod on the bottom surface, the locking rod has a T-shaped structure, a junction box is provided on one side of the mounting frame, and multiple male connectors and wire terminals are provided on the side of the junction box; the fill light, high-definition industrial camera, and laser rangefinder are all provided with locking connectors at the top, the locking connectors have a locking slot in the middle, the locking slot engages with the locking rod, and a clamping knob is provided on the side of the locking connector aligned with the locking slot; the fill light, high-definition industrial camera, and laser rangefinder are all provided with wires on the side, and female connectors are provided at the ends of the wires, which are respectively connected to the male connectors on the junction box.

[0022] Preferably, the cloud platform management module has a built-in defect database that records information such as the type, location, discovery time, and processing progress of interior component defects. It combines historical data to perform trend analysis, predict defect development patterns, and provide targeted maintenance suggestions for maintenance personnel. The cloud platform management module supports inspection task scheduling and can issue control commands for adjusting inspection parameters and robot inspection tasks to the on-board edge computing module based on train operation plans and interior component maintenance needs. The cloud platform management module supports continuous model training and optimization, summarizes defect samples uploaded by the on-board edge computing module, regularly updates the AI ​​defect detection model, and distributes the optimized model to each on-board edge computing module.

[0023] Preferably, the alarm module includes an in-vehicle local alarm unit and a cloud platform remote alarm unit. The in-vehicle local alarm unit is deployed inside the passenger compartment, and the cloud platform remote alarm unit establishes communication with the maintenance personnel's terminal. When the in-vehicle edge computing module identifies a serious defect, it immediately triggers the audible and visual alarm of the in-vehicle local alarm unit and simultaneously uploads the defect information to the cloud platform. The cloud platform triggers a remote alarm to notify the crew and maintenance personnel to handle the issue promptly. The cloud platform remote alarm unit pushes the defect information to the maintenance personnel's terminal via SMS and platform messages, specifying the defect location, type, and severity, and records the alarm handling status, forming a closed-loop management system.

[0024] Preferably, the inspection system includes an inspection mode during operation and an inspection mode during off-operation, achieving coordinated operation. During operation, when the train is carrying passengers, the fixed sensing module is activated and continuously collects image, vibration, and sound data of the interior components of the carriage, which is transmitted to the on-board edge computing module for real-time analysis and reasoning. Based on the severity of the defect, corresponding cache upload or alarm operations are performed. During off-operation, when the train is returned to the depot, the cloud platform management module issues inspection tasks, and the on-board edge computing module controls the inspection robot module to start, complete a full-coverage inspection and collect data, which is then analyzed by the on-board edge computing module to identify defects.

[0025] Compared with the prior art, the beneficial effects of the present invention are:

[0026] 1. This invention combines a fixed sensing module and an inspection robot module to carry out inspection work during train operation and non-operation periods. During operation, the fixed sensing module enables real-time monitoring around the clock, while the inspection robot enables full coverage and close-range inspection during non-operation periods. This effectively compensates for the blind spots of fixed cameras and solves the problem that traditional manual inspection and single monitoring modes cannot cover the entire scene, thus achieving comprehensive capture of defects in the interior parts of the carriage.

[0027] 2. This invention uses an in-vehicle edge computing module for local data processing, eliminating the need to transmit all raw data to the cloud, significantly reducing network bandwidth pressure, shortening defect identification response latency to the millisecond level, and enabling real-time identification and alarm of defects during runtime; at the same time, the in-vehicle edge computing module filters and performs preliminary analysis on the data, uploading only structured data and key information, improving data transmission and processing efficiency.

[0028] 3. This invention adopts an improved lightweight AI defect detection model, which combines multi-dimensional perception data of images, vibrations, and sounds to achieve accurate identification of appearance defects, structural defects, and dimensional defects of interior parts; the close-range perception and laser ranging of the inspection robot module further improve the detection accuracy of defects and dimensional defects in hidden parts. Combined with the collaborative verification during operation and non-operation periods, the missed detection rate and false detection rate are significantly reduced, which is far superior to the level of manual inspection.

[0029] 4. This invention achieves centralized management, trend analysis, and task scheduling of defect data through a cloud platform management module. Combined with continuous optimization of AI models, it enables predictive maintenance of defects and reduces ineffective inspection work. At the same time, it replaces traditional manual inspections, significantly reducing the intensity of manual labor and labor costs, especially reducing the safety risks of manual inspections at night, and improving the level of intelligent maintenance of subway train interiors.

[0030] 5. This invention, through real-time monitoring and timely alarms, can quickly detect and address serious defects that affect passenger safety, preventing the defects from escalating and causing safety accidents; at the same time, it can promptly handle minor defects, maintain the integrity of the interior components of the carriage, improve the passenger riding experience, and meet the safety and service requirements of subway operation. Attached Figure Description

[0031] Figure 1 This is a block diagram of the inspection system of the present invention;

[0032] Figure 2 This is an architectural block diagram of the vehicle-mounted edge computing module of the present invention;

[0033] Figure 3 This is an architectural block diagram of the fixed sensing module of the present invention;

[0034] Figure 4 This is a block diagram of the internal architecture of the inspection robot module of the present invention;

[0035] Figure 5 This is a block diagram of the architecture of the near-field sensing unit of the present invention;

[0036] Figure 6 This is a structural schematic diagram of the inspection robot module of the present invention;

[0037] Figure 7 This is a schematic diagram of the structure of the robot body of the present invention;

[0038] Figure 8 This is a schematic diagram of the structure of the robotic arm of the present invention;

[0039] Figure 9 This is a schematic diagram of the mounting bracket of the present invention;

[0040] Figure 10 This is a schematic diagram of the structure of the high-definition industrial camera of the present invention;

[0041] Figure 11 This is an architectural block diagram of the cloud platform management module of the present invention;

[0042] Figure 12 This is an architectural block diagram of the alarm module of the present invention.

[0043] In the diagram: 1. Robot body; 11. Wheel; 12. Control box; 13. Connecting slot; 14. Control panel; 2. Robotic arm; 21. Base; 22. Fixing hole; 23. Connecting cable; 24. Connecting plug; 25. Mounting base; 26. Mounting slot; 3. Mounting bracket; 31. Mounting column; 32. Mounting plate; 33. Mounting hole; 34. Clamping rod; 35. Male connector; 36. Terminal block; 4. Fill light; 5. High-definition industrial camera; 51. Clamping connector; 52. Bayonet; 53. Clamping knob; 54. Wire; 55. Female connector; 6. Laser rangefinder sensor. Detailed Implementation

[0044] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0045] The following description, in conjunction with the accompanying drawings and specific embodiments, provides further details:

[0046] Example 1

[0047] like Figure 1 and Figure 2As shown, this embodiment provides a multi-dimensional intelligent inspection system for subway trains that integrates edge computing. The system comprises an onboard edge computing module, a fixed sensing module, an inspection robot module, a cloud platform management module, and an alarm module. These modules interact stably via an onboard local area network and a wireless communication network. The onboard edge computing module establishes bidirectional communication connections with both the fixed sensing module and the inspection robot module. Deployed inside each subway train car, this module uses a lightweight edge computing server as its hardware carrier and incorporates a pruned and quantized optimized AI defect detection model. It receives various types of sensing data collected by the fixed sensing module and the inspection robot module, performs defect identification, feature extraction, and multi-source data fusion processing, and uploads the processed structured data to the cloud platform management module. Simultaneously, it can receive various commands from the cloud platform management module, such as those for adjusting inspection parameters and controlling equipment. The vehicle-mounted edge computing module has local real-time data processing, model inference, and result caching capabilities, eliminating the need to transmit all raw perception data to the cloud, effectively reducing network transmission pressure. It also has built-in functions for resuming data transmission after network outage and data caching. When the communication network is temporarily interrupted, alarm events and key defect data can be stored locally and automatically synchronized to the cloud platform management module after the network is restored. The module also has a local rule engine, which can independently complete data analysis and defect alarm operations according to preset rules even if the connection with the cloud is lost.

[0048] like Figure 1 and Figure 3 As shown, the fixed sensing module is installed at key locations such as both ends, the middle, above the seats, and both sides of the doors of the subway train car. It mainly consists of high-definition infrared cameras, vibration sensors, and sound sensors, and is used for all-weather real-time monitoring during the train's passenger operation. The sensing data it collects is transmitted to the on-board edge computing module in real time. The high-definition infrared camera in this module adopts a modular design, supporting adaptive zoom and intelligent region of interest analysis. It only performs full-resolution image acquisition on key areas of interior components, while low-resolution analysis is used for other areas. This effectively captures exterior images of interior components in the vehicle cabin, identifying surface defects such as scratches, damage, color differences, and stains, as well as structural defects such as loose armrests and detached interior panels. Vibration sensors are installed at the connection points of interior components such as armrests and seats to detect the firmness of the installation and capture vibration signals corresponding to looseness and abnormal noises. Sound sensors are used to collect abnormal sound signals generated by loose or damaged interior components. These signals, along with image and vibration data, complete multi-dimensional defect identification. The on-board edge computing module performs real-time analysis and reasoning on the data transmitted by the fixed sensing module, filtering out invalid data and uploading only the defect-related structured data and key image fragments to the cloud platform management module.

[0049] like Figure 1 , Figure 4 , Figure 5 , Figure 6 , Figure 7 , Figure 8 , Figure 9 and Figure 10 As shown, the inspection robot module is deployed in the subway depot and is only used during train return and depot entry and non-operational periods. This module consists of an autonomous navigation unit, a proximity sensing unit, a robot body 1, and a local control unit. The robot body 1 has wheels 11 at its bottom corners, and a control box 12 is mounted on it. Both the autonomous navigation unit and the local control unit are integrated inside the control box 12. The control box 12 has a connection slot 13 on its side and multiple control panels 14 on its upper surface. A robotic arm 2 is also mounted on the robot body 1. The robotic arm 2 has a base 21 at its bottom with multiple fixing holes 22. The robotic arm 2 is fixed to the robot body 1 by bolts passing through the fixing holes 22. A connection cable 23 is connected to the bottom side of the robotic arm 2. The connection plug 24 at the end of the connection cable 23 is inserted into the connection slot 13 of the control box 12. The output end of the robotic arm 2 has a mounting base 25 with multiple mounting slots 26 for mounting the proximity sensing unit. The top center of the mounting frame 3 at the output end of the robotic arm 2 is provided with a mounting post 31, and the top of the mounting post 31 is provided with a mounting plate 32. Multiple mounting holes 33 are opened on the mounting plate 32. The bottom surface of the mounting frame 3 is provided with a T-shaped clamping rod 34, and a junction box is provided on one side. Multiple male connectors 35 and wire terminals 36 are provided on the side of the junction box. The near-field sensing unit consists of a high-definition industrial camera 5, a laser rangefinder 6, and a supplementary light 4. Each of the three is provided with a clamping connector 51 at the top. A bayonet 52 is opened in the middle of the clamping connector 51. The bayonet 52 is engaged with the clamping rod 34 of the mounting frame 3. A clamping knob 53 is provided on the side of the clamping connector 51 aligned with the bayonet 52 for fixing. Wires 54 are connected to the sides of the three. The female connectors 55 at the ends of the wires 54 are connected to the male connectors 35 on the junction box. The autonomous navigation unit of this module adopts LiDAR and visual SLAM fusion navigation technology. It has a pre-loaded map of the interior of the vehicle and can autonomously complete obstacle avoidance, path planning and precise positioning. It can achieve full coverage inspection of the vehicle without blind spots along the preset route. The close-range perception unit can supplement the light with supplementary light 4 when the light is insufficient. The high-definition industrial camera 5 takes close-up pictures of the interior parts details and captures defects in the blind spots of the fixed perception module. The laser range sensor 6 measures the dimensional parameters such as the assembly gap and installation deviation of the interior parts. At the same time, the inspection robot module is equipped with an embedded AI computing box, which can complete the preliminary defect identification during the inspection process and only transmit the defect results and key images back to the vehicle edge computing module.

[0050] like Figure 1 and Figure 11As shown, the cloud platform management module is deployed in the subway operation control center and operates using a "cloud-edge-device" collaborative architecture. This architecture refers to the collaborative work and division of labor among the cloud (cloud platform management module), edge (onboard edge computing unit), and terminal (fixed sensing module, inspection robot module) in this inspection system, forming a complete inspection data processing and control system. The cloud platform management module has a built-in defect database that can record information such as the type, location, discovery time, and processing progress of interior component defects. It can also perform defect trend analysis based on historical data to predict defect development patterns and provide targeted maintenance suggestions for maintenance personnel. This module can receive structured data and defect information uploaded by the onboard edge computing modules in each carriage, and complete data aggregation, storage, analysis and visualization. It also supports inspection task scheduling, and issues control commands such as inspection parameter adjustment and robot inspection tasks to the onboard edge computing modules according to the train operation plan and interior maintenance needs. It can also summarize defect samples uploaded by the onboard edge computing modules, continuously train and optimize the AI ​​defect detection model, and regularly send the updated model to each onboard edge computing module to improve defect detection accuracy.

[0051] like Figure 1 and Figure 12 As shown, the alarm module includes an onboard local alarm unit and a cloud platform remote alarm unit. Data interaction is achieved through the onboard local area network, wireless communication network, onboard edge computing module, and cloud platform management module. The onboard local alarm unit is installed inside the carriage, while the cloud platform remote alarm unit establishes a communication connection with the maintenance personnel's terminal. When the onboard edge computing module identifies a serious train defect, it immediately triggers the onboard local alarm unit to issue an audible and visual alarm. Simultaneously, the defect information is urgently uploaded to the cloud platform management module. The cloud platform management module then triggers a remote alarm, pushing the defect location, type, and severity to the maintenance personnel's terminal via SMS, platform messages, etc., notifying the crew and maintenance personnel to handle the issue promptly. The entire alarm handling process is recorded, forming a closed-loop management system for defect handling.

[0052] Example 2

[0053] This embodiment provides an inspection method for a multi-dimensional intelligent inspection system for subway train interiors that integrates edge computing, as described in Embodiment 1. The method is divided into an inspection mode during operation and an inspection mode during non-operational periods. The two modes work together to achieve full-time, full-coverage inspection of train interior components.

[0054] The on-duty inspection mode is applied to the passenger-carrying operation of the train. First, the fixed sensing module is activated, continuously collecting images, vibration, and sound data of the interior components of the carriage through high-definition infrared cameras, vibration sensors, and sound sensors. The data is transmitted in real time to the on-board edge computing module. The on-board edge computing module uses a built-in AI defect detection model to analyze and infer the collected data in real time, accurately identifying appearance defects, structural defects, and dimensional defects of the interior components, and simultaneously filtering invalid data. If a minor defect is detected, the on-board edge computing module caches the defect type, location, and corresponding image information and then uploads it to the cloud platform management module, which records and archives the information for unified processing during off-duty periods. If a serious defect is detected, the on-board edge computing module immediately triggers the on-board local alarm unit to issue an audible and visual alarm, and simultaneously uploads the defect information to the cloud platform management module. The cloud platform management module activates a remote alarm, promptly notifying the crew and maintenance personnel to carry out emergency response.

[0055] The off-peak inspection mode is applied during train return and depot entry. First, the cloud platform management module, based on the train entry plan, sends inspection tasks to the onboard edge computing module. Upon receiving the task, the onboard edge computing module sends a start command to the inspection robot module. The inspection robot module then starts, and its autonomous navigation unit navigates the carriages autonomously, avoiding obstacles, based on a pre-set carriage map and planned path, completing a comprehensive, blind-spot-free inspection of the entire carriage area. During the inspection, the inspection robot module's close-range sensing unit continuously collects high-definition close-range images and dimensional deviation data of the interior components, transmitting this data in real-time to the onboard edge computing module. This focuses on identifying blind spots that cannot be covered by the fixed sensing module. The onboard edge computing module then processes the data transmitted by the inspection robot module. The data undergoes in-depth analysis, and defect records cached during runtime are used to confirm defects and supplement information, generating a complete defect report for the interior components of the vehicle. The report is then uploaded to the cloud platform management module, which classifies and grades defects according to four quality levels: A, B, C, and D. Level A represents excellent quality (no defects), Level B represents minor defects (not affecting safety, requiring regular maintenance), Level C represents general defects (affecting user experience, requiring timely repair), and Level D represents serious defects (affecting safety, requiring immediate repair). The defect information is then pushed to the maintenance personnel's terminals, where they conduct targeted repairs based on the defect reports. After the repairs are completed, the defect processing progress is updated in the cloud platform management module, completing the closed-loop management of the entire defect inspection and handling process.

[0056] In summary, compared with existing technologies, this application combines a fixed sensing module and an inspection robot module to conduct inspections during both train operation and non-operation periods. During operation, the fixed sensing module enables real-time monitoring around the clock, while the inspection robot provides full coverage and close-range inspection during non-operation periods. This effectively compensates for the blind spots of fixed cameras and solves the problem that traditional manual inspection and single monitoring modes cannot cover all scenarios, achieving comprehensive capture of defects in the interior components of the carriage. The use of an onboard edge computing module for local data processing eliminates the need to transmit all raw data to the cloud, significantly reducing network bandwidth pressure and shortening defect identification response latency to milliseconds, enabling immediate identification and alarm of defects during operation. Simultaneously, the onboard edge computing module filters and performs preliminary analysis on the data, uploading only structured data and key information, improving data transmission and processing efficiency. An improved lightweight AI defect detection model, combined with multi-dimensional perception data including images, vibrations, and sounds, enables accurate identification of appearance, structural, and dimensional defects in interior components. The inspection robot module's close-range perception and laser ranging further enhance the detection accuracy of defects in concealed areas and dimensions. Combined with collaborative verification during both running and non-running periods, the false negative and false positive rates are significantly reduced, far exceeding the level of manual inspection. A cloud platform management module enables centralized management, trend analysis, and task scheduling of defect data. Combined with continuous optimization of the AI ​​model, predictive maintenance of defects is achieved, reducing ineffective inspections. Simultaneously, it replaces traditional manual inspections, significantly reducing labor intensity and costs, especially mitigating the safety risks of late-night manual inspections, and improving the intelligence level of subway train interior maintenance. Real-time monitoring and timely alarms allow for the rapid detection and handling of serious defects affecting passenger safety, preventing defects from escalating and causing accidents. At the same time, minor defects are addressed promptly, maintaining the integrity of interior components, improving the passenger experience, and meeting the safety and service requirements of subway operations.

[0057] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A multi-dimensional intelligent inspection system for subway train interiors integrating edge computing, characterized in that, It includes an in-vehicle edge computing module, a fixed sensing module, an inspection robot module, a cloud platform management module, and an alarm module; The vehicle-mounted edge computing module establishes communication connections with the fixed sensing module and the inspection robot module, respectively. The fixed sensing modules are deployed in key locations inside the subway train carriages, including both ends of the carriage, the middle of the carriage, above the seats, and both sides of the doors. The inspection robot module is deployed in the subway depot and is used during train return and non-operational periods. The cloud platform management module is deployed in the subway operation control center and interacts with the on-board edge computing module through a wireless communication network. The alarm module interacts with the vehicle edge computing module and cloud platform management module through the vehicle local area network and wireless communication network.

2. The multi-dimensional intelligent inspection system for subway train interiors integrating edge computing as described in claim 1, characterized in that, The onboard edge computing module is deployed in each subway train carriage, employing a lightweight edge computing server and incorporating a pruned and quantized optimized AI defect detection model. This AI defect detection model supports real-time local data processing, model inference, and result caching, eliminating the need to transmit all raw data to the cloud. The onboard edge computing module also features built-in network interruption recovery and caching capabilities, allowing it to locally cache alarm events and critical data during temporary network outages and automatically synchronize them to the cloud platform once the network is restored. Furthermore, the onboard edge computing module incorporates a local rule engine, enabling independent analysis and alarm generation based on preset rules even when disconnected from the cloud.

3. The multi-dimensional intelligent inspection system for subway train interiors integrating edge computing as described in claim 1, characterized in that, The fixed sensing module includes a high-definition infrared camera, a vibration sensor, and a sound sensor, used for all-weather real-time monitoring during train operation. The high-definition infrared camera adopts a modular design, supports adaptive zoom and region of interest analysis, and only performs full-resolution acquisition of key areas of interior components, while other areas are analyzed at low resolution. The high-definition infrared camera is used to collect image data of interior components in the carriage, capturing surface defects such as scratches, damage, color differences, and stains, as well as structural defects such as loose armrests and detached interior panels. The vibration sensor is deployed at the connection points of armrests and seats to detect the installation firmness of interior components and capture vibration signals corresponding to looseness and abnormal noises. The sound sensor is used to collect abnormal sound signals generated by loose or damaged interior components to assist in defect identification.

4. The multi-dimensional intelligent inspection system for subway train interiors integrating edge computing as described in claim 1, characterized in that, The inspection robot module includes an autonomous navigation unit, a near-field perception unit, a robot body (1), and a local control unit. The autonomous navigation unit adopts a fusion navigation technology of lidar and visual SLAM, pre-loads the map inside the carriage, supports autonomous obstacle avoidance, path planning and precise positioning, and can move autonomously in the carriage along the preset inspection route to achieve full coverage inspection without blind spots in the carriage. The close-range sensing unit consists of a high-definition industrial camera (5), a laser rangefinder (6), and a supplementary light (4). The high-definition industrial camera (5) can capture details of interior parts at close range and capture defects in the blind spots of the fixed sensing module. The laser rangefinder (6) is used to measure the assembly gaps and installation deviations of the interior parts. The supplementary light (4) is used to provide supplementary lighting for the high-definition industrial camera (5) when there is insufficient light. The inspection robot module is equipped with an embedded AI computing box, which can perform preliminary defect identification in real time during the inspection process and only transmit the defect results and key images back.

5. The multi-dimensional intelligent inspection system for subway train interiors integrating edge computing as described in claim 4, characterized in that, The robot body (1) is equipped with a control box (12), and the autonomous navigation unit and the local control unit are both integrated inside the control box; the robot body (1) is equipped with a robotic arm (2), and the output end of the robotic arm (2) is equipped with a mounting frame (3). The high-definition industrial camera (5), laser rangefinder (6) and fill light (4) of the near-field sensing unit are all mounted on the mounting frame (3).

6. The multi-dimensional intelligent inspection system for subway train interiors integrating edge computing as described in claim 5, characterized in that, The robot body (1) has wheels (11) at the bottom corners, the control box (12) has a connection slot (13) on the side, and the control box (12) has multiple control panels (14) on its upper surface; the robotic arm (2) has a base (21) at its bottom, and the base (21) has multiple fixing holes (22) on its bottom, and the robotic arm (2) is fixed to the robot body (1) by bolts passing through the fixing holes (22); the robotic arm (2) has a connection line (23) on its bottom side, and the end of the connection line (23) has a connection plug (24) connected to the connection slot (13); the output end of the robotic arm (2) has a mounting base (25), and the mounting base (25) has multiple mounting slots (26).

7. A multi-dimensional intelligent inspection system for subway train interiors integrating edge computing as described in claim 6, characterized in that, The mounting bracket (3) has a mounting post (31) at the top center, a mounting plate (32) at the top of the mounting post (31), and multiple mounting holes (33) on the mounting plate (32); the mounting bracket (3) has a clamping rod (34) on the bottom surface, the clamping rod (34) is T-shaped, a junction box is provided on one side of the mounting bracket (3), and multiple male connectors (35) and connectors (36) are provided on the side of the junction box; the supplementary light (4), the high-definition industrial camera (5) and the laser rangefinder (6) are located on the top. Each of the components is equipped with a snap-fit ​​connector (51), and the snap-fit ​​connector (51) has a snap-fit ​​opening (52) in the middle. The snap-fit ​​opening (52) is engaged with the snap-fit ​​rod (34), and the snap-fit ​​connector (51) has a clamping knob (53) on its side aligned with the snap-fit ​​opening (52). The fill light (4), the high-definition industrial camera (5), and the laser rangefinder (6) are all equipped with wires (54) on their sides. The ends of the wires (54) are equipped with female connectors (55), and the female connectors (55) are respectively connected to the male connectors (35) on the junction box.

8. The multi-dimensional intelligent inspection system for subway train interiors integrating edge computing as described in claim 1, characterized in that, The cloud platform management module has a built-in defect database that records information such as the type, location, discovery time, and processing progress of interior component defects. It combines historical data to perform trend analysis, predict defect development patterns, and provide targeted maintenance suggestions for maintenance personnel. The cloud platform management module supports inspection task scheduling and can issue inspection parameter adjustment and robot inspection task control commands to the on-board edge computing module according to the train operation plan and interior component maintenance requirements. The cloud platform management module supports continuous model training and optimization, summarizes the defect samples uploaded by the on-board edge computing module, regularly updates the AI ​​defect detection model, and distributes the optimized model to each on-board edge computing module.

9. A multi-dimensional intelligent inspection system for subway train interiors integrating edge computing as described in claim 1, characterized in that, The alarm module includes an in-vehicle local alarm unit and a cloud platform remote alarm unit. The in-vehicle local alarm unit is deployed inside the carriage, and the cloud platform remote alarm unit establishes communication with the maintenance personnel's terminal. When the in-vehicle edge computing module identifies a serious defect, it immediately triggers the audible and visual alarm of the in-vehicle local alarm unit and simultaneously uploads the defect information to the cloud platform. The cloud platform triggers a remote alarm to notify the crew and maintenance personnel to handle the issue promptly. The cloud platform remote alarm unit pushes the defect information to the maintenance personnel's terminal via SMS and platform messages, specifying the defect location, type, and severity, and records the alarm handling status, forming a closed-loop management system.

10. A multi-dimensional intelligent inspection system for subway train interiors integrating edge computing as described in any one of claims 1-9, characterized in that, The inspection system includes two modes: a running inspection mode and an off-run inspection mode, enabling coordinated operation. During the running inspection, when the train is in passenger service, the fixed sensing module is activated and continuously collects image, vibration, and sound data of the interior components of the carriage. This data is transmitted to the on-board edge computing module for real-time analysis and reasoning, and appropriate cache upload or alarm operations are performed based on the severity of the defects. During the off-run inspection, when the train is returned to the depot, the cloud platform management module issues inspection tasks, and the on-board edge computing module controls the inspection robot module to start, completing a full-coverage inspection and collecting data. After analysis by the on-board edge computing module, defects are identified.

Images