A multifunctional modular track inspection system

Abstract

This invention discloses a multifunctional modular integrated track inspection system, belonging to the field of track inspection technology. Using a lightweight mobile trolley as a carrier, it integrates a system body, a positioning module, six multifunctional inspection modules, and a data processing module. The system body provides a running platform, power supply, and operation control, and uses differential pulse signals to control and synchronize different functional inspection modules. The positioning module is used to align the position of data within the system modules. The multifunctional inspection modules cover visual inspection, track geometric parameter detection, rail profile detection, rail creep displacement detection, turnout detection, and high-precision fastener detection. The data processing module processes multi-source inspection data and outputs inspection results and maintenance suggestions. This invention solves the problems of traditional equipment having single functions, low inspection efficiency and accuracy, high reliance on manpower, and poor data coordination, thus improving the scientific nature and efficiency of track maintenance.

Description

Technical Field

[0001] This invention relates to the field of track inspection technology, and in particular to a multifunctional modular track integrated inspection system. Background Technology

[0002] As a core mode of land transportation, railway transportation undertakes heavy passenger and freight transport tasks. With the speed increase of existing lines to 200km / h and the large-scale operation of high-speed railways, the railway traffic density has increased and the train interval has shortened, which puts forward higher requirements for the efficiency and accuracy of track maintenance and repair in the engineering system.

[0003] Existing track inspection equipment has the following shortcomings: 1) Single function: Most equipment can only perform single parameter detection (such as only measuring track gauge or only inspecting fasteners), so multiple equipment are required to operate multiple times, making the inspection process cumbersome; 2) Low efficiency: The maximum detection speed of traditional track inspection instruments does not exceed 5km / h, and the daily inspection mileage is limited, which is difficult to meet the needs of large-scale line inspection; 3) High dependence on manpower: The use and maintenance of equipment require a lot of human participation, and manual inspection is easily affected by subjective factors, resulting in low accuracy of defect identification; 4) Poor data collaboration: Data from each inspection device is stored independently, making it impossible to achieve position alignment and fusion analysis, and making it difficult to generate systematic maintenance recommendations.

[0004] To address the aforementioned issues, there is an urgent need for a modular intelligent testing device that integrates multiple testing functions, high precision, and high efficiency, enabling multi-parameter testing and data linkage analysis in a single operation, thereby reducing labor costs and improving the scientific nature of maintenance. Summary of the Invention

[0005] In view of this, this application provides a multifunctional modular track integrated inspection system to address the shortcomings of the existing technology.

[0006] The first aspect of this application provides a multifunctional modular track inspection system, comprising:

[0007] The system body is used to provide a stable running platform, power supply and operation control, and to control and synchronize all modules in the multi-functional detection module by collecting differential pulse signals output by the high-precision encoder in the positioning module and sending them to the vehicle controller.

[0008] The positioning module is used to align the positions of data across multiple modules within the system.

[0009] The multi-functional inspection module includes a visual inspection module, a track geometry parameter inspection module, a rail profile inspection module, a rail creep displacement inspection module, a turnout inspection module, and a high-precision fastener inspection module. All inspection modules are mechanically connected through reserved mechanical interfaces on the system body and electrically connected through reserved electrical interfaces.

[0010] The data processing module is used to process the multi-source detection data from the multi-functional detection module and output the detection results and maintenance suggestions.

[0011] In one possible implementation of the first aspect, the visualization inspection module includes a 3D camera, a 2D camera, an LED light source, and a detection beam;

[0012] The 3D camera emits a line laser and receives the reflected signal to acquire a corresponding depth image; the 2D camera scans line by line to acquire a corresponding texture image; the LED light source provides an auxiliary light source, and the detection beam provides support;

[0013] The depth image and texture image are processed by a preset neural network to accurately identify the track type;

[0014] Based on structural differences data of different track types and prior data on defect distribution, the system can intelligently detect the condition and defects of track components.

[0015] In one possible implementation of the first aspect, the track geometry parameter detection module includes an inertial navigation measurement unit, an odometer, a visual measurement unit, a synchronization controller, and a processing computer.

[0016] The inertial navigation measurement unit uses laser inertial navigation to collect motion attitude data of the traveling platform;

[0017] The odometer is used to collect relative mileage data of the traveling platform within the area of ​​the rail to be measured;

[0018] The visual measurement unit is used to acquire images of the rail and calculate the attitude change information of the traveling platform relative to the rail based on the images of the rail.

[0019] The synchronization controller is used for the synchronous control of the system, and triggers the odometer, the vision measurement unit and the inertial navigation measurement unit to synchronously collect data by generating high-precision pulses;

[0020] The processing computer is used to solve the relative mileage data and the motion posture data through a combined navigation algorithm to generate the relative motion trajectory of the walking platform. The combined navigation algorithm is based on the Kalman filter algorithm.

[0021] The processing computer is also used to perform serpentine correction on the relative motion trajectory; the serpentine correction of the relative motion trajectory is achieved by using the attitude change information of the traveling platform relative to the rail to compensate for the spatial displacement caused by the serpentine motion of the traveling platform;

[0022] The processing computer obtains the motion trajectory of the left and right rail vertices based on the corrected relative motion trajectory, and extracts the track geometric parameters by calculating the motion trajectory of the left and right rail vertices.

[0023] In one possible implementation of the first aspect, the rail profile detection module is used for:

[0024] High-precision point cloud data of rail cross-sections are obtained using 3D cameras.

[0025] Based on 3D point cloud technology, the measured rail profile features in the high-precision point cloud data are extracted and accurately registered with the preset standard rail profile.

[0026] The measured rail profile features after precise registration are compared point by point with the standard rail profile features. The deviation values ​​in the vertical direction, the deviation values ​​in the horizontal direction, and the total wear are calculated, and a continuous rail profile curve is generated.

[0027] In one possible implementation of the first aspect, the rail crawling displacement detection module includes a crawling identification module, a synchronous trigger control module, an image acquisition module, and a data processing module.

[0028] The crawling marking module is used to set multiple crawling marking modules at equal intervals in the track area to be tested. Each crawling marking module includes the following markings: two sets of symmetrically arranged rail displacement markings fixed to the web of the rail, namely markings A1 and A2; two sets of symmetrically arranged sleeper reference markings fixed to the top surface of the rail sleeper, namely markings B1 and B2; and a positioning marking fixed to a fixed building beside the track, denoted as marking C.

[0029] The synchronous trigger control module includes an embedded microcontroller. The embedded microcontroller acquires the pulse count of the wheel encoder and generates synchronous trigger pulses at preset trigger intervals to control the image acquisition module to perform synchronous image acquisition, thereby ensuring that the image data acquired by the image acquisition module has time synchronization.

[0030] The image acquisition module consists of multiple area scan cameras for synchronous image acquisition. All area scan cameras transmit the acquired image data to the data processing module in real time. Specifically: the first area scan camera is aligned with the A1 and B1 tables to acquire high-definition grayscale images; the second area scan camera is aligned with the A2 and B2 tables to acquire high-definition grayscale images; and the third and / or fourth area scan camera is aligned with the C table to acquire high-definition grayscale images.

[0031] The data processing module pre-stores calibration and displacement calculation algorithms. It calculates the crawling measurement data of the rail measurement points on the rail side where the A1 rail is located using the displacement difference between rail C and rail A1; it calculates the crawling measurement data of the rail measurement points on the rail side where the A2 rail is located using the displacement difference between rail C and rail A2; it calculates the sleeper crawling measurement data on the rail side where the B1 rail is located using the displacement difference between rail C and rail B1; and it calculates the sleeper crawling measurement data on the rail side where the B2 rail is located using the displacement difference between rail C and rail B2.

[0032] In one possible implementation of the first aspect, the branch area detection module includes hardware devices and a configured branch area detection algorithm. The hardware devices reuse the hardware devices of the visual inspection module, and the branch area detection algorithm includes:

[0033] Based on CAD parametric modeling and correction using on-site measured data, four types of 3D model libraries for branch areas were established. Each type of 3D model library integrates the dimensional parameters, spatial coordinate information, and detection thresholds of various components in the branch area.

[0034] Two-level positioning is used for model matching. The turnout area is coarsely located by the mileage data collected by the mileage encoder. Then, multiple strong feature points are extracted by collecting the point cloud data of the coarsely located turnout area. The RANSAC algorithm is used to match all strong feature points with the standard model in the 3D model library to determine the theoretical position of each component in the turnout area.

[0035] In the standard model of the 3D model library, the detection area boundary is pre-set for each component in the turnout area, and the pre-set detection area boundary is projected onto the collected point cloud data to form the coarse annotation area of ​​each component in the turnout area.

[0036] Based on the collected point cloud data, the shape features and size parameters of each component in the turnout area are analyzed, and the target components are screened and segmented based on the corresponding detection thresholds. At the same time, fitting processing is performed to accurately determine the measurement point position of the target component.

[0037] Within the coarsely marked area of ​​each component, based on the measured points of the determined target components, qualitative and quantitative detection is carried out on 7 types of defects. The 7 types of defects include bolt missing defects, elastic clip displacement defects, switch rail contact gap defects, limiter centering defects, guard rail wheel groove exceeding standard defects, guard rail wear defects, and sleeper spacing exceeding standard defects.

[0038] In one possible implementation of the first aspect, the high-precision fastener detection module includes a 3D camera and a data unit, wherein the data unit integrates a fastener spring gap detection algorithm, a pad thickness measurement algorithm, and an insulating block and track gauge baffle model identification algorithm.

[0039] The fastener latch gap detection algorithm includes:

[0040] Acquire 3D point cloud data of the fasteners, including depth information, using a 3D camera;

[0041] By combining a preset neural network, the various components in the fastener are identified, segmented and located from the three-dimensional point cloud data of the fastener, and the spatial position of each component in the fastener is determined.

[0042] The radius of the spring bar is calculated by fitting the segmented spring bar point cloud data using a 3D point cloud algorithm, and the position of the spring tongue is located and the height of the spring tongue pressing position is measured.

[0043] The height of the insulating gauge block at the corresponding spring latching position is calculated by fitting the point cloud data of the identified and segmented insulating gauge block using a 3D point cloud algorithm.

[0044] Based on the elastic bar radius, the corresponding elastic bar diameter is obtained, and based on the elastic bar diameter, the height of the spring tongue clamping position, and the height of the insulating gauge block, the spring tongue gap value is measured.

[0045] The algorithm for measuring the thickness of the pad includes:

[0046] The fastener area is scanned using a 3D camera to collect three-dimensional point cloud data including the rail, rail support platform, iron pad, and elastic clip;

[0047] Based on deep learning algorithms, semantic segmentation is performed on the collected 3D point cloud data of rails, rail supports, iron pads and elastic rails to accurately distinguish and locate the spatial range of rails, rail supports and iron pads.

[0048] After semantic segmentation, obtain the three-dimensional point cloud data of rail, rail support platform and iron pad, and perform plane fitting using 3D point cloud algorithm to obtain the corresponding rail plane, rail support platform plane and iron pad plane.

[0049] The thickness of the upper pad is obtained by calculating the vertical distance between the rail plane and the iron pad plane; the thickness of the lower pad is obtained by calculating the vertical distance between the rail support platform plane and the iron pad plane; and the total thickness of the pad is obtained by calculating the vertical distance between the rail plane and the rail support platform plane.

[0050] The algorithm for identifying the model of the insulating block and track gauge baffle includes:

[0051] The fastener area is scanned using a 3D camera to collect three-dimensional point cloud data including rails, insulating blocks, gauge baffles, and elastic strips.

[0052] Based on deep learning algorithms, semantic segmentation is performed on the collected 3D point cloud data of rails, insulating blocks, gauge baffles and elastic bars to accurately distinguish and locate the spatial range of rails, insulating blocks and gauge baffles.

[0053] Based on the semantically segmented 3D point cloud data of the rail, insulation block and gauge baffle, the edge regions of the rail, insulation block and gauge baffle are extracted.

[0054] The edge regions of the rail, insulating block and gauge baffle are fitted with straight lines using 3D point cloud algorithm to obtain straight lines at the rail edge, straight lines on both sides of the insulating block and straight lines of the gauge baffle.

[0055] Using the straight line of the rail edge as a reference, the straight lines on both sides of the insulating block and the straight line of the gauge baffle are calibrated; the distance between the straight lines on both sides of the insulating block after calibration is calculated and compared with the size of the standard insulating block to achieve the identification of the insulating block model; the distance between the straight line of the gauge baffle and the straight line of the insulating block near the iron pad after calibration is calculated and compared with the distance between the standard gauge baffle and the insulating block to achieve the identification of the gauge baffle model.

[0056] In one possible implementation of the first aspect, the system body includes a walking platform:

[0057] The traveling platform adopts a detachable structure, consisting of an active unit, a driven unit, an operating platform, a middle seat, and a rear seat; the active unit and the driven unit are interlocked by a mechanical structure; the operating platform is positioned by a pin and then locked using a spinning quick-locking structure; the middle seat, the rear seat, the active unit, and the driven unit are positioned by pins and then locked using a spinning quick-locking structure.

[0058] The system body also includes a power module for power supply, which is composed of multiple lithium batteries;

[0059] The system body also includes an electrical module and an integrated control module for operation control;

[0060] The electrical module includes a system control unit and an audible and visual warning unit. The system control unit is used to control the travel direction and speed of the traveling platform by receiving user commands, and to communicate with each module to display relevant information on the display screen. The audible and visual warning unit consists of a spotlight, a yellow warning light, and a high-frequency speaker.

[0061] The integrated control module provides an interactive UI interface, allowing users to interact with each module.

[0062] In one possible implementation of the first aspect, the positioning module includes a high-precision encoder, a BeiDou GPS positioning module, and a sleeper number image recognition camera, used to achieve data position alignment of multiple modules within the system.

[0063] In one possible implementation of the first aspect, the data processing module is equipped with a multi-sensor data fusion algorithm and an intelligent diagnostic model. The multi-sensor data fusion algorithm fuses the multi-source detection data, and the intelligent diagnostic model outputs detection results and maintenance suggestions.

[0064] Its beneficial effects are as follows: This invention discloses a multifunctional modular track inspection system, which uses a lightweight mobile trolley as a carrier and integrates a system body, a positioning module, six multifunctional inspection modules, and a data processing module. The system body provides a running platform, power supply, and operation control, and controls different functional inspection modules through differential pulse signals. The positioning module is used to align the position of the module data within the system. The multifunctional inspection modules cover visual inspection, track geometric parameter detection, rail profile detection, rail creep displacement detection, turnout detection, and high-precision fastener detection. The data processing module processes multi-source inspection data and outputs inspection results and maintenance suggestions. This invention solves the problems of traditional equipment having single functions, low inspection efficiency and accuracy, high reliance on manpower, and poor data coordination, thus improving the scientific nature and efficiency of track maintenance. Attached Figure Description

[0065] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0066] Figure 1 This is a schematic diagram of the composition of a multifunctional modular track integrated inspection system provided in an embodiment of this application;

[0067] Figure 2 This is a schematic diagram of the overall structure of a multifunctional modular track inspection system provided in an embodiment of this application;

[0068] Figure 3 This is a schematic diagram of the structure of the traveling platform in a multifunctional modular track integrated inspection system provided in this application embodiment;

[0069] Figure 4 This is a schematic diagram of a multi-functional detection module controlled by differential pulse signals according to an embodiment of this application;

[0070] Figure 5 This is a schematic diagram of the rail creep displacement detection module provided in the embodiments of this application;

[0071] Among them, 1-travel platform, 2-visual inspection module, 3-track geometric parameter detection module, 4-rail profile detection module, 5-data processing module, 6-active part, 7-driven part, 8-operating table, 9-middle seat, 10-rear seat. Detailed Implementation

[0072] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0073] In this application, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0074] Example 1

[0075] Existing track inspection equipment has the following shortcomings: 1) Single function: Most equipment can only perform single parameter detection (such as only measuring track gauge or only inspecting fasteners), so multiple equipment are required to operate multiple times, making the inspection process cumbersome; 2) Low efficiency: The maximum detection speed of traditional track inspection instruments does not exceed 5km / h, and the daily inspection mileage is limited, which is difficult to meet the needs of large-scale line inspection; 3) High dependence on manpower: The use and maintenance of equipment require a lot of human participation, and manual inspection is easily affected by subjective factors, resulting in low accuracy of defect identification; 4) Poor data collaboration: Data from each inspection device is stored independently, making it impossible to achieve position alignment and fusion analysis, and making it difficult to generate systematic maintenance recommendations.

[0076] Therefore, this application provides a multifunctional modular track integrated inspection system, such as Figure 1 As shown, it includes:

[0077] The system body is used to provide a stable running platform, power supply and operation control, and to control and synchronize all modules in the multi-functional detection module by collecting differential pulse signals output by the high-precision encoder in the positioning module and sending them to the vehicle controller.

[0078] The positioning module is used to align the positions of data across multiple modules within the system.

[0079] The multi-functional inspection module includes a visual inspection module, a track geometry parameter inspection module, a rail profile inspection module, a rail creep displacement inspection module, a turnout inspection module, and a high-precision fastener inspection module. All inspection modules are mechanically connected through reserved mechanical interfaces on the system body and electrically connected through reserved electrical interfaces.

[0080] The data processing module is used to process the multi-source detection data from the multi-functional detection module and output the detection results and maintenance suggestions.

[0081] This embodiment provides a multifunctional modular track inspection system, which includes a system body, a positioning module, a multifunctional inspection module (six inspection modules), and a data processing module, specifically:

[0082] The system itself includes a traveling platform, a power module, an electrical module, and a comprehensive control module.

[0083] Walking platform 1 (please refer to) Figure 2 and Figure 3 ):

[0084] The main body adopts a detachable structure, consisting of an active part 6, a driven part 7, an operating table 8, a middle seat 9, and a rear seat 10. The active part 6 and the driven part 7 are interlocked by a mechanical structure. The operating table 8 and the active part 6 are locked together by a spinning quick-locking structure after being positioned by a pin. The seat is locked together with the active part 6 and the driven part 7 by a spinning quick-locking structure after being positioned by a pin.

[0085] The active unit 6 mainly consists of a drive frame, drive wheel assembly, motor assembly, and brake assembly. It is the core part of the traveling platform for power output and driving control, providing power for the equipment's movement and ensuring driving safety.

[0086] The driven unit mainly consists of a driven frame, driven wheel sets, and a traction assembly. The traction assembly, located at the front of the driven frame (i.e., the front in the direction of travel), serves as the mechanical interface for the visual inspection module, providing the necessary traction force for its movement during system operation. Simultaneously, a mechanical interface for installing the rail profile detection module 4 is designed at the rear of the driven frame (i.e., the rear in the direction of travel). It should be noted that this mechanical interface also serves as the mechanical interface for the rail creep displacement detection module, the turnout detection module, and the high-precision fastener detection module. These modules share the same mechanical interface and are assembled using pin positioning, then connected via a spin-lock quick-release method, facilitating the installation and replacement of each module.

[0087] The control panel 8 is mainly composed of a main frame, sheet metal covering and electrical control components. The control panel 8 is also designed with a mechanical interface for a visual inspection module. The visual inspection module is positioned with a pin and then quickly fixed by spinning after being combined with the control panel.

[0088] The middle seat 9 and the rear seat 10 are mainly composed of a main frame welded to a sheet metal seat surface, providing seating space for operators.

[0089] Power module:

[0090] Composed of multiple lithium batteries, the power source for the running platform is provided by a single battery. All batteries have equalization function, overcharge and over-discharge protection function, short circuit protection function, temperature protection function, and shock and vibration protection function, ensuring a continuous, stable and safe power source for the platform.

[0091] Electrical modules:

[0092] This includes a system control unit and an audible and visual warning unit. The system control unit is installed in the control panel 8 and the electrical box under the vehicle, which saves space and ensures stable operation between the various systems.

[0093] The system control unit, as the core control device of the track trolley, receives user commands via push rods, controls the trolley's motor, and thus controls the trolley's direction and speed. It also communicates with the battery, encoder, and detection modules, displaying relevant information on the screen for operators to monitor the trolley's and detection status in real time. It features high-precision pulse conversion; the vehicle controller converts the differential pulse signal output from the high-precision encoder into the high-precision pulses required by the detection modules and other equipment, providing trigger signals and distributing them as multiple outputs. This high-precision pulse signal is independent of the trolley's speed and does not change with acceleration or deceleration. Figure 4 As shown. Furthermore, the system control unit also has the following functions:

[0094] a. Two-way speed control function, which communicates with the motor driver to control the rotation direction and speed of the motor, thereby controlling the track trolley to achieve two-way travel, with a maximum travel speed of 20km / h;

[0095] b. Directional driving control function: The car can be controlled to drive at a constant speed using the speed control button. The speed control levels are 5km / h, 12km / h and 20km / h.

[0096] c. Speed ​​holding control function, which provides precise and stable control of motor speed, with no obvious jerking during acceleration or deceleration, and less likelihood of stalling when going downhill;

[0097] d. Mileage positioning function: calculates the accurate mileage of the car based on the encoder signal, with a mileage positioning error of no more than 2m per kilometer;

[0098] e. High-precision pulse conversion function, which converts high-precision encoder signals into high-precision pulses required by detection modules and other equipment, provides trigger signals for them and distributes them into multiple outputs. The accuracy of this signal is independent of the speed of the trolley and does not change due to the acceleration or deceleration of the trolley.

[0099] f. Communication function of the detection module, enabling communication with the multi-functional detection module;

[0100] g. Display screen communication function: Communicates with the display screen to display the following information: driving parameters, battery parameters, positioning parameters, and fault codes; driving parameters include the trolley's speed, mileage, and motor speed; battery parameters include battery voltage and charge; positioning parameters display the trolley's coordinates; fault codes display motor and battery fault codes to facilitate troubleshooting by operators.

[0101] The audible and visual warning unit consists of a spotlight, a yellow warning light, and a high-frequency horn. The spotlight provides workers with a clear view in both directions, while the yellow warning light and high-frequency horn alert other personnel on the line, ensuring the safety of both the workers and other personnel.

[0102] Integrated control module:

[0103] An interactive UI interface is provided, through which users can interact with each module; all modules in the multi-functional detection module are uniformly managed by the integrated control module, which establishes communication with each module, sets tasks for each detection module, unifies the mileage of each module, and monitors the status of each module in real time.

[0104] Positioning module:

[0105] It includes a high-precision encoder (pulse resolution ≤0.1mm), a Beidou GPS positioning module, and a sleeper number image recognition camera, all integrated into the system control unit of the running platform. Through the fusion of multiple technologies, it achieves accurate mileage positioning with a mileage error of ≤2%, ensuring the alignment of data positions from multiple detection modules.

[0106] Data processing module:

[0107] Integrated into the industrial control computer of the running platform, it is equipped with multi-sensor data fusion algorithms and intelligent diagnostic models, supports 4G real-time data transmission, automatically generates defect reports, spring and autumn inspection workload statistics, and operation suggestions such as track gauge fine-tuning, track lifting, and track realignment. It also has historical data comparison functions to analyze defect trends and provide early warnings of track breakage risks.

[0108] Multifunctional detection module:

[0109] The system comprises six main detection modules: a visual inspection module, a track geometry parameter detection module, a rail profile detection module, a rail creep displacement detection module, a turnout detection module, and a high-precision fastener detection module. All modules are connected via pre-reserved mechanical and electrical interfaces. Specifically: the visual inspection module 2 is installed via the pre-reserved mechanical interface on the control panel, using pins for positioning and assembly, and then secured with a spin-lock after assembly. The track geometry parameter detection module is pre-placed under the platform frame and assembled by rotating the traction component handle to the traction position. The remaining detection modules share the mechanical mounting interface at the rear of the driven part, using pins for positioning and assembly, and then secured with a spin-lock after assembly. The visual inspection module connects to the trolley via eight circular connectors. Two connector specifications are available: a six-pin connector for power and control signal transmission, and a ten-pin connector for network communication. Each camera requires one six-pin and one ten-pin connector. The pin angles of the connectors differ between cameras to prevent incorrect installation. The track geometry parameter detection module only needs to be connected to the pre-installed network port connector on the trolley's control panel via a single network cable. The remaining detection modules share the heavy-duty connector at the rear of the driven frame and are connected to the track trolley. The power supply, control, and network communication connections within the heavy-duty connectors of each module are completely identical, allowing for rapid switching between modules without altering the electrical wiring.

[0110] Visual inspection module:

[0111] The equipment consists of a 3D camera, a 2D camera, an LED light source, and a corresponding detection beam.

[0112] Working principle: By integrating imaging technology from 3D and 2D cameras with neural network technology, it accurately identifies track types, including ballasted track, ballastless track, turnout sections, and non-turnout sections. Then, based on structural data and prior data on defect distribution (defect distribution patterns) for different track bed types, it efficiently detects and intelligently diagnoses the condition and defects of track components. It can accurately identify common defects such as missing fasteners, breakage, displacement, rail abrasion, and spalling, providing accurate defect information for track maintenance.

[0113] Track geometry parameter detection module:

[0114] The equipment consists of an inertial navigation measurement unit, an odometer, a visual measurement unit, a synchronization controller, and a processing computer.

[0115] The working principle involves the laser inertial navigation system in the inertial navigation measurement unit acquiring motion attitude data of the chassis body, while the odometer acquires relative mileage data of the chassis body traveling within the area of ​​the track to be measured. The acquired relative mileage data and motion attitude data are processed using a combined navigation algorithm (Kalman filter algorithm) to generate the relative motion trajectory of the chassis body. Then, attitude change information acquired by the visual measurement unit is used to perform serpentine correction on the relative motion trajectory, resulting in a corrected relative motion trajectory, thereby achieving accurate extraction of track relative parameters. Further, the positioning module acquires the absolute spatial position information of the chassis body, and the multi-source data acquired by each device is aligned in time and space using a synchronous controller. During data processing, the accuracy of the rail profile data is further improved by sequentially performing serpentine correction and matching processing on the original rail profile data, resulting in matched rail profile data, thus effectively eliminating data errors caused by the misalignment of the chassis wheels and the track during data acquisition. In terms of track geometry parameter calculation, a Kalman filter algorithm is used for integrated navigation to generate the motion trajectory of the vehicle body in the geodetic coordinate system. This motion trajectory is then projected onto the matched rail profile data to obtain a trajectory that represents the rail vertices. The track geometry parameters are then extracted by calculating from this trajectory. The specific calculation method is as follows:

[0116] For the elevation and orientation parameters of the left and right tracks, the sliding window measurement method, combined with the midpoint chord measurement method, is used to calculate the true trajectories of the left and right track vertices.

[0117] For the track level parameters, the elevation difference between the top of the left rail and the top of the right rail is obtained by calculating the actual trajectory of the left rail vertex and the actual trajectory of the right rail vertex, and then the track level parameters are determined. The triangular pit parameters are calculated based on the algebraic difference of the level difference within the sliding window size.

[0118] For the track gauge value, firstly, extract the left and right rail profile data of the rail to be measured from the matched rail profile data. Then, draw tangents to the vertices of the left and right rail profile data respectively to obtain the first and second tangents. Shift the first and second tangents at the same cross-section downwards by 16mm and calculate the coordinates of the intersection point of the left and right rail profile data. Finally, calculate the Euclidean distance based on the coordinates of the intersection point of the left and right rail profile data to obtain the track gauge value.

[0119] Rail profile detection module 4:

[0120] The equipment consists of a high-precision line structured light 3D camera.

[0121] The working principle involves acquiring high-precision point cloud data of the rail cross-section using a high-precision line structured light 3D camera imaging technology. Further, 3D point cloud technology and multi-segment profile features are used to register a standard profile. The registered profile is then compared and measured with the standard profile to achieve high-precision detection of rail head wear. This allows for accurate calculation of vertical wear, lateral wear, and total wear, generating continuous profile curves to provide precise data support for rail maintenance. The camera used is the SICK XR300, which boasts a Z-axis resolution of 0.011mm, an X-axis resolution of 0.094mm, a Z-axis measurement range of 180mm, and a maximum full-frame scanning frame rate of 7kHz.

[0122] Rail creep displacement detection module (such as) Figure 5 (as shown)

[0123] The equipment consists of a crawling identification module, a synchronous trigger control module, an image acquisition module, and a data processing module.

[0124] The crawling marking module consists of multiple crawling marking modules evenly spaced in the track area to be tested. Each crawling marking module includes the following types of markings: two sets of symmetrically arranged rail displacement markings fixed to the web of the rail, namely markings A1 and A2; two sets of symmetrically arranged sleeper reference markings fixed to the top surface of the rail sleeper, namely markings B1 and B2; and positioning markings fixed to fixed structures along the track, namely marking C.

[0125] The synchronous trigger control module includes an embedded microcontroller. This controller calculates the pulses obtained from the wheel encoder and generates synchronous trigger pulses at preset trigger intervals to control the image acquisition module to perform synchronous image acquisition, ensuring that the acquired image data has time synchronization.

[0126] The image acquisition module consists of multiple area scan cameras that acquire images synchronously. These area scan cameras transmit the acquired image information to the data processing module in real time. The specific division of labor is as follows: the first area scan camera 21 is aimed at the A1 and B1 targets to acquire high-definition grayscale images; the second area scan camera 31 is aimed at the A2 and B2 targets to acquire high-definition grayscale images; and the third area scan camera 11 is aimed at the C target to acquire high-definition grayscale images to obtain clear image information of the C target.

[0127] The data processing module, pre-stored with calibration and displacement calculation algorithms, operates as follows: By calculating the displacement difference between C and A1, it obtains the crawling measurement data of the rail measurement points on the rail side where A1 is located; by calculating the displacement difference between C and A2, it obtains the crawling measurement data of the rail measurement points on the rail side where A2 is located; by calculating the displacement difference between C and either B1 or B2, it obtains the sleeper crawling measurement data on the rail side where B1 or B2 is located. The core algorithm principle: After acquiring image information from the image acquisition module, the data processing module uses the PNP algorithm to solve for the relative poses between multiple area array cameras, thereby achieving global coordinate unification. The specific process is as follows:

[0128] Extract the subpixel center coordinates of the circular array markers in the calibration board images captured by the first and second area array cameras, sort the center coordinates in ascending order in the XY direction, and then generate the corresponding physical coordinates based on the spacing between each center point and the known circular array spacing. These physical coordinates are unified coordinates in the calibration board coordinate system.

[0129] Combining the intrinsic parameter matrix and distortion parameters of the first area array camera, the PNP algorithm is used to calculate the rotation matrix and translation vector from the calibration coordinate system to the coordinate system of the first area array camera.

[0130] Continuing to combine the intrinsic parameter matrix and distortion parameters of the second area array camera, the PNP algorithm is used to calculate the rotation matrix and translation vector from the calibration plate coordinate system to the second area array camera coordinate system;

[0131] Based on the above calculations, the rotation matrix and translation vector from the first area-array camera coordinate system to the second area-array camera coordinate system are derived. Using the same method, the rotation matrix and translation vector from the second area-array camera coordinate system to the third area-array camera coordinate system, and the rotation matrix and translation vector from the second area-array camera coordinate system to the fourth area-array camera coordinate system (if the third and / or fourth area-array camera is used to acquire C-standard high-definition grayscale images) coordinate system are obtained sequentially. Finally, a unified coordinate system for multiple area-array cameras is achieved, i.e., a unified global coordinate system is completed. After completing the unified global coordinate system, the creep displacement of the rails and sleepers is accurately calculated using the above displacement calculation method, providing an important basis for analyzing rail stability and formulating maintenance measures.

[0132] Branch area detection module:

[0133] The equipment consists of hardware that reuses the visual inspection module, but is equipped with a branch area detection algorithm. Different pulses are used to trigger whether the hardware is working as the visual inspection module or the branch area inspection module.

[0134] The working principle and steps are as follows:

[0135] A multi-type standard model was constructed based on CAD parametric modeling technology, and the model was corrected by combining on-site measurement data to establish a 3D model library containing four types of turnout areas (such as single turnouts and symmetrical turnouts). The model library integrates key data such as the dimensional parameters of each component of the turnout area (such as the standard width of the guard rail is 70mm), spatial coordinate information, and detection thresholds (such as the threshold for switch rail gap is 2mm), providing a standard reference for subsequent inspection work.

[0136] Two-level positioning is used for model matching, including coarse mileage matching and fine feature matching. Coarse mileage matching uses mileage data collected by the mileage encoder to quickly locate the current detection section and preliminarily determine the turnout position corresponding to the detection area, laying the foundation for subsequent fine matching. Fine feature matching extracts strong feature points in the turnout area, such as bolt centers (extracted using the RANSAC cylindrical detection algorithm) and switch rail tips (extracted using the linear endpoint fitting algorithm). Then, the RANSAC algorithm is used to match the extracted feature points with feature points in the standard model to determine the theoretical position of each component of the turnout. The matching error is controlled within ≤1mm to ensure the accuracy of the detection.

[0137] Model-driven pre-annotation involves pre-setting the boundary range of the detection area in the standard model, such as the bolt area (with a radius of 15mm and a height of 20mm) and the guard rail wheel groove measurement section (divided into opening section, buffer section, etc.). Then, these pre-set detection area boundaries are projected onto the actual collected point cloud data to form coarsely annotated areas, which indicate the key areas for subsequent fine-grained detection.

[0138] The location of measurement points is determined by precise segmentation of 3D features. By analyzing the shape characteristics (such as the cylindrical shape of bolts and the specific shape of elastic clips) and size information of each component in the turnout area, the target components are screened and segmented, and the structure of the target components is fitted, such as fitting the switch rail / guard rail plane and the cuboid structure of the sleeper, so as to accurately determine the location of each measurement point, such as the measurement point of the distance between the power rods in the switch rail contact area.

[0139] Specific qualitative and quantitative tests are conducted, including defects such as missing bolts, displacement of elastic clips, gaps between switch rails, centering of limiters, excessive grooves in guard rails, wear of guard rails, and excessive sleeper spacing.

[0140] Specifically, the detection methods for the seven types of diseases are as follows:

[0141] For bolt missing defects detection, the number of cylindrical features is counted within the pre-marked bolt area, and the count is compared with the number of bolts that should be present in the area in the standard model. If the standard model should contain 8 bolts for a certain area, and the number of detected bolts is less than 8, then the area is determined to have bolt missing defects.

[0142] The detection of elastic clip displacement defects involves identifying the position of the elastic clip using a 3D template matching algorithm, and then calculating the deviation of the clip's fixing point along the X-axis. If the deviation is greater than 3mm, the elastic clip is determined to have displacement defects.

[0143] For the detection of switch rail contact gap defects, within the "switcher-stock rail contact area between the power rods" that has been precisely identified, five measurement points are evenly selected (one point is set every 20mm along the mileage direction). The planes of the inner side of the switch rail and the inner side of the stock rail are fitted respectively, and the distance between the two planes in the X-axis at each measurement point (i.e., the gap value) is calculated. The average gap value of these five measurement points is taken. If the average value is >2mm, it is judged as a switch rail contact defect exceeding the standard.

[0144] For centering defect detection of the limit switch, first identify the contact area of ​​the limit switch. By identifying the U-shaped structural features of the limit switch, extract the gaps between the limit switch and the rail on both sides. Select three measurement points in each of these two gaps and calculate the gap width at each measurement point (i.e., the distance between the side of the limit switch and the side of the rail along the Y-axis). Take the average of the widths at each measurement point. If the average of the two gap widths is >7mm and the difference between the two averages is <2mm, the limit switch is considered to be centered correctly; otherwise, the limit switch is considered to be centered incorrectly.

[0145] For detecting defects in the guard rail wheel groove, four measurement points are selected for each of the "open section, buffer section, and straight section" of the guard rail wheel groove. Straight lines are fitted to the inner side of the guard rail and the inner side of the straight rail, and the distance between the two straight lines in the X-axis at each measurement point (i.e., the wheel groove width) is calculated. The calculated actual wheel groove width is then compared with the standard wheel groove width (e.g., the standard width of the open section is 2612mm, and the standard width of the buffer section is 1376mm) to obtain the deviation value. If the deviation value of any measurement point is ≥2mm, it is determined to be a defect in the guard rail wheel groove that exceeds the standard.

[0146] For guardrail wear detection, extract cross-sectional point cloud data in the guardrail flange contact area, fit the working edge of the cross-section (i.e. the two sides of the guardrail in contact with the flange), measure the distance between the two sides of the edge in the X-axis (actual guardrail width), and calculate the wear value = standard guardrail width (70mm) - actual guardrail width. If the wear value ≥ 2mm, it is judged as guardrail wear exceeding the standard.

[0147] The detection of sleeper spacing exceeding the standard involves identifying the cuboid region of two adjacent sleepers, calculating the geometric center coordinates of each sleeper, and then calculating the distance between the centers of adjacent sleepers along the Y-axis (mileage direction) (i.e., sleeper spacing). The actual sleeper spacing is compared with the standard sleeper spacing. If the deviation is ≥2mm, it is determined to be a sleeper spacing exceeding the standard.

[0148] High-precision fastener inspection module:

[0149] The equipment consists of a high-precision 3D camera and is equipped with algorithms for detecting gaps in fastener springs, measuring pad thickness, and identifying the type of insulating block / gauge baffle.

[0150] The fastener spring tongue gap detection algorithm uses a high-precision 3D camera to collect fastener depth data, and combines it with a preset neural network (YOLO) to intelligently identify and locate components such as fastener springs, bolts, and insulating gauge blocks. Then, it uses a 3D point cloud algorithm to fit and calculate the spring radius and locate the spring tongue clamping position by fitting the identified and segmented spring point cloud data. It measures the height of the spring tongue clamping position, further fits the insulating gauge block point cloud plane, and calculates the height of the insulating gauge block at the corresponding spring tongue clamping position. Finally, it measures the spring tongue gap value by using the spring tongue clamping position height, spring diameter, and corresponding insulating gauge block height. If the gap value is ≥1mm, the spring tongue is determined to be gapped.

[0151] The pad thickness measurement algorithm uses a high-precision 3D camera to collect 3D point cloud data of the fastener area, and combines deep learning algorithms to achieve semantic segmentation and localization of the rail, rail support platform, iron pad, and elastic strip. Then, the 3D point cloud algorithm performs planar fitting on the point cloud data corresponding to the segmented rail area, iron pad area, and rail support platform area. Finally, by calculating the planar distance between each pair of the rail plane, iron pad plane, and rail support platform plane, the total thickness of the pad, the thickness of the pad on top of the iron pad, and the thickness of the pad below the iron pad are measured.

[0152] The algorithm for identifying the model of insulating blocks / gauge baffles uses a high-precision 3D camera to collect 3D point cloud data of the fastener area. Combined with deep learning algorithms, it achieves semantic segmentation and localization of the rail, insulating block, gauge baffle, and elastic strip. Then, the point cloud algorithm performs straight-line fitting on the edge regions of the segmented rail, insulating block, and gauge baffle, fitting four straight lines: the rail edge line, the lines on the left and right sides of the insulating block, and the gauge baffle line. Finally, after calibration using the rail edge line, the distance between the two lines on either side of the insulating block is calculated and compared with the size of a standard insulating block to identify the insulating block model. Similarly, the distance between the gauge baffle line and the line on the side of the insulating block closest to the iron pad is calculated and compared with the distance between a standard gauge baffle and a standard insulating block to identify the gauge baffle model.

[0153] Compared with the prior art, this embodiment has at least the following technical advantages:

[0154] Significantly improved testing efficiency: Testing speed reaches 12-15km / h, with a daily testing mileage exceeding 50km, which is more than 20 times more efficient than traditional manual inspection and reduces labor costs by 90%.

[0155] High precision and comprehensive: Track geometry parameter measurement accuracy is ±0.3~±0.5mm, rail wear resolution is better than 0.2mm, and defect identification accuracy is ≥90%, covering all dimensions of track inspection needs;

[0156] It is highly flexible, adopts a modular and detachable design, and each detection module can operate independently or synchronously to adapt to different detection scenarios;

[0157] Good economic efficiency: The equipment has a design life of ≥5 years, and early warning of defects reduces maintenance costs for sudden failures by about 40%;

[0158] High level of intelligence: It realizes the alignment and automatic analysis of multi-source data, generates maintenance suggestions, reduces manual intervention, and promotes the transformation of track maintenance towards refinement and intelligence.

[0159] In some embodiments, the visual inspection module includes a 3D camera, a 2D camera, an LED light source, and a detection beam;

[0160] The 3D camera emits a line laser and receives the reflected signal to acquire a corresponding depth image; the 2D camera scans line by line to acquire a corresponding texture image; the LED light source provides an auxiliary light source, and the detection beam provides support;

[0161] The depth image and texture image are processed by a preset neural network to accurately identify the track type;

[0162] Based on structural differences data of different track types and prior data on defect distribution, the system can intelligently detect the condition and defects of track components.

[0163] In some embodiments,

[0164] The track geometry parameter detection module includes an inertial navigation measurement unit, an odometer, a visual measurement unit, a synchronization controller, and a processing computer;

[0165] The inertial navigation measurement unit uses laser inertial navigation to collect motion attitude data of the traveling platform;

[0166] The odometer is used to collect relative mileage data of the traveling platform within the area of ​​the rail to be measured;

[0167] The visual measurement unit is used to acquire images of the rail and calculate the attitude change information of the traveling platform relative to the rail based on the images of the rail.

[0168] The synchronization controller is used for the synchronous control of the system, and triggers the odometer, the vision measurement unit and the inertial navigation measurement unit to synchronously collect data by generating high-precision pulses;

[0169] The processing computer is used to solve the relative mileage data and the motion posture data through a combined navigation algorithm to generate the relative motion trajectory of the walking platform. The combined navigation algorithm is based on the Kalman filter algorithm.

[0170] The processing computer is also used to perform serpentine correction on the relative motion trajectory; the serpentine correction of the relative motion trajectory is achieved by using the attitude change information of the traveling platform relative to the rail to compensate for the spatial displacement caused by the serpentine motion of the traveling platform;

[0171] The processing computer obtains the motion trajectory of the left and right rail vertices based on the corrected relative motion trajectory, and extracts the track geometric parameters by calculating the motion trajectory of the left and right rail vertices.

[0172] In some embodiments, the rail profile detection module is used for:

[0173] High-precision point cloud data of rail cross-sections are obtained using 3D cameras.

[0174] Based on 3D point cloud technology, the measured rail profile features in the high-precision point cloud data are extracted and accurately registered with the preset standard rail profile.

[0175] The measured rail profile features after precise registration are compared point by point with the standard rail profile features. The deviation values ​​in the vertical direction, the deviation values ​​in the horizontal direction, and the total wear are calculated, and a continuous rail profile curve is generated.

[0176] In some embodiments, the rail crawling displacement detection module includes a crawling identification module, a synchronous trigger control module, an image acquisition module, and a data processing module;

[0177] The crawling marking module is used to set multiple crawling marking modules at equal intervals in the track area to be tested. Each crawling marking module includes the following markings: two sets of symmetrically arranged rail displacement markings fixed to the web of the rail, namely markings A1 and A2; two sets of symmetrically arranged sleeper reference markings fixed to the top surface of the rail sleeper, namely markings B1 and B2; and a positioning marking fixed to a fixed building beside the track, denoted as marking C.

[0178] The synchronous trigger control module includes an embedded microcontroller. The embedded microcontroller acquires the pulse count of the wheel encoder and generates synchronous trigger pulses at preset trigger intervals to control the image acquisition module to perform synchronous image acquisition, thereby ensuring that the image data acquired by the image acquisition module has time synchronization.

[0179] The image acquisition module consists of multiple area scan cameras for synchronous image acquisition. All area scan cameras transmit the acquired image data to the data processing module in real time. Specifically: the first area scan camera is aligned with the A1 and B1 tables to acquire high-definition grayscale images; the second area scan camera is aligned with the A2 and B2 tables to acquire high-definition grayscale images; and the third and / or fourth area scan camera is aligned with the C table to acquire high-definition grayscale images.

[0180] The data processing module pre-stores calibration and displacement calculation algorithms. It calculates the crawling measurement data of the rail measurement points on the rail side where the A1 rail is located using the displacement difference between rail C and rail A1; it calculates the crawling measurement data of the rail measurement points on the rail side where the A2 rail is located using the displacement difference between rail C and rail A2; it calculates the sleeper crawling measurement data on the rail side where the B1 rail is located using the displacement difference between rail C and rail B1; and it calculates the sleeper crawling measurement data on the rail side where the B2 rail is located using the displacement difference between rail C and rail B2.

[0181] In some embodiments, the branch area detection module includes a hardware device and a configured branch area detection algorithm. The hardware device reuses the hardware device of the visual inspection module, and the branch area detection algorithm includes:

[0182] Based on CAD parametric modeling and correction using on-site measured data, four types of 3D model libraries for branch areas were established. Each type of 3D model library integrates the dimensional parameters, spatial coordinate information, and detection thresholds of various components in the branch area.

[0183] Two-level positioning is used for model matching. The turnout area is coarsely located by the mileage data collected by the mileage encoder. Then, multiple strong feature points are extracted by collecting the point cloud data of the coarsely located turnout area. The RANSAC algorithm is used to match all strong feature points with the standard model in the 3D model library to determine the theoretical position of each component in the turnout area.

[0184] In the standard model of the 3D model library, the detection area boundary is pre-set for each component in the turnout area, and the pre-set detection area boundary is projected onto the collected point cloud data to form the coarse annotation area of ​​each component in the turnout area.

[0185] Based on the collected point cloud data, the shape features and size parameters of each component in the turnout area are analyzed, and the target components are screened and segmented based on the corresponding detection thresholds. At the same time, fitting processing is performed to accurately determine the measurement point position of the target component.

[0186] Within the coarsely marked area of ​​each component, based on the measured points of the determined target components, qualitative and quantitative detection is carried out on 7 types of defects. The 7 types of defects include bolt missing defects, elastic clip displacement defects, switch rail contact gap defects, limiter centering defects, guard rail wheel groove exceeding standard defects, guard rail wear defects, and sleeper spacing exceeding standard defects.

[0187] In some embodiments, the high-precision fastener detection module includes a 3D camera and a data unit, wherein the data unit integrates a fastener spring gap detection algorithm, a pad thickness measurement algorithm, and an insulating block and track gauge baffle model identification algorithm.

[0188] The fastener latch gap detection algorithm includes:

[0189] Acquire 3D point cloud data of the fasteners, including depth information, using a 3D camera;

[0190] By combining a preset neural network, the various components in the fastener are identified, segmented and located from the three-dimensional point cloud data of the fastener, and the spatial position of each component in the fastener is determined.

[0191] The radius of the spring bar is calculated by fitting the segmented spring bar point cloud data using a 3D point cloud algorithm, and the position of the spring tongue is located and the height of the spring tongue pressing position is measured.

[0192] The height of the insulating gauge block at the corresponding spring latching position is calculated by fitting the point cloud data of the identified and segmented insulating gauge block using a 3D point cloud algorithm.

[0193] Based on the elastic bar radius, the corresponding elastic bar diameter is obtained, and based on the elastic bar diameter, the height of the spring tongue clamping position, and the height of the insulating gauge block, the spring tongue gap value is measured.

[0194] The algorithm for measuring the thickness of the pad includes:

[0195] The fastener area is scanned using a 3D camera to collect three-dimensional point cloud data including the rail, rail support platform, iron pad, and elastic clip;

[0196] Based on deep learning algorithms, semantic segmentation is performed on the collected 3D point cloud data of rails, rail supports, iron pads and elastic rails to accurately distinguish and locate the spatial range of rails, rail supports and iron pads.

[0197] After semantic segmentation, obtain the three-dimensional point cloud data of rail, rail support platform and iron pad, and perform plane fitting using 3D point cloud algorithm to obtain the corresponding rail plane, rail support platform plane and iron pad plane.

[0198] The thickness of the upper pad is obtained by calculating the vertical distance between the rail plane and the iron pad plane; the thickness of the lower pad is obtained by calculating the vertical distance between the rail support platform plane and the iron pad plane; and the total thickness of the pad is obtained by calculating the vertical distance between the rail plane and the rail support platform plane.

[0199] The algorithm for identifying the model of the insulating block and track gauge baffle includes:

[0200] The fastener area is scanned using a 3D camera to collect three-dimensional point cloud data including rails, insulating blocks, gauge baffles, and elastic strips.

[0201] Based on deep learning algorithms, semantic segmentation is performed on the collected 3D point cloud data of rails, insulating blocks, gauge baffles and elastic bars to accurately distinguish and locate the spatial range of rails, insulating blocks and gauge baffles.

[0202] Based on the semantically segmented 3D point cloud data of the rail, insulation block and gauge baffle, the edge regions of the rail, insulation block and gauge baffle are extracted.

[0203] The edge regions of the rail, insulating block and gauge baffle are fitted with straight lines using 3D point cloud algorithm to obtain straight lines at the rail edge, straight lines on both sides of the insulating block and straight lines of the gauge baffle.

[0204] Using the straight line of the rail edge as a reference, the straight lines on both sides of the insulating block and the straight line of the gauge baffle are calibrated; the distance between the straight lines on both sides of the insulating block after calibration is calculated and compared with the size of the standard insulating block to achieve the identification of the insulating block model; the distance between the straight line of the gauge baffle and the straight line of the insulating block near the iron pad after calibration is calculated and compared with the distance between the standard gauge baffle and the insulating block to achieve the identification of the gauge baffle model.

[0205] In some embodiments, the system body includes a walking platform:

[0206] The traveling platform adopts a detachable structure, consisting of an active unit, a driven unit, an operating platform, a middle seat, and a rear seat; the active unit and the driven unit are interlocked by a mechanical structure; the operating platform is positioned by a pin and then locked using a spinning quick-locking structure; the middle seat, the rear seat, the active unit, and the driven unit are positioned by pins and then locked using a spinning quick-locking structure.

[0207] The system body also includes a power module for power supply, which is composed of multiple lithium batteries;

[0208] The system body also includes an electrical module and an integrated control module for operation control;

[0209] The electrical module includes a system control unit and an audible and visual warning unit. The system control unit is used to control the travel direction and speed of the traveling platform by receiving user commands, and to communicate with each module to display relevant information on the display screen. The audible and visual warning unit consists of a spotlight, a yellow warning light, and a high-frequency speaker.

[0210] The integrated control module provides an interactive UI interface, allowing users to interact with each module.

[0211] In some embodiments, the positioning module includes a high-precision encoder, a BeiDou GPS positioning module, and a sleeper number image recognition camera, used to achieve data position alignment among multiple modules within the system.

[0212] In some embodiments, the data processing module is equipped with a multi-sensor data fusion algorithm and an intelligent diagnostic model. The multi-sensor data fusion algorithm fuses the multi-source detection data, and the intelligent diagnostic model outputs detection results and maintenance suggestions.

[0213] Example 2

[0214] The multifunctional modular track inspection system provided in this embodiment is applied to the daily inspection of high-speed railways, specifically as follows:

[0215] Application scenario: A high-speed railway line (designed speed 350km / h) needs to conduct monthly track condition inspections, focusing on monitoring the stability of track geometric parameters, rail wear trends, and the condition of key structural components;

[0216] Configuration modules: running platform (5-seater, titanium alloy frame), track visualization inspection module (including 2D / 3D camera), track geometry parameter detection module (inertial navigation + visual fusion), rail profile detection module (SICK XR300 3D camera), positioning module (Beidou RTK + encoder + sleeper recognition camera) and data processing module (industrial computer + 4G / 5G transmission).

[0217] Work process: Before departure, set the task as monthly inspection of high-speed rail in the integrated control module, select the inspection speed of 15km / h, and enable synchronous data collection of the three modules;

[0218] Driving and Data Acquisition: The trolley runs at a constant speed of 15km / h, and the controller triggers each module to acquire data synchronously via high-precision differential pulses; the track geometry parameter detection module outputs parameters such as track gauge, level, and elevation (accuracy ±0.4mm); the rail profile detection module generates a rail cross-section point cloud every 10cm and calculates vertical and lateral wear; the visual inspection module identifies defects such as missing fasteners and rail scratches in real time (accuracy ≥92%).

[0219] Data processing: After the day's work is completed, the hard drive is inserted into the data processing workstation. The system automatically parses and calls the corresponding algorithm: integrates track geometry parameters and rail profile data to generate a track status heat map; compares with historical data to issue a warning that the monthly increase in rail side wear in a certain section has reached 0.3mm (exceeding the threshold of 0.2mm); automatically generates the "High-speed Rail Monthly Inspection Report" and recommends arranging grinding work at K128+500;

[0220] Output: The report is simultaneously uploaded to the maintenance platform of the engineering section to guide the maintenance plan;

[0221] Advantages include: daily inspection mileage exceeding 50km, with an efficiency three times that of traditional track inspection instruments; multi-parameter position alignment to avoid data fragmentation; and intelligent early warning to reduce the workload of manual verification.

[0222] Example 3

[0223] This embodiment provides a multifunctional modular track inspection system applied to the spring and autumn inspections of conventional speed railways, specifically as follows:

[0224] Application scenario: A conventional railway line (speed 120km / h) is conducting its annual spring inspection, which requires a comprehensive assessment of track geometry, structural defects, rail creep, and fastener fit.

[0225] Configuration modules: traveling platform, visual inspection module, track geometry parameter detection module, rail crawl displacement detection module (including A / B / C marking system), high-precision fastener detection module (DeepVision intelligent 3D camera), positioning module and data processing module;

[0226] Work process: Initial marking: Set up a set of creeping markers (A1 / A2 rail markers, B1 / B2 sleeper markers, and C trackside fixed markers) every 500m in the section to be inspected (approximately 20km).

[0227] Module switching and installation: In the morning, the track geometry parameter detection module and the visual inspection module are installed to complete track geometry parameter detection and preliminary screening of defects at 12km / h; in the afternoon, the module is replaced with the rail creep displacement detection module and the high-precision fastener detection module to complete fine inspection at 5km / h.

[0228] Crawling detection: When the trolley passes the marker point, four area array cameras are simultaneously triggered to take pictures of markers A, B, and C; the data processing module uses the PnP algorithm to unify the coordinate system and calculates the displacement of the rail relative to marker C; it is found that the left rail at K45+200 has crawled by 18mm (exceeding the limit by 10mm), triggering an early warning.

[0229] Fastener inspection: A 3D camera scans the fastener area and calculates the spring gap value; 37 failed fasteners with gaps ≥1mm are identified, with the location accurate to the sleeper number;

[0230] Comprehensive Report: The system integrates geometric deviation, creepage amount, fastener status, and generates the "Spring Inspection Comprehensive Assessment Report," proposing a joint rectification suggestion of "track lifting + fastener re-tightening + stress release."

[0231] Advantages include: enabling five types of inspection tasks to be deployed at once, avoiding multiple trips to the track; achieving rail crawling measurement and automated high-precision fastener measurement; and supporting modular quick-change design and multi-task operation in different time periods.

[0232] Example 4

[0233] This embodiment provides a multifunctional modular track inspection system applied to specialized inspection of subway turnout areas, specifically:

[0234] Application scenario: Abnormal noises from passing trains have recently occurred in the turnout area (single No. 9 turnout) of a subway main line in a certain city. Millimeter-level diagnosis of the geometric dimensions and structural condition of the turnout area is required.

[0235] Configuration modules: traveling platform, branch area detection module, positioning module, and data processing module;

[0236] Workflow: Model matching, the system locates the K8+320 turnout area through coarse mileage matching; strong feature points such as the tip of the switch rail and the center of the bolt are extracted, and the RANSAC algorithm is used to match the No. 9 turnout model (error ≤ 0.8mm).

[0237] Intelligent Detection: Switch rail contact gap: Average gap value of 2.3mm across 5 detection points → judged as excessive contact; Limiter centered: Average gap on both sides 8.1mm / 9.5mm, difference 1.4mm → acceptable; Guard rail wheel groove: Actual measurement of 1378mm at a certain point in the buffer section (standard 1376mm), deviation +2mm → exceeds the standard; Missing bolts: Detected 1 bolt area with only 7 cylindrical features (should be 8) → missing bolt alarm; 3D visualization: The system generates a point cloud of the switch area and an overlay map of the standard model, highlighting the areas exceeding the standard; Maintenance recommendations: The report recommends adjusting the switch rail contact device, replacing the guard rail buffer section, and reinstalling missing bolts, and includes a 3D positioning screenshot;

[0238] Advantages include: no new hardware required, high-precision turnout diagnosis achieved by reusing the visual inspection module; detection logic driven by 3D models, adaptable to multiple types of turnouts; millimeter-level quantitative analysis supports precise maintenance, avoiding empiricist misjudgments.

[0239] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computing software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0240] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0241] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A multi-functional modular rail integrated inspection system, characterized in that, The light-weight mobile trolley is used as a carrier, and comprises: a system body, which is used to provide a stable running platform, power supply and operation control, and realizes control and synchronization of all modules in a multifunctional detection module by collecting differential pulse signals output by high-precision encoders in a positioning module into a vehicle controller; a positioning module, which is used to realize position alignment of multi-module data in the system; a multifunctional detection module, which comprises a visual inspection module, a track geometric parameter detection module, a rail profile detection module, a rail climbing displacement detection module, a turnout detection module and a high-precision fastener detection module, all of which are mechanically connected through a reserved mechanical interface and electrically connected through a reserved electrical interface on the system body; a data processing module, which is used to process multi-source detection data of the multifunctional detection module and output detection results and maintenance suggestions; the turnout detection module comprises hardware devices and a set of turnout detection algorithm, the hardware devices reuse hardware devices of the visual inspection module, and the turnout detection algorithm comprises: based on CAD parameterized modeling and combined with field test data for correction, 3D model libraries of four types of turnout are established, each of which integrates size parameters, spatial coordinate information and detection thresholds of each component in the turnout; two-stage positioning is adopted for model matching, mileage data collected by a mileage encoder is used for coarse positioning of the turnout area, point cloud data of the coarse positioning turnout area is collected to extract a plurality of strong feature points, and RANSAC algorithm is used to match all strong feature points with standard models in the 3D model library to determine the theoretical positions of each component in the turnout area; detection area boundaries of each component in the turnout area are pre-set in the standard models in the 3D model library, and the pre-set detection area boundaries are projected into the collected point cloud data to form coarse labeling areas of each component in the turnout area; based on the collected point cloud data, shape features and size parameters of each component in the turnout area are analyzed, and based on corresponding detection thresholds, target components are screened and segmented, and fitting processing is performed to accurately determine the positions of target component measurement points; in the coarse labeling areas of each component, 7 types of diseases are qualitatively and quantitatively detected based on the determined target component measurement points, the 7 types of diseases including bolt missing disease, spring bar displacement disease, point rail close-to-seam disease, limiter centering disease, guard rail groove over-standard disease, guard rail wear disease and sleeper spacing over-standard disease.

2. The multi-functional modular track integrated detection system according to claim 1, wherein, the visual inspection module comprises a 3D camera, a 2D camera, an LED light source and a detection beam; the 3D camera emits linear laser and receives reflected signals to obtain corresponding depth images; the 2D camera scans line by line to obtain corresponding texture images; the LED light source provides auxiliary light source, and the detection beam provides support; the depth images and the texture images are processed by a pre-set neural network to accurately identify track types; based on structural difference data and disease distribution prior data of different track types, track component states and diseases are intelligently detected.

3. The multi-functional modular track integrated detection system according to claim 1, wherein, The track geometry parameter detection module comprises an inertial navigation measurement unit, an odometer, a visual measurement unit, a synchronous controller and a processing computer; The inertial navigation measurement unit collects motion posture data of the walking platform by laser inertial navigation; The odometer is used to collect relative mileage data of the walking platform in the steel rail area to be measured; The visual measurement unit is used to collect images of the steel rail and calculate the posture change information of the walking platform relative to the steel rail based on the images of the steel rail; The synchronous controller is used for synchronous control of the system, and high-precision pulse is generated to trigger the odometer, the visual measurement unit and the inertial navigation measurement unit to synchronously collect data; The processing computer is used to solve the relative mileage data and the motion posture data by a combined navigation algorithm to generate the relative motion trajectory of the walking platform, and the combined navigation algorithm is constructed based on a Kalman filtering algorithm; The processing computer is also used to correct the relative motion trajectory in a serpentine manner; the correction of the relative motion trajectory in a serpentine manner is achieved by compensating the spatial displacement caused by the serpentine motion of the walking platform by using the posture change information of the walking platform relative to the steel rail; The processing computer obtains the motion trajectories of the left and right rail vertices according to the corrected relative motion trajectory, and extracts the track geometry parameters by calculating the motion trajectories of the left and right rail vertices.

4. The multi-functional modular track integrated detection system of claim 1, wherein, The steel rail profile detection module is used to: acquire high-precision point cloud data of a steel rail section by a 3D camera; extract measured steel rail profile features from the high-precision point cloud data based on 3D point cloud technology, and accurately register the measured steel rail profile features with a preset standard steel rail profile; compare the accurately registered measured steel rail profile features with the standard steel rail profile feature point by point, calculate the deviation values in the vertical direction, the deviation values in the horizontal direction and the total wear amount, and generate a continuous steel rail profile curve.

5. The multi-functional modular track integrated detection system according to claim 1, wherein, The steel rail crawling displacement detection module comprises a crawling identification module, a synchronous trigger control module, an image acquisition module and a data processing module; The crawling identification module is used to set multiple crawling identification modules at equal intervals in the track area to be measured, and each crawling identification module comprises the following identification: two groups of symmetrically arranged steel rail displacement identification fixed on the rail waist, which are A1 mark and A2 mark respectively; two groups of symmetrically arranged sleeper reference identification fixed on the top surface of the rail sleeper, which are B1 mark and B2 mark respectively; and a positioning identification fixed on the adjacent fixed building, which is C mark; The synchronous trigger control module comprises an embedded microcontroller, which generates a synchronous trigger pulse for controlling the image acquisition module to synchronously acquire images according to a preset trigger condition interval by acquiring wheel encoder pulse count, so as to ensure that the image data acquired by the image acquisition module has time synchronization. The image acquisition module is composed of multiple area array cameras for image synchronous acquisition, and all the area array cameras transmit the collected image data to the data processing module in real time, wherein: a first area array camera is aligned with A1 mark and B1 table for high-definition gray image acquisition; a second area array camera is aligned with A2 mark and B2 mark for high-definition gray image acquisition; a third and / or fourth area array camera is aligned with C mark for high-definition gray image acquisition; The data processing module pre-stores calibration algorithm and displacement calculation algorithm, and obtains the rail measurement point creep measurement data on the side of the A1 mark through the displacement difference between C mark and A1 mark; obtains the rail measurement point creep measurement data on the side of the A2 mark through the displacement difference between C mark and A2 mark; obtains the sleeper creep measurement data on the side of the B1 mark through the displacement difference between C mark and B1 mark; and obtains the sleeper creep measurement data on the side of the B2 mark through the displacement difference between C mark and B2 mark.

6. The multi-functional modular track integrated detection system according to claim 1, wherein, The high-precision fastener detection module includes a 3D camera and a data unit, and the data unit is integrated with a fastener tongue gap detection algorithm, a pad thickness measurement algorithm, and an insulating block and gauge baffle type identification algorithm; The fastener tongue gap detection algorithm includes: collecting three-dimensional point cloud data of the fastener containing depth information through the 3D camera; combining a preset neural network, identifying, segmenting and positioning each component in the fastener from the three-dimensional point cloud data of the fastener, and determining the spatial position of each component in the fastener; calculating the radius of the spring strip by fitting the segmented spring strip point cloud data using a 3D point cloud algorithm, positioning the tongue clamping position, and measuring the height of the tongue clamping position; calculating the height of the insulating gauge block at the corresponding tongue clamping position by fitting the segmented insulating gauge block point cloud data using a 3D point cloud algorithm; based on the spring strip radius, obtaining the corresponding spring strip diameter, and based on the spring strip diameter, the tongue clamping position height and the insulating gauge block height, realizing the measurement of the tongue gap value; The pad thickness measurement algorithm includes: scanning the fastener area through the 3D camera to collect three-dimensional point cloud data of the steel rail, the rail support platform, the iron pad and the spring strip; based on a deep learning algorithm, performing semantic segmentation on the three-dimensional point cloud data of the steel rail, the rail support platform, the iron pad and the spring strip to accurately distinguish and position the spatial range of the steel rail, the rail support platform and the iron pad; obtaining the three-dimensional point cloud data of the steel rail, the rail support platform and the iron pad after semantic segmentation, and respectively performing plane fitting on the three-dimensional point cloud data of the steel rail, the rail support platform and the iron pad to obtain the corresponding steel rail plane, rail support platform plane and iron pad plane; by calculating the vertical distance between the steel rail plane and the iron pad plane, the thickness of the upper pad of the iron pad is obtained; by calculating the vertical distance between the rail support platform plane and the iron pad plane, the thickness of the lower pad of the iron pad is obtained; by calculating the vertical distance between the steel rail plane and the rail support platform plane, the total thickness of the pad is obtained; The insulating block and gauge baffle type identification algorithm includes: scanning the fastener area through the 3D camera to collect three-dimensional point cloud data of the steel rail, the insulating block, the gauge baffle and the spring strip; Based on the deep learning algorithm, the three-dimensional point cloud data of the collected rail, insulating block, track gauge baffle and spring strip are subjected to semantic segmentation, and the spatial range of the rail, insulating block and track gauge baffle is accurately distinguished and positioned; Based on the three-dimensional point cloud data of the rail, insulating block and track gauge baffle after semantic segmentation, the edge region of the rail, insulating block and track gauge baffle is extracted; The edge region of the rail, insulating block and track gauge baffle is subjected to straight line fitting through the 3D point cloud algorithm, and the rail edge straight line, insulating block two-side straight line and track gauge baffle straight line are obtained; The insulating block two-side straight line and track gauge baffle straight line are calibrated based on the rail edge straight line as the reference; the distance between the calibrated insulating block two-side straight lines is calculated and compared with the standard insulating block size to realize the insulating block model identification; the distance between the calibrated track gauge baffle straight line and the straight line close to the iron pad side of the insulating block is calculated, and compared with the distance between the standard track gauge baffle and the insulating block to realize the track gauge baffle model identification.

7. The multi-functional modular rail integrated inspection system of claim 1, wherein, The system body includes a walking platform: The walking platform adopts a detachable structure and is composed of a driving part, a driven part, an operation table, a middle seat and a tail seat; the driving part and the driven part are interlocked through a mechanical structure; the operation table is positioned by a bolt and locked by a spinning type quick structure; the middle seat, the tail seat, the driving part and the driven part are positioned by a bolt and locked by a spinning type quick structure; The system body also includes a power module for realizing power supply, which is composed of multiple lithium batteries; The system body also includes an electrical module and a comprehensive control module for realizing operation control; The electrical module includes a system control unit and an audible and visual warning unit; the system control unit is used to control the driving direction and speed of the walking platform by receiving user instructions, and communicates with each module to display relevant information on the display screen; the audible and visual warning unit is composed of a spotlight, a yellow warning light and a high-frequency horn; The comprehensive control module is used to provide an interactive UI interface for users to interact with each module.

8. The multi-functional modular rail integrated inspection system of claim 1, wherein, The positioning module includes a high-precision encoder, a Beidou GPS positioning module and a sleeper number image recognition camera, which is used to realize the position alignment of multiple module data in the system.

9. The multi-functional modular rail integrated inspection system of claim 1, wherein, The data processing module is equipped with a multi-sensor data fusion algorithm and an intelligent diagnosis model, which fuses the multi-source detection data through the multi-sensor data fusion algorithm and outputs the detection results and maintenance suggestions through the intelligent diagnosis model.

Images