Water tunnel disease detection device and method based on laser line scanning camera array

Abstract

The invention discloses a water conveyance tunnel disease detection device and method based on a laser line scanning camera array. The device comprises a mobile detection platform, a laser line scanning camera array, a posture and position measurement unit, a synchronous control unit, a data acquisition and processing unit and a disease identification module. The laser line scanning camera array is annularly arranged in the circumferential direction of the section of the tunnel, and line scanning gray imaging is carried out under active illumination of near-infrared laser. The synchronous control unit controls the trigger frequency and the exposure time based on the minimum recognizable scale of the disease, and motion blur is inhibited; the posture and position measuring unit provides PPS time synchronization and event time marks, and high-precision alignment of images and posture data is achieved; the data acquisition and processing unit completes image correction, splicing and spatial registration to generate a two-dimensional expanded image of the inner wall of the tunnel; the disease identification module automatically identifies crack and leakage diseases based on the image. Full-coverage scanning of the inner wall of the tunnel can be realized under the condition of high-speed continuous movement, and the inspection efficiency and the detection precision are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of structural inspection and safety monitoring technology for water conservancy engineering tunnels, specifically to a device and method for detecting defects in the inner wall of water conveyance tunnels based on a laser line scan camera array. It is applicable to the detection of cracks and / or leakage defects in the inner wall of water conveyance tunnels, water diversion tunnels, pressure tunnels, and other hydraulic tunnels under the condition of venting and inspection. Background Technology

[0002] Water conveyance tunnels are common water distribution structures in water conservancy projects, typically characterized by large spatial dimensions, humid environments, and long service lives. Affected by factors such as changes in surrounding rock stress, variations in construction quality, aging of lining materials, and seepage and hydraulic erosion, the tunnel walls may develop cracks and leaks during long-term operation, potentially accompanied by localized erosion and exposed reinforcement. If these defects are not detected and addressed promptly, they can lead to further leakage, decreased lining performance, and even localized instability, impacting the safety of the project and the reliability of water supply. Therefore, regular inspection of tunnel wall defects is necessary.

[0003] Existing methods for detecting tunnel defects mainly include manual inspection, image detection based on ordinary area array cameras, and 3D laser scanning. Manual inspection relies on experience, is inefficient, and the consistency of results is affected by human factors. Ordinary area array camera imaging is highly dependent on lighting conditions such as low illumination and damp reflection in tunnels, and is also easily affected by motion blur in mobile inspection conditions, making it difficult to reliably acquire images that meet the requirements for identifying minute cracks or leakage traces. Although 3D laser scanning can acquire spatial information, the equipment cost and deployment complexity are high, and the scanning efficiency and adaptability to continuous movement are limited, making it difficult to balance detection efficiency and data quality in long-distance tunnel scenarios.

[0004] Furthermore, the inner walls of water conveyance tunnels are typically continuous curved surfaces with localized construction joints, expansion joints, or ancillary structures. Single-view or single-point acquisition methods are easily affected by obstructions, creating blind spots in the detection. To improve detection efficiency, the detection equipment often needs to move continuously within the tunnel. However, under high-speed continuous movement conditions, how to suppress spatial image shift corresponding to a single scan while maintaining resolution, and how to ensure stable correspondence between the acquired image data and the pose, mileage, and other information of the moving platform under the same temporal reference, are key issues affecting the accuracy and repeatability of the detection.

[0005] Therefore, there is an urgent need for a defect detection device and method for tunnel curved surface environment and high-speed continuous movement conditions, which can realize the synchronous correlation of image acquisition with position and mileage information while collecting data across the entire cross section, thereby improving the engineering applicability and stability of defect detection such as cracks and leaks. Attached Figure Description

[0006] Figure 1This is a schematic diagram of the overall structure of a water conveyance tunnel defect detection device based on a laser line scan camera array according to an embodiment of the present invention;

[0007] Figure 2 A schematic diagram of a laser line scanning camera array arranged in a ring along the cross-section of a water conveyance tunnel;

[0008] Figure 3 A schematic diagram of the structure and imaging relationship for integrating linear laser active illumination and performing line scanning grayscale imaging in a single laser line scanning camera;

[0009] Figure 4 This is a connection diagram for synchronization control and trigger allocation based on PPS time base, event time scale and DMI mileage constraints;

[0010] Figure 5 This is a schematic diagram illustrating the timing relationship between PPS, camera trigger / event input, and exposure (or line sampling);

[0011] Figure 6 A schematic diagram illustrating the collaborative constraints of image line frequency, shutter time, and movement speed under high-speed continuous movement conditions, as well as the relationship between the single-line spatial image shift threshold.

[0012] Figure 7 A flowchart of the data processing process for the data acquisition and processing unit;

[0013] Figure 8 This is a schematic diagram of the processing flow of the disease identification module (including image correction, image stitching, spatial registration, and crack / leakage identification).

[0014] Figure 9 This is a flowchart illustrating a method for detecting defects in water conveyance tunnels based on a laser line scan camera array, according to an embodiment of the present invention. Summary of the Invention

[0015] The purpose of this invention is to provide a device and method for detecting defects in water conveyance tunnels based on a laser line scan camera array, so as to stably acquire line scan grayscale images of the tunnel inner wall under continuous movement (including high-speed continuous movement) during the emptying inspection condition, and achieve high-precision and continuous detection of cracks and / or leakage defects, thereby improving the efficiency and detection accuracy of long-distance water conveyance tunnel inspection.

[0016] To achieve the above objectives, the present invention adopts the following technical solution:

[0017] A device for detecting defects in water conveyance tunnels based on a laser line scan camera array includes:

[0018] A mobile inspection platform is used to carry a laser line scanning camera array and move continuously along the axial direction of the water conveyance tunnel.

[0019] A laser line scan camera array, installed on the mobile detection platform, includes multiple laser line scan cameras equipped with near-infrared linear laser active illumination components. The multiple laser line scan cameras are arranged in an array along the circumference of the tunnel cross section. They are used to project linear near-infrared lasers onto the inner wall of the tunnel during movement to form illumination strips and simultaneously acquire line scan grayscale image data at the corresponding positions.

[0020] The attitude and position measurement unit is used to acquire the displacement and attitude information of the mobile detection platform during the movement process, and to provide PPS time synchronization signal and event input terminal;

[0021] The synchronization control unit is used to receive the PPS time synchronization signal to establish a unified time reference, and to coordinate the trigger line frequency and single-line exposure integration time of the laser line scanning camera array, and to perform unified timing control of the laser line scanning camera array and the mobile detection platform to ensure the time synchronization of each collected data during the movement, and to control the mobile detection platform to move at a constant speed along the central axis of the water conveyance tunnel through the signal to maintain a stable speed and relative position.

[0022] The data acquisition and processing unit is used to simultaneously acquire the line scan grayscale image data of the laser line scan camera array and the displacement and attitude information obtained by the attitude and position measurement unit, and to perform image correction, stitching and spatial registration processing on the line scan grayscale image data to generate two-dimensional unfolded image information of the tunnel inner wall.

[0023] The defect identification module is used to identify and analyze cracks and / or leakage defects in the tunnel inner wall based on the two-dimensional unfolded image information.

[0024] Furthermore, the synchronization control unit is used to pre-set a maximum axial spatial sampling interval threshold based on the minimum identifiable scale of cracks and / or leakage defects. With the threshold for allowing single-line spatial image shift And during the continuous movement of the mobile detection platform, based on the platform's instantaneous speed The trigger line frequency of the laser line scan camera array With single-line exposure time To achieve coordinated control, the following can be achieved:

[0025] , ;

[0026] This simultaneously satisfies the imaging conditions of sufficient axial sampling and controlled single-row image shift, ensuring effective identification of cracks and / or leakage defects under continuous movement conditions.

[0027] Furthermore, the threshold for allowing single-line spatial image shifting The size should not exceed 1 mm, or be no larger than the object length corresponding to a single pixel or a preset multiple thereof, to ensure the distinguishability of fine cracks and seepage textures in two-dimensional grayscale images.

[0028] Furthermore, the laser line scanning camera array is arranged in a ring or segmented ring along the tunnel cross-section to achieve full or near full cross-section coverage scanning of the tunnel inner wall and reduce detection blind spots caused by the tunnel's curved surface structure or ancillary structures.

[0029] Furthermore, the attitude and position measurement unit includes a POS / IMU and a DMI odometry interface, and provides a PPS time synchronization signal output terminal and an event input terminal;

[0030] The synchronization control unit includes a time base module, a trigger generation and allocation module, and an event marking module;

[0031] The time reference module is used to receive the PPS time synchronization signal to establish a unified time reference;

[0032] The trigger generation and allocation module is used to generate and allocate camera array trigger signals based on the unified time reference.

[0033] The event marking module is used to input the camera array trigger signal, exposure valid signal or line valid signal into the event input terminal of the attitude and position measurement unit to record the event timestamp and establish the time correspondence between the image acquisition reference time and the displacement information and attitude information.

[0034] Furthermore, the trigger generation and allocation module is implemented using programmable logic devices. Under the unified time reference constraint, it generates low-jitter trigger pulses and outputs them in parallel to the trigger input terminals of each line scan camera through a fan-out, buffer, and impedance-matched allocation network. This ensures that the trigger timing difference between cameras is controlled within the microsecond range, and the single-line exposure integration time is also controlled. satisfy ,in The set trigger line frequency for the laser line scanner camera, and Not less than 1 kHz.

[0035] Furthermore, the trigger generation and allocation module includes at least the following two trigger modes:

[0036] Time-triggered mode, triggered by the set laser line scan camera's trigger frequency. Output trigger;

[0037] In distance-triggered mode, the distance increment or pulse output from the DMI odometry interface of the attitude and position measurement unit is triggered according to a preset axial distance step. Generate a trigger signal to ensure that the axial spatial sampling interval of adjacent scan lines meets the following conditions. ;

[0038] When configuring dual DMIs, it is preferable to designate one of them as the distance trigger reference and the other as the redundancy consistency check and slippage identification. When the output difference of the dual DMIs exceeds the preset threshold and outputs an abnormal quality flag, the synchronization control unit switches to the time trigger mode and / or adaptively adjusts the trigger parameters to improve trigger stability and image scale consistency under complex working conditions.

[0039] Furthermore, the data acquisition and processing unit includes an image correction submodule, an image stitching submodule, and a spatial registration submodule, wherein:

[0040] The image correction submodule is used to perform uniformity correction processing on the line scan grayscale data output line by line by line from each laser line scan camera, including at least dark current and gain uniformity and brightness equalization.

[0041] The image stitching submodule is used to register and stitch the imaging strips of multiple cameras in the circumferential direction of the tunnel cross section according to the camera array installation calibration parameters, and to perform stitching consistency verification and stitching parameter constraints in the overlapping areas of adjacent fields of view in order to suppress stitching misalignment and scale drift.

[0042] The spatial registration submodule is used to bind line scan data with displacement information, attitude information, and DMI mileage one-to-one under the unified time reference and event time stamp recording framework established by the synchronization control unit. The binding result is used as a constraint to perform motion compensation and spatial registration on the stitching result, generating two-dimensional unfolded image information with axial mileage as row coordinates and circumferential unfolded coordinates as column coordinates. The two-dimensional unfolded image information is a two-dimensional unfolded grayscale image or image block, and the synchronization index field bound to the two-dimensional unfolded image information is output.

[0043] Furthermore, the near-infrared linear laser active illumination component includes a near-infrared linear laser source and a cylindrical lens. The cylindrical lens is used to spread the laser beam to form a linear illumination strip. A bandpass filter matching the emission band of the laser source is placed in front of the camera lens to suppress the influence of ambient stray light on the linear scan grayscale imaging.

[0044] A method for detecting defects in water conveyance tunnels based on a laser line scan camera array, using the aforementioned device, includes the following steps:

[0045] S1. Control the mobile detection platform to move continuously along the axis of the water conveyance tunnel at a preset speed;

[0046] S2. During the movement, the laser line scan camera array synchronously acquires line scan grayscale image data of the tunnel inner wall under near-infrared laser active illumination, and records the event timestamp.

[0047] S3. Obtain the displacement and attitude information of the mobile detection platform, perform time matching and binding of the line scan grayscale image data with the displacement and attitude information according to the event time stamp, and complete motion compensation and spatial registration.

[0048] S4. The line scan grayscale image data after motion compensation and spatial registration is stitched together to generate two-dimensional unfolded image information of the tunnel inner wall.

[0049] S5. Based on the two-dimensional unfolded image information, perform automatic identification and analysis of cracks and / or leakage defects.

[0050] Compared with existing technologies, this invention, through the combined design of "near-infrared linear laser active illumination + multi-line scanning camera circumferential array coverage + PPS event time marker alignment + line frequency / exposure and speed coordinated constraints", enables full or near full cross-section coverage scanning of the tunnel inner wall and high-precision defect detection under continuous movement, especially high-speed continuous movement conditions, which significantly improves the efficiency and detection accuracy of long-distance water conveyance tunnel inspection. Detailed Implementation

[0051] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. It should be understood that the embodiments described in this specification are only used to illustrate the technical solutions of the present invention and do not constitute a limitation on the scope of protection of the present invention; any equivalent modifications or substitutions made by those skilled in the art without departing from the spirit and substance of the present invention should fall within the scope of protection of the present invention.

[0052] Example 1

[0053] See Figures 1 to 2 This embodiment provides a device for detecting defects in water conveyance tunnels based on a laser line-scanning camera array. The cross-section of the water conveyance tunnel being tested is exemplified by a city gate shape (arched upper section, straight lower section). For instance, the upper arch can be approximated as a semicircle with a diameter of approximately 6.4 m, and the lower straight section is approximately 6.4 m wide and 4 m high. These dimensions are merely parameters for this embodiment and are for illustrative purposes only, not constituting a limitation of the invention. The device includes a mobile detection platform 1, a laser line-scanning camera array 2, an attitude and position measurement unit 3, a synchronization control unit 4, a data acquisition and processing unit 5, and a defect identification module 6.

[0054] 1. Mobile Detection Platform

[0055] The mobile detection platform 1 is used to carry the laser line scan camera array 2, attitude and position measurement unit 3, synchronization control unit 4 and data acquisition and processing unit 5, and is in a controlled continuous movement state along the axis of the water conveyance tunnel, so as to provide a stable motion carrier and installation benchmark for the continuous acquisition of line scan grayscale images of the tunnel inner wall.

[0056] Considering that the bottom plate of the water conveyance tunnel may have silt, water accumulation and local unevenness under the condition of emptying inspection, and it is usually difficult to lay fixed guide rails in advance, the mobile inspection platform 1 preferably adopts a wheeled adaptive mobile structure, and improves the stability and reliability in humid environment and long-distance operation through vibration reduction and waterproof sealing design.

[0057] The mobile detection platform 1 can be a self-propelled or towed wheeled structure. The wheels can have a wide tread structure and be equipped with suspension and vibration damping components to enhance the ability to traverse uneven surfaces and reduce the impact of platform vibration on the stability of line scanning imaging. The walking mechanism can adopt a waterproof and sealed structure to adapt to the environment of tunnel floor slabs with water accumulation. When the platform is self-propelled, its drive system preferably uses a motor drive and is equipped with speed closed-loop control; when the platform is towed, a controlled traction speed can be provided by a traction device.

[0058] The platform's operating speed meets the continuity and efficiency requirements of the detection task. The rated operating speed of the mobile detection platform is not less than 1 m / s, preferably not less than 2 m / s, and more preferably 2 to 5 m / s. The platform speed is coordinated with the line frequency and exposure time of the laser line scanning camera to meet the preset single-line spatial image shift threshold control requirements under high-speed continuous movement conditions.

[0059] To improve the stability of speed control and timing coordination, the walking and drive control module of the mobile detection platform can communicate with the synchronous control unit 4 and receive displacement and speed information output by the attitude and position measurement unit 3. This information is used to compensate for or correct speed deviations caused by slippage, bumps, etc., thereby providing stable motion input for the synchronous control unit 4 to trigger allocation and sampling timing control based on a unified time reference, event time stamp recording and mileage constraints.

[0060] 2. Laser line scanning camera array

[0061] like Figures 1 to 3As shown in the figure, the laser line-scanning camera array 2 is installed on the mobile detection platform 1, and includes multiple line-scanning cameras and near-infrared linear active illumination components supporting each line-scanning camera. When the mobile detection platform 1 travels continuously along the axial direction of the water conveyance tunnel, each line-scanning camera sequentially collects the reflected gray-scale information of the corresponding position on the inner wall of the tunnel under the condition of the illumination strip formed by the active illumination; the fields of view of multiple cameras in the circumferential direction of the tunnel cross-section are sequentially covered and stitched and consistency checked in the overlapping area, so as to form a full-section or near-full-section two-dimensional unfolded image of the inner wall of the tunnel for subsequent disease identification and processing. In this embodiment, a two-dimensional line-scanning gray-scale imaging system using near-infrared active illumination is adopted, and three-dimensional reconstruction is not involved.

[0062] (2-1) Line-scanning Camera and Active Illumination Configuration

[0063] In this embodiment, the laser line-scanning camera array 2 preferably includes 4 to 6 line-scanning cameras, for example, 4 cameras; each line-scanning camera is equipped with a linear near-infrared active illumination component, such as Figure 3 the line-scanning camera body 2-1-1 and the linear near-infrared active illumination component 2-1-2 shown. The line array resolution of the line-scanning camera is not less than 4096×1 pixel, and the pixel size is about 7 μm; the camera trigger line frequency is not less than 1 kHz, and can be configured as 2 to 28 kHz under high-speed mobile detection conditions to meet the requirements of spatial axial sampling accuracy.

[0064] The linear near-infrared active illumination component preferably uses a linear near-infrared laser with a central wavelength of about 808 nm, and forms a linear illumination strip through a cylindrical lens. To suppress the influence of residual visible light, reflected stray light and device spontaneous radiation in the tunnel environment on gray-scale imaging, a band-pass filter matching the emission band is set in front of the camera lens, such as 808 nm±25 nm, so that the camera mainly receives the reflected energy in the illumination strip area, improves the contrast and consistency of the gray-scale image, and thus is conducive to subsequent two-dimensional unfolding and disease identification and processing.

[0065] (2-2) Spatial Resolution and Parameter Constraints

[0066] The spatial resolution of line-scanning imaging includes the lateral resolution along the scanning line direction and the longitudinal resolution along the platform movement direction . The lateral resolution is determined by the working distance and imaging geometry, and the longitudinal resolution is determined by the platform speed and trigger line frequency:

[0067]

[0068] When the physical width of the sensor is, equations (1) and (2) can be reduced to an approximate relationship:

[0069]

[0070] The vertical resolution is:

[0071]

[0072] in, , The unit is mm / pixel. Working distance (mm) Horizontal pixel count. is the physical width of the sensor (mm), and is the focal length (mm). The sensor pixel size (mm) is given. The platform's movement speed (mm / s) Trigger line frequency (line / s) for the camera.

[0073] To meet the requirements of controlled axial sampling and image movement for identifying cracks and / or leaks under high-speed continuous movement conditions, the synchronous control unit pre-sets a maximum axial spatial sampling interval threshold based on the minimum identifiable scale of the defect. With single-line image shift threshold and trigger line frequency With single-line exposure time Apply collaborative constraints so that:

[0074]

[0075] In one embodiment, Preferably less than 1 mm, when the platform speed When the speed is not less than 2 m / s, it is preferable to set the line frequency. Not less than 2 kHz, and combined with The control is used to suppress motion blur, thereby ensuring the distinguishability of crack and leakage features in the two-dimensional unfolded image.

[0076] (2-3) Field of view—assembly and stability of the lighting strip

[0077] The line scan camera and the linear near-infrared active illumination assembly are fixedly mounted on an integrated bracket, ensuring that the illumination strip stably falls within the camera's effective imaging field of view during platform movement. Assembly standards and calibration procedures constrain the position of the illumination strip, the angle of incidence, and the camera's optical axis direction to guarantee the consistency and repeatability of grayscale imaging under different cameras and operating conditions. This fixed mounting relationship is used for stable illumination of 2D grayscale imaging and does not involve 3D reconstruction.

[0078] (2-4) Array arrangement and coverage

[0079] like Figure 2As shown, multiple line scanning cameras are arranged in a ring or segmented ring along the circumference of the tunnel cross-section (see...). Figure 2 Each camera's optical axis points towards the tunnel's inner wall, with multiple fields of view sequentially covering the tunnel circumferentially. Adjacent camera fields of view maintain a 10° to 30° overlap in the circumferential direction for stitching and consistency verification of the two-dimensional unfolded images. This arrangement enables full-section or near-full-section coverage scanning of the tunnel's inner wall and reduces blind spots caused by the tunnel's curved surface structure or ancillary structures. For areas with significant obstruction, the blind spot range can be reduced by adjusting the installation angle or adding supplementary cameras.

[0080] Table 1 provides an example configuration of 4 cameras with approximately 255° circumferential coverage. Angles are defined as clock-like angular coordinates established around the tunnel cross-section, with 0° at 9 o'clock and 90° at 12 o'clock. Clockwise is positive and counterclockwise is negative along the circumference; the coverage target ranges from −37.5° to +217.5°.

[0081] Table 1 Recommended configuration parameters for laser line scan camera array (Example, 4 cameras, approximately 255° circumferential coverage)

[0082]

[0083] In this embodiment, the linear array resolution No less than 4096, pixel size Approximately 7 μm, working distance The typical range is 3 to 5 meters; line frequency Based on equation (4) and combined with equations (5) and (6), the settings are adaptively adjusted according to the platform speed, so that... The preset threshold requirements are met.

[0084] 3. Attitude and Position Measurement Unit

[0085] like Figure 4 , Figure 5 As shown, the attitude and position measurement unit 3 is used to acquire the displacement and attitude information of the mobile detection platform during its continuous axial movement through the tunnel, and provides a unified time reference and event time stamp data to the synchronization control unit 4 and the data acquisition and processing unit 5, enabling the line scan grayscale image data to be bound one-to-one with the pose information at the corresponding sampling reference time. The attitude and position measurement unit 3 includes a POS / IMU 31, a DMI odometry interface 32, a PPS time synchronization signal output terminal 33, and an event input terminal 34. The POS / IMU 31 is a combined navigation device integrating an inertial measurement unit and a GNSS receiver. The dual-antenna GNSS component is used for heading assistance, and the dual DMI odometry is used for mileage constraints and redundancy checks.

[0086] (3-1) POS / IMU 31 and Dual-Antenna Heading Assist

[0087] The POS / IMU 31 is used to output the three-dimensional position, three-dimensional attitude angles, velocity, and time information of the mobile detection platform, and outputs continuously at a high rate to meet the needs of continuous pose calculation under high-speed continuous movement conditions. The dual-antenna GNSS assembly is mounted on the mobile detection platform 1 using an ultra-short baseline method, preferably fixed to a rigid mounting bracket or base on the upper part of the platform to reduce relative displacement errors caused by driving vibrations. The baseline length between the phase centers of the two antennas is denoted as L. In this embodiment, L is 1.2 m, preferably 0.8 m to 1.5 m, and can be appropriately increased to improve heading stability when the platform structure allows. The dual-antenna baseline is used to provide heading assistance, shorten the initial heading alignment time, and enhance short-term heading stability, thereby reducing the impact of heading drift on the geometric consistency of the two-dimensional unfolded image.

[0088] In GNSS-enabled areas such as tunnel entrances or shaft openings, the POS / IMU 31 uses GNSS observations to constrain position and heading. When the platform enters the main tunnel section, causing GNSS limitations or obstruction, the POS / IMU 31 primarily relies on inertial estimation, incorporating DMI odometer information as velocity and displacement constraints for combined navigation filtering to obtain continuous pose output. To ensure the reproducibility of pose and image geometry, the assembly phase prioritizes completing the installation offset calibration of the POS / IMU coordinate system and the platform body coordinate system, the boom measurement from the POS / IMU reference point to the camera array reference point, and the consistency calibration of the dual-antenna baseline direction with the platform's longitudinal axis. These parameters serve as fixed inputs for image and pose binding and two-dimensional unfolding geometric correction.

[0089] (3-2) Dual DMI Odometer and DMI Odometer Interface 32

[0090] This embodiment is configured with dual DMI odometers, each associated with a different wheel / shaft system on the platform. Displacement increments are observed by counting pulses output from the encoder and received by the DMI odometer interface 32. The DMI output can be a pulse count and direction signal, or orthogonal A-phase and B-phase signals. The DMI odometer interface 32 counts the pulses and generates mileage increments. Mileage conversion is expressed and calibrated using the following relationship:

[0091]

[0092] in, No. The pulse count of the DMI per unit time. The mileage ratio coefficient is preferably determined by the wheel diameter and the equivalent number of pulses per revolution, and is corrected by a straight calibration section or a known distance scale to compensate for the ratio deviation caused by tire compression, wear and assembly errors.

[0093] Dual DMI is used to improve robustness under complex operating conditions: firstly, it addresses the two displacement increments. , Consistency verification is performed. When the difference exceeds the preset threshold, it is determined that there may be abnormalities such as slippage, freewheeling, or wheel diameter changes, and a quality flag is output. This flag is used to adaptively adjust the mileage observation weights in the integrated navigation filter or to remove abnormal observations. Secondly, the effective DMI velocity / displacement increment is used as the filter observation input POS / IMU 31 to suppress the accumulation of pure inertial navigation calculation errors, improve the continuity and stability of pose over time, and thus improve the axial scale consistency of line scan image stitching and two-dimensional unfolding.

[0094] (3-3) PPS time synchronization signal output terminal 33 and event input terminal 34

[0095] To ensure consistency between the sampling reference time and pose reference time of the line scan image, this embodiment employs a synchronization method that uses a unified time base (PPS) and the event input terminal to record edge timestamps. Figure 4 , Figure 5 As shown.

[0096] 1) PPS Unified Time Base

[0097] The attitude and position measurement unit 3 outputs a PPS time synchronization signal through the PPS time synchronization signal output terminal 33. This signal is input to the time reference module 41 of the synchronization control unit 4 to establish a unified time reference shared by the laser line scan camera array 2, the data acquisition and processing unit 5, and the attitude and position measurement unit 3. Under this unified time reference, the synchronization control unit 4 generates and distributes camera array trigger signals to drive multiple line scan cameras to sample synchronously.

[0098] 2) Event Mark records the sampling reference time.

[0099] The event marking module 43 of the synchronization control unit 4 sends the event input signal (one or more of the camera array trigger signal, camera exposure valid signal, and line valid signal) to the event input terminal 34 of the attitude and position measurement unit 3. The attitude and position measurement unit 3 records the event timestamp of the corresponding edge received by the event input terminal 34, forming event timestamp data on the same time axis as the pose output.

[0100] 3) Pose matching / interpolation binding based on event timestamps

[0101] After receiving line-scanned grayscale image data, pose data, and event time-stamped data, the data acquisition and processing unit 5 uses the event time-stamped data as the image sampling reference time and extracts the pose corresponding to the event time from the pose epoch sequence. When the event time and the pose output epoch do not coincide, the pose at the event time is obtained by time interpolation between two adjacent epochs.

[0102]

[0103] in, For event time stamp, This represents the set of pose parameters. This enables the binding of images and poses at the same reference time, providing traceable and reproducible spatiotemporal constraints for subsequent 2D unfolding, image stitching, and lesion identification.

[0104] The aforementioned structure and process enable the attitude and position measurement unit 3 to output continuous pose under both GNSS availability and GNSS obstruction conditions, and to accurately mark and match the image sampling reference time through PPS unified time reference and event time stamp recording. To ensure that the event time stamp is as close as possible to the actual sampling reference time of the line scan camera, when the camera has stable exposure valid or line valid output, its preset edge is preferably used as the event source; when the camera does not provide exposure valid or line valid output or the output is unstable, the edge of the camera array trigger signal is used as the event source. The event source signal shaping, level matching, isolation, and array unified event generation are completed by the event marking module 43 of the synchronization control unit 4.

[0105] 4. Synchronization control unit

[0106] Synchronization control unit 4 is used to perform unified timing control on laser line scanning camera array 2, attitude and position measurement unit 3, and motion detection platform 1, so that the line scanning grayscale image data and pose data achieve a one-to-one correspondence at the same reference time within the framework of unified time base and event time stamp recording; synchronization control unit 4 is also used to trigger line frequency With single-line exposure time Coordinated control is implemented to ensure that the spatial image shift of the mobile detection platform on the tunnel inner wall during single-line scanning under high-speed continuous movement does not exceed a preset threshold, thereby ensuring the distinguishability and identifiability of crack and / or leakage features in two-dimensional grayscale images.

[0107] like Figure 4 , Figure 5As shown, the synchronization control unit 4 includes a time reference module 41, a trigger generation and allocation module 42, and an event marking module 43. The time reference module 41 establishes a unified time reference with the PPS time synchronization signal output terminal 33 of the attitude and position measurement unit 3; the event marking module 43 forms an event time stamp recording link with the event input terminal 34 of the attitude and position measurement unit 3; the trigger generation and allocation module 42 outputs synchronous trigger signals to multiple line scan cameras to achieve array synchronous sampling. The trigger generation and allocation module 42 can also receive mileage increment or pulse information output from the DMI odometer interface 32 of the attitude and position measurement unit 3 for distance triggering or trigger quality control; in the dual DMI configuration, the two mileage differentials can be used to form a quality flag for trigger parameter adaptation or trigger mode switching.

[0108] (4-1) Time base module 41

[0109] The time reference module 41 receives the PPS signal output from the PPS time synchronization signal output terminal 33 of the attitude and position measurement unit 3 as the time reference for the entire system. It uses a local high-stability clock source to count and subdivide adjacent PPS pulses to form a unified time scale, preferably with a time resolution of no more than 1 μs. After isolation and fan-out, the PPS signal can be distributed to the trigger generation and distribution module 42 and the data acquisition and processing unit 5, so that the camera trigger timing, event time stamp recording and pose time axis share the same time reference, thereby providing a traceable time synchronization basis for subsequent image and pose alignment based on event time stamps.

[0110] The trigger generation and allocation module 42 is used to generate camera array trigger signals and perform fan-out allocation to drive multiple line scan cameras to sample synchronously. This module is preferably implemented using a programmable logic device. Under the unified clock constraint provided by the time base module 41, it generates low-jitter trigger pulses and outputs them in parallel to the trigger input terminals of each line scan camera through a fan-out, buffer, and impedance matching allocation network to ensure that the sampling time is consistent among the array cameras. Preferably, the difference in trigger timing between cameras is controlled within the microsecond range.

[0111] (4-2) Trigger Generation and Allocation Module 42

[0112] Trigger line frequency of trigger generation and allocation module 42 With single-line exposure time To perform collaborative settings, the generation and allocation module 42 is triggered to generate line frequencies. The single-line exposure integration time Δt is set in conjunction with the axial spatial sampling interval constraint and the single-line spatial image shift constraint, such as... Figure 6 As shown.

[0113] 1) Axial spatial sampling interval constraint

[0114] The time interval between two adjacent rows of scans is On the platform at speed During continuous movement, the axial spatial sampling interval corresponding to adjacent scan lines on the tunnel inner wall. satisfy:

[0115]

[0116] To avoid missing cracks or seepage textures due to excessively sparse axial sampling, it is required that... Not greater than the preset maximum axial sampling interval threshold Thus we get:

[0117]

[0118] in It can be preset based on the minimum identifiable size of the crack or leak to be inspected.

[0119] 2) Single-line spatial image shift constraint

[0120] Single-line scanning at exposure integration time Equivalent spatial image shift within satisfy:

[0121]

[0122] To suppress motion blur caused by high-speed movement, it is required that... Not greater than the allowed single-line image shift threshold Thus we get:

[0123]

[0124] in, Preferably, the diameter is no greater than 1 mm, and more preferably no greater than the object length corresponding to a single pixel or a preset multiple thereof, to ensure the distinguishability of fine cracks and seepage textures in the two-dimensional grayscale image. To ensure the timing feasibility of the line scan camera, It should also satisfy the constraint of not exceeding the row period:

[0125] Through the aforementioned synergistic constraints, imaging conditions that simultaneously satisfy sufficient axial sampling and controlled single-row image shift can be met under conditions of high-speed continuous platform movement.

[0126] The trigger generation and allocation module 42 can operate in both time-triggered and distance-triggered modes. In time-triggered mode, it operates according to a set line frequency. The output trigger signal is suitable for operating conditions with relatively stable speed control; the distance trigger mode is based on the distance increment or pulse output by the DMI odometer interface 32 of the attitude and position measurement unit 3, according to a preset axial distance step. Generate triggers so that adjacent scan lines naturally meet the requirements. The constraints ensure that the axial sampling interval of adjacent scan lines remains stable despite speed fluctuations. For the dual DMI configuration, one path can be designated as the distance trigger reference, while the other is used for redundancy consistency verification and slippage identification. When the quality indicator shows an abnormality, the trigger mode can be switched or the trigger parameters can be adaptively adjusted to improve trigger stability and image scale consistency under complex working conditions.

[0127] (4-3) Event Marker Module 43

[0128] Event marking module 43 is used to convert the image sampling reference time into an event edge signal that can be recorded by attitude and position measurement unit 3, and input it into event input terminal 34, so that attitude and position measurement unit 3 can form event time-stamped data under PPS unified time reference, which is used for subsequent image and pose alignment, such as Figure 5 As shown. The event signal source can be a camera array trigger signal, or an exposure valid signal or a line valid signal output by a line scan camera. To make the event time closer to the actual sampling reference time, it is preferable to use the preset edge of the exposure valid signal or the line valid signal as the event reference; when the camera does not provide or the output is unstable, the trigger signal edge is used as the event reference. The event marking module 43 performs level matching, isolation and shaping on the event source signal, and outputs a narrow pulse event signal according to the electrical specifications of the event input terminal 34.

[0129] In multi-camera array synchronous sampling scenarios, the event marking module 43 can use a common trigger signal as a unified event for the array, or select the valid exposure or valid line signal of the reference camera as a unified event, or logically synthesize multiple valid exposure or valid line signals after synchronous shaping to form a unified event signal, so as to ensure that each line or group of image data corresponds to a unique and traceable event time stamp input. Considering that the trigger link and event link may introduce a fixed propagation delay, this fixed delay can be measured during the assembly or calibration stage and recorded as a system constant for the data acquisition and processing unit 5 to perform delay compensation when aligning images and poses; the image and pose matching and interpolation methods and their calculation formulas are described in Section 3, Pose Interpolation Relationship.

[0130] 5. Data Acquisition and Processing Unit

[0131] The data acquisition and processing unit 5 is used to synchronously acquire the line scan grayscale data output by the laser line scan camera array 2 and the pose solution, DMI mileage, and quality marker output by the attitude and position measurement unit 3, under the unified time reference and event time stamp recording framework established by the synchronization control unit 4. It sequentially completes synchronous acquisition and storage, time alignment and data binding, image correction, multi-camera image stitching, motion compensation and spatial registration, and 2D unfolded image generation, outputting a 2D unfolded grayscale image or 2D unfolded image blocks divided by mileage, and outputting the synchronous index field bound to the image to the defect identification module 6. The data acquisition and processing unit 5 preferably includes an image correction submodule, an image stitching submodule, and a spatial registration submodule, which correspondingly complete... Figure 7 The processing links shown are (5-3) to (5-5).

[0132] (5-1) Synchronous acquisition and storage

[0133] The data acquisition and processing unit 5 receives line-scan grayscale data or trigger counts output line by line from multiple laser line scan cameras via a high-speed data interface. Simultaneously, it receives the pose solution, DMI odometer, and quality flag output from the attitude and position measurement unit 3, as well as the PPS time reference and trigger / exposure valid event input provided by the synchronization control unit 4. To ensure traceability of the processing, at least the following fields are recorded during the acquisition phase:

[0134] a) Image side: Camera number, row number or trigger count, single-row grayscale vector, trigger row frequency and single-row exposure integration time parameter, exposure valid or row valid status, and index field associated with event number;

[0135] b) Measurement side: pose output timestamp, position and attitude angle and velocity information, DMI cumulative mileage and mileage increment, DMI quality flag, event time stamp record data, wherein the event time stamp record data includes at least the event number and the event time.

[0136] The acquired data is written to the buffer queue and storage medium in a streaming manner, and the acquisition parameters and correction parameter version numbers are recorded for subsequent backtracking and verification. During the acquisition process, the row data can be preprocessed online, including bad pixel / bad line correction, dark current correction, gain uniformity and grayscale normalization, and the parameters used are recorded along with the data.

[0137] (5-2) Time alignment and data binding

[0138] The data acquisition and processing unit 5 receives pose timeline and event timestamp recording data under a unified time reference, and uses the event timestamp as the image sampling reference time to achieve one-to-one binding between image and pose. The event timestamp recording data is obtained by the attitude and position measurement unit 3 at the event input end. The event source is generated by the synchronization control unit 4 according to the strategy described in Section 4. The event source is preferably a preset edge of the exposure valid or line valid signal. When a stable edge cannot be obtained, the trigger signal edge is used as a substitute. When multiple camera arrays are sampled synchronously, the event timestamp recording adopts a unified event strategy for the array, so that each row or row group of image data corresponds to a unique and traceable event number and event time.

[0139] Data acquisition and processing unit 5 establishes an index relationship based on event numbers, mapping row numbers or trigger counts to event numbers. And obtain the event time from the event timestamp record. In the pose output sequence, locate and... Interpolate the time of two adjacent epochs to obtain the pose at the event time. When the attitude and position measurement unit provides event-triggered output pose, event-triggered output matching is used to directly obtain the pose. This results in an event-level binding record:

[0140]

[0141] Considering that the trigger link and event link may introduce a fixed propagation delay, the data acquisition and processing unit 5 reads the fixed delay constant measured during the assembly or calibration stage, performs delay compensation on the event time and / or image time index, and then performs matching and interpolation to improve the image and pose alignment accuracy.

[0142] (5-3) Image Correction

[0143] The data acquisition and processing unit 5 performs consistency correction processing on the grayscale line data from each camera, including at least dark current and gain consistency and brightness equalization; and can perform stripe suppression, local contrast stabilization, and geometric / scale consistency processing as needed to reduce the impact of differences in laser illumination, sensor response, and sampling conditions on the consistency of subsequent stitching and recognition. The parameters, version numbers, applicable ranges, and line data used for image correction are recorded together for result verification and traceability.

[0144] (5-4) Multi-camera image stitching

[0145] Data acquisition and processing unit 5 registers and stitches the imaging strips of each camera in the circumferential direction of the tunnel cross-section according to the camera array installation calibration parameters. For adjacent fields of view with circumferential overlap, stitching alignment is completed based on the overlapping area, and consistency checks and constraints are performed on the stitching seams to suppress stitching misalignment and scale drift caused by differences in illumination, viewing angle, or trigger jitter. If necessary, weighted fusion or gradient fusion is used in the overlapping area to reduce stitching seams caused by brightness differences, forming a circumferentially continuous line scan imaging strip, and outputting stitching parameters and verification results for subsequent traceability.

[0146] (5-5) Motion compensation and spatial registration to generate a 2D unfolded image

[0147] Data acquisition and processing unit 5 uses the event-level binding record 𝐴𝑘 as a constraint, and combines the pose solution and DMI odometry constraints to perform motion compensation and spatial registration on the stitching result, establish a two-dimensional unfolded coordinate system, and generate a two-dimensional unfolded grayscale image or image patch. The two-dimensional unfolded coordinate system uses the axial odometry 𝑍 as the row coordinate and the circumferential unfolded coordinate 𝑆 as the column coordinate:

[0148] a) Axial mileage consistency: When the synchronization control unit 4 is working in distance trigger mode, the camera triggers according to the preset distance step size, and the data acquisition and processing unit 5 maps the row index to the axial mileage 𝑍 and uses it for unfolding and writing; when working in time trigger mode, the data acquisition and processing unit 5 performs consistent mapping of the row spacing based on the DMI mileage or the axial displacement obtained by pose calculation, and performs remapping or resampling of the axial direction when necessary to maintain the consistency of the axial scale of the unfolded image.

[0149] b) Circumferential unfolding mapping: The mapping of circumferential unfolding coordinates 𝑆 is generated by camera intrinsic parameters, installation extrinsic parameters and tunnel cross-section geometric priors to generate mapping relationships or lookup tables, realizing a unified mapping from cell column index to circumferential unfolding position, and registering multi-camera stitched strips to unified unfolding coordinates.

[0150] c) Quality control and segmentation: When the dual DMI consistency check or slippage identification outputs an abnormal quality flag, the data acquisition and processing unit 5 marks the affected section with quality and reduces the mileage constraint weight, switches the mileage benchmark, or adopts conservative segmentation processing according to the preset strategy to ensure the controllability and traceability of the expansion scale and alignment results.

[0151] d) Block Management and Disk Writing: For long-distance continuous acquisition, the two-dimensional unfolded image is segmented by mileage or divided into blocks by a fixed number of rows to generate and write to the disk. The block number, the start and end mileage and time range of the block, the participating camera number, the stitching and fusion parameters, the calibration parameter version number, the DMI quality mark and the event sequence range are recorded for subsequent retrieval, positioning and backtracking.

[0152] The data acquisition and processing unit 5 outputs a two-dimensional unfolded grayscale image or a two-dimensional unfolded image block divided by mileage, and outputs the corresponding synchronization index field to the disease identification module 6; the synchronization index field includes at least the block number, the start and end mileage of the block and the time range, and the event sequence. and its corresponding pose solution The system includes DMI quality indicators, splicing and fusion parameters, and row sequence numbers or trigger counts that can be traced back to the original row data, thereby supporting the location, tracing, and verification of the defect identification results in the tunnel's axial mileage and circumferential position.

[0153] 6. Disease Identification Module

[0154] The defect identification module 6 receives the two-dimensional unfolded grayscale image or the two-dimensional unfolded image block divided by mileage output by the data acquisition and processing unit 5, and receives the synchronous pose and mileage index data bound to the image, which is used to establish the correspondence between the identification results and the tunnel's axial mileage and circumferential position. The defect identification module 6 follows... Figure 8 The process shown identifies cracks and / or leaks, measures and locates their parameters, and outputs traceable identification results.

[0155] (6-1) Input and Quality Control

[0156] The defect identification module 6 reads the two-dimensional unfolded grayscale image and its corresponding synchronization index field. The synchronization index field includes at least the image block number, block start and end distances and time range, event sequence and its corresponding pose solution, DMI quality flag, and row number or trigger count that can be traced back to the original row data. The defect identification module 6 performs checks and quality control on the input data, including at least missing row or block detection, time jump and index inconsistency detection, DMI quality flag verification and suspicious interval marking, and outputs the quality control results. For affected intervals, weight downgrading, result downgrading labeling or removal can be performed according to preset strategies to ensure the stability and traceability of the identification output.

[0157] (6-2) Residual image correction and block-level processing before identification

[0158] After the data acquisition and processing unit 5 has completed image correction, image stitching, and spatial registration and output a two-dimensional unfolded image, the defect identification module 6 can perform residual error correction and consistency processing on the unfolded image as needed to reduce non-defect interference introduced by the unfolding, fusion, or block boundaries. The processing includes at least brightness uniformity and local contrast stabilization, suppression of residual stripes and interline brightness fluctuations, marking of local overexposed or underexposed areas, and masking and ROI cropping of the detection area; the processing does not change the two-dimensional unfolding scale and positioning reference, but only improves the recognition quality.

[0159] When the two-dimensional unfolded image is output in the form of mileage segments or fixed number of rows blocks, the disease identification module 6 performs block-level splicing and seam consistency processing on adjacent image blocks according to the image block index field and fusion parameters to form an unfolded image sequence for continuous detection and statistics, and reduces the interference of seams on crack and leakage identification.

[0160] (6-3) Coordinate mapping and establishment of positioning reference

[0161] Based on the unfolded coordinate definition and index data provided by the data acquisition and processing unit 5, the defect identification module 6 establishes a mapping relationship between image coordinates and tunnel unfolded coordinates, and establishes a traceable positioning benchmark. The axial position is mapped to the axial mileage Z of the two-dimensional unfolded image based on the mapping relationship between the block start and end mileage and the row index; the circumferential position is mapped to the circumferential unfolded coordinate S based on the mapping parameters or lookup table from the pixel column index to the circumferential unfolded coordinate S, and can be converted to the circumferential angle θ as needed.

[0162] To support retrospection and verification, the disease identification module 6 establishes a traceability chain and binds and saves it with the identification results. The traceability chain includes at least the image block number, row number or trigger count, event index and its corresponding pose solution, and is associated with DMI quality mark and quality control results.

[0163] (6-4) Crack identification, measurement and evidence output

[0164] Cracks appear as thin, elongated lines or bands in a 2D unfolded grayscale image, exhibiting geometric abrupt changes and structural boundary features. Crack identification includes candidate generation, connectivity constraints and enhancement, false detection suppression, parameter measurement, and evidence output.

[0165] Candidate generation is based on edge response, gradient response, linear structural response or learning model to obtain crack candidates and suppress gradually varying background and random noise; connectivity constraint and enhancement are applied to the slender structure to constrain orientation consistency and connectivity, enhance the main crack structure and suppress non-crack textures; false detection suppression combines features such as slenderness ratio, orientation stability, local contrast and texture consistency to reduce false detections caused by lining texture, scratches, joints and so on.

[0166] Parameter measurements are used to calculate the length and orientation of the crack's centerline, estimate the crack width based on the distance between the two edges in the normal direction, and convert it to actual dimensions according to the spatial calibration scale of the two-dimensional unfolded image. Crack identification and measurement are based on the imaging conditions constrained by the axial sampling interval threshold and the allowable single-line spatial image shift threshold in Section 4, ensuring that crack features remain distinguishable and measurable in the unfolded image. The output includes at least the crack results, parameters, and corresponding evidence images.

[0167] (6-5) Leakage identification, characterization and evidence output

[0168] Leakage in a 2D unfolded grayscale image manifests as sheet-like or strip-like continuous regions, accompanied by grayscale and texture differences and boundary variation features. Leakage identification includes candidate generation, region growing and morphological constraints, false detection suppression, degree representation, and evidence output.

[0169] Candidate generation can extract leakage candidate regions based on grayscale statistical differences, local variance, gradient changes, or learning models. Region growing and morphological constraints screen candidate connected regions based on area, compactness, extension scale, and continuity to eliminate isolated noise patches and stabilize candidate morphology. False detection suppression combines features such as background texture consistency, boundary stability, and local contrast to reduce false detections caused by stains, shadows, or non-uniform lighting. Degree characterization outputs leakage area, axial and circumferential distribution range, and degree indicators for subsequent classification and statistics. The output should include at least the leakage result, characterization parameters, and corresponding evidence images.

[0170] The crack identification and leakage identification can be performed in parallel and share data. Figure 8 The quality control results and positioning benchmarks established as shown in (6-1) to (6-3) are illustrated.

[0171] (6-6) Results Summary and Output

[0172] The disease identification module 6 summarizes the identification results of cracks and / or leaks according to their location coordinates, generates a disease list, statistical results and distribution representation, and outputs traceable identification results. The output content includes at least the disease type, disease location, disease parameters, evidence images and traceability information.

[0173] The location of the defect includes at least the axial mileage Z and the circumferential position S or the circumferential angle θ; the defect parameters include at least the crack length, width and direction, as well as the leakage area, range and degree indicators; the evidence images include at least a partial screenshot of the unfolded image and an annotated overlay image; the traceability information includes at least the image block number, the start and end mileage of the block and the time range, the event index or equivalent index, the row number or trigger count, the pose solution and the quality mark, so as to support the engineering review, archiving and maintenance decision-making.

[0174] In the optional configuration of the embodiment, the disease identification module 6 can output candidate points and verification fields for manual review, and generate statistical indicators such as recognition rate and missed detection rate for quantitative verification and continuous optimization.

[0175] Example 2

[0176] like Figure 9As shown, this embodiment provides a method for detecting defects in water conveyance tunnels based on a laser line scan camera array. This method establishes a unified time reference using the PPS time synchronization signal output by the attitude and position measurement unit, and uses the camera trigger signal and / or valid exposure signal as inputs to the event terminal to form an event time stamp. Combined with DMI odometer constraints, it achieves a one-to-one correspondence between the line scan grayscale image data and the pose data at the same reference time. Based on this, motion compensation, spatial registration, two-dimensional unfolding and stitching, and defect identification and location output are completed. The method steps are as follows.

[0177] S0, Power-on self-test and parameter initialization

[0178] After power-on, the system completes self-tests for each unit and enters the detection process. Based on the minimum identifiable scale of the defect to be inspected, a maximum axial spatial sampling interval threshold is preset. With the threshold for allowing single-line spatial image shift And complete the initial configuration of the camera array trigger line frequency and single-line integration time.

[0179] S1. Control the mobile detection platform to move continuously along the axial direction of the water conveyance tunnel at a preset speed.

[0180] The mobile detection platform is controlled to move continuously along the axial direction of the water conveyance tunnel at a preset speed. The synchronous control unit operates according to... and The image line frequency and the integration time of the line scan camera are set or adjusted in a coordinated manner to ensure that the image displacement corresponding to a single line scan does not exceed a preset threshold and the axial sampling interval meets the preset threshold requirement under continuous movement conditions, thereby ensuring the distinguishability of disease features in the two-dimensional line scan grayscale image. The constraint relationship and setting basis of line frequency and integration time have been described in Example 1, and will not be repeated in this step.

[0181] S2. Synchronously acquire line scan grayscale image data and record event timestamps.

[0182] The attitude and position measurement unit outputs a PPS time synchronization signal to establish a unified time reference for the system. The synchronization control unit generates camera array trigger signals and performs fan-out allocation based on this unified time reference, enabling each laser line-scan camera to sample synchronously according to a preset trigger line frequency. Each laser line-scan camera acquires the reflected grayscale information of the tunnel interior wall line by line under near-infrared linear active illumination, forming line-scan grayscale image data. The synchronization control unit connects the camera trigger signal and / or the valid exposure signal to the event input terminal of the attitude and position measurement unit, which then timestamps the event input edges, forming an event time-stamp sequence corresponding to the image line data. The image line data and event timestamps are synchronously saved and transmitted to the data acquisition and processing unit.

[0183] S3. Acquire pose and distance information, perform time matching based on event timestamps, and complete motion compensation and spatial registration.

[0184] The attitude and position measurement unit continuously outputs the pose information of the moving detection platform under a unified time reference, and simultaneously outputs DMI (Distance Mitigation Index) odometry information as an axial distance constraint. The data acquisition and processing unit reads the event time-stamp sequence and matches the pose output using the event time-stamp as an index. Event-triggered matching is used for directly corresponding event time-stamps, and time interpolation is used for image row data between adjacent time-stamps to obtain pose data that corresponds one-to-one with each image row data. The data acquisition and processing unit further introduces DMI odometry constraints to perform uniformity correction and resampling on the axial sampling position, reducing the non-uniformity of axial sampling caused by velocity fluctuations. After completing the binding of image row data and pose data, motion compensation and spatial registration are performed on the image row data, and circumferential corresponding offset is corrected based on pose changes, so that the line scan data from multiple cameras and at multiple times are aligned under a unified unfolded reference frame.

[0185] S4. Stitch together to generate a two-dimensional unfolded image of the tunnel inner wall.

[0186] The registered line scanning data obtained in step S3 is written into a two-dimensional unfolded coordinate system according to axial mileage and circumferential position to generate two-dimensional unfolded grayscale image information of the tunnel inner wall. The overlapping areas of the multi-camera fields of view are fused to achieve smooth grayscale transition, and the unfolded image is resampled to obtain a two-dimensional unfolded image with uniform spatial resolution, providing a unified input for subsequent defect identification.

[0187] S5. Identify cracks and leaks based on 2D unfolded image information and output traceable results.

[0188] Disease identification module Figure 8 The process described receives the 2D unfolded grayscale image or 2D unfolded image blocks divided by mileage output from step S4, and receives the index field bound to the image. The index field includes at least the image block number, the start and end mileage and time range of the block, the event time-stamp sequence and its corresponding pose solution, the DMI quality flag, and the row number or trigger count that can be traced back to the original row data. After performing quality control and necessary preprocessing on the input image, the defect identification module identifies cracks and / or leakage defects and measures their parameters. It then maps the identification results to a 2D unfolded coordinate system and outputs the defect location represented by axial mileage and circumferential position. The identification output includes at least the defect type, defect location, geometric parameters, and result index information, forming a traceable defect list and distribution map.

[0189] Output the disease detection results and location information, and the detection process ends.

[0190] In this embodiment, to quantitatively verify the effectiveness of crack identification, several tunnel sections containing cracks were selected as test samples, and a true list of cracks was compiled through manual review. The crack identification results output by the defect identification module were compared item by item with the true list. The crack identification rate was calculated as the proportion of the number of cracks that matched the true list to the total number of true cracks, and the false negative rate was calculated as the proportion of the number of cracks that did not match the true list to the total number of true cracks. The statistical results are as follows: 162 cracks were manually recorded; the algorithm identified 136, and 26 were not identified, resulting in a crack identification rate of 83.95% and a false negative rate of 16.05%. In addition, the algorithm output 71 additional suspected crack candidate points beyond the manually recorded points for subsequent manual review and engineering treatment decision-making.

[0191] In this invention, the device and method involve a mobile detection platform moving continuously and in a controlled manner along the axial direction of a water conveyance tunnel. An array of laser line-scanning cameras performs line-by-line grayscale imaging of the tunnel's inner wall under near-infrared active illumination, generating line-scanning grayscale image data that accumulates continuously with the axial mileage. Multiple laser line-scanning cameras are arranged in a ring or segmented ring array along the circumferential direction of the tunnel cross-section. The fields of view of each camera are interconnected in the circumferential direction and have necessary overlap areas, thereby achieving continuous coverage acquisition of the entire or near-entire cross-section of the tunnel's inner wall.

[0192] To achieve a one-to-one correspondence between image line data and pose data at the same reference time, the synchronization control unit establishes a unified system time reference using the PPS time synchronization signal output by the attitude and position measurement unit. It also connects the camera trigger signal and / or valid exposure signal to the event input to form an event time-stamped sequence, while using DMI odometer information to provide axial distance constraints. The data acquisition and processing unit uses the event time-stamped sequence as an index to perform time alignment and interpolation matching on the pose output, and combines the DMI odometer constraints to perform axial sampling consistency correction and resampling, thus completing the binding of image line data and pose data.

[0193] In the 2D unfolded image generation stage, the data acquisition and processing unit performs motion compensation and spatial registration on the line scan data after binding, corrects the circumferential offset caused by attitude changes, and writes the line scan strips from multiple cameras and at multiple times into a unified unfolded coordinate system according to axial mileage and circumferential position, forming a 2D unfolded grayscale image of the tunnel inner wall. The overlapping areas of the multi-camera fields of view are fused, and the unfolded image is resampled to obtain an output with consistent spatial resolution. The defect identification module... Figure 8 The process shown receives a two-dimensional unfolded grayscale image or a two-dimensional unfolded image block divided by mileage and its index field, identifies cracks and / or leakage defects, measures parameters, locates and outputs the results, and forms traceable results.

[0194] Under the device configuration and data acquisition process described in Examples 1 and 2, the present invention can achieve the following technical effects and verification conclusions.

[0195] 1. Imaging resolvability is guaranteed under high-speed continuous movement conditions.

[0196] The synchronous control unit determines the permissible single-line spatial image shift threshold required for crack and leakage identification. With the maximum axial sampling interval threshold By coordinating and adjusting the image row frequency, single-row integration time, and moving speed, the axial sampling interval and single-row spatial image shift are simultaneously controlled under continuous movement conditions, thereby ensuring the distinguishability of disease features in the two-dimensional line-scanned grayscale image. Under the test conditions of this embodiment, when the platform moves continuously at 1 m / s to 5 m / s, the two-dimensional unfolded grayscale image still maintains clear and distinguishable crack and leakage features, and can stably detect cracks with a width of not less than 1 mm. Under the constraint of the same detection mileage and near-full cross-section coverage requirements, the unit mileage operation time of continuous moving line-scan detection and stop-and-go detection are compared, and the detection efficiency can be improved by about 5 to 10 times.

[0197] 2. Arrayed circumferential arrangement improves coverage and reduces blind spots.

[0198] Multiple laser line scanning cameras are arranged in a circumferential array along the tunnel cross-section and stitched together circumferentially, enabling continuous scanning of the entire or near-entire cross-section of the tunnel wall. Under near-entire cross-section conditions, this embodiment can achieve a coverage rate of no less than 95%, effectively reducing obstruction and blind spots, and improving the completeness and consistency of the defect results.

[0199] 3. Linear laser active illumination improves grayscale uniformity and signal-to-noise ratio.

[0200] Each camera integrates a linear laser active illumination component to form a strip illumination that matches the imaging field of view, providing stable illumination conditions in low-light or unevenly illuminated environments inside the tunnel, reducing the impact of ambient light changes on line-scan grayscale imaging, improving the grayscale consistency and disease feature contrast of the unfolded image, and providing stable input for crack and leakage identification.

[0201] 4. PPS time base, event time scale, and DMI odometry constraints improve the reliability of image pose alignment.

[0202] A precise time correspondence between image acquisition and pose measurement is established based on the PPS unified time base and event time-stamped records. DMI mileage constraints are used to improve axial distance consistency and pose calculation continuity. Under the test conditions of this embodiment, the time alignment accuracy can reach the microsecond level, and the registration error between image and pose data can be controlled at the millimeter level, providing a reliable foundation for 2D unfolded image generation and lesion identification.

[0203] 5. The disease identification results on the two-dimensional unfolded image have engineering applicability.

[0204] In this embodiment, to quantitatively verify the crack identification effect, several tunnel sections containing crack defects were selected as test samples, and a true list of crack defects was generated through manual review. The crack identification results output by the defect identification module were compared item by item with the true list. The crack identification rate was calculated as the proportion of the number of cracks that matched the true value out of the total number of true value cracks, and the false negative rate was calculated as the proportion of the number of cracks that did not match the true value out of the total number of true value cracks. The statistical results are as follows: 162 crack defects were manually recorded, 136 were automatically identified, and 26 were not identified, resulting in a crack identification rate of 83.95% and a false negative rate of 16.05%. In addition, the system output 71 additional suspected crack candidate points beyond the manually recorded points for subsequent manual review and engineering treatment decision-making reference.

[0205] The device and method of this invention have been applied to the detection of defects in multiple water conveyance tunnels. Compared with traditional manual inspection and common instrument detection methods, it has advantages such as fast detection speed, stable imaging quality, large coverage, good result consistency and high degree of intelligence, and can provide technical support for the operation and maintenance of water conveyance tunnels.

[0206] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Modifications, equivalent substitutions, or improvements made by those skilled in the art without departing from the spirit and principles of the present invention should all be included within the scope of protection of the present invention.

Claims

1. A device for detecting defects in water conveyance tunnels based on a laser line scan camera array, characterized in that, include: A mobile inspection platform is used to carry a laser line scanning camera array and move continuously along the axial direction of the water conveyance tunnel. A laser line scan camera array, installed on the mobile detection platform, includes multiple laser line scan cameras equipped with near-infrared linear laser active illumination components. The multiple laser line scan cameras are arranged in an array along the circumference of the tunnel cross section. They are used to project linear near-infrared lasers onto the inner wall of the tunnel during movement to form illumination strips and simultaneously acquire line scan grayscale image data at the corresponding positions. The attitude and position measurement unit is used to acquire the displacement and attitude information of the mobile detection platform during the movement process, and to provide PPS time synchronization signal and event input terminal; The synchronization control unit is used to receive the PPS time synchronization signal to establish a unified time reference, and to coordinate the trigger line frequency and single-line exposure integration time of the laser line scanning camera array, and to perform unified timing control of the laser line scanning camera array and the mobile detection platform to ensure the time synchronization of each collected data during the movement, and to control the mobile detection platform to move at a constant speed along the central axis of the water conveyance tunnel through the signal to maintain a stable speed and relative position. The data acquisition and processing unit is used to simultaneously acquire the line scan grayscale image data of the laser line scan camera array and the displacement and attitude information obtained by the attitude and position measurement unit, and to perform image correction, stitching and spatial registration processing on the line scan grayscale image data to generate two-dimensional unfolded image information of the tunnel inner wall. The defect identification module is used to identify and analyze cracks and / or leakage defects in the tunnel inner wall based on the two-dimensional unfolded image information.

2. The apparatus according to claim 1, characterized in that, The synchronization control unit is used to pre-set a maximum axial spatial sampling interval threshold based on the minimum identifiable scale of cracks and / or leakage defects. With the threshold for allowing single-line spatial image shift And during the continuous movement of the mobile detection platform, based on the platform's instantaneous speed The trigger line frequency of the laser line scan camera array With single-line exposure time To achieve coordinated control, the following can be achieved: ， ； This simultaneously satisfies the imaging conditions of sufficient axial sampling and controlled single-row image shift, ensuring effective identification of cracks and / or leakage defects under continuous movement conditions.

3. The apparatus according to claim 2, characterized in that, The threshold for allowing single-line spatial image shift The length should not exceed 1 mm, or be no greater than the object length corresponding to a single pixel or a preset multiple thereof, to ensure the distinguishability of fine cracks and seepage textures in two-dimensional grayscale images.

4. The apparatus according to claim 1, characterized in that, The laser line scanning camera array is arranged in a ring or segmented ring along the tunnel cross-section to achieve full or near full cross-section coverage scanning of the tunnel inner wall and reduce detection blind spots caused by the tunnel's curved surface structure or ancillary structures.

5. The apparatus according to claim 1, characterized in that, The attitude and position measurement unit includes a POS / IMU and a DMI odometry interface, and provides a PPS time synchronization signal output terminal and an event input terminal; The synchronization control unit includes a time base module, a trigger generation and allocation module, and an event marking module; The time reference module is used to receive the PPS time synchronization signal to establish a unified time reference; The trigger generation and allocation module is used to generate and allocate camera array trigger signals based on the unified time reference. The event marking module is used to input the camera array trigger signal, exposure valid signal or line valid signal into the event input terminal of the attitude and position measurement unit to record the event timestamp and establish the time correspondence between the image acquisition reference time and the displacement information and attitude information.

6. The apparatus according to claim 3, characterized in that, The trigger generation and allocation module is implemented using programmable logic devices. Under the unified time reference constraint, it generates low-jitter trigger pulses and outputs them in parallel to the trigger input terminals of each line scan camera through a fan-out, buffer, and impedance-matched allocation network. This ensures that the trigger timing difference between cameras is controlled within the microsecond range, and the single-line exposure integration time is also controlled. satisfy ,in The set trigger line frequency for the laser line scanner camera, and Not less than 1 kHz.

7. The apparatus according to claim 5, characterized in that, The trigger generation and allocation module includes at least the following two trigger modes: Time-triggered mode, triggered by the set laser line scan camera's trigger frequency. Output trigger; In distance-triggered mode, the distance increment or pulse output from the DMI odometry interface of the attitude and position measurement unit is triggered according to a preset axial distance step. Generate a trigger signal to ensure that the axial spatial sampling interval of adjacent scan lines meets the following conditions. ; When configuring dual DMIs, it is preferable to designate one of them as the distance trigger reference and the other as the redundancy consistency check and slippage identification. When the output difference of the dual DMIs exceeds the preset threshold and outputs an abnormal quality flag, the synchronization control unit switches to the time trigger mode and / or adaptively adjusts the trigger parameters to improve trigger stability and image scale consistency under complex working conditions.

8. The apparatus according to claim 1, characterized in that, The data acquisition and processing unit includes an image correction submodule, an image stitching submodule, and a spatial registration submodule, wherein: The image correction submodule is used to perform uniformity correction processing on the line scan grayscale data output line by line by line from each laser line scan camera, including at least dark current and gain uniformity and brightness equalization. The image stitching submodule is used to register and stitch the imaging strips of multiple cameras in the circumferential direction of the tunnel cross section according to the camera array installation calibration parameters, and to perform stitching consistency verification and stitching parameter constraints in the overlapping areas of adjacent fields of view in order to suppress stitching misalignment and scale drift. The spatial registration submodule is used to bind line scan data with displacement information, attitude information, and DMI mileage one-to-one under the unified time reference and event time stamp recording framework established by the synchronization control unit. The binding result is used as a constraint to perform motion compensation and spatial registration on the stitching result, generating two-dimensional unfolded image information with axial mileage as row coordinates and circumferential unfolded coordinates as column coordinates. The two-dimensional unfolded image information is a two-dimensional unfolded grayscale image or image block, and the synchronization index field bound to the two-dimensional unfolded image information is output.

9. The apparatus according to claim 1, characterized in that, The near-infrared linear laser active illumination component includes a near-infrared linear laser source and a cylindrical lens. The cylindrical lens is used to spread the laser beam to form a linear illumination strip. A bandpass filter matching the emission band of the laser source is placed in front of the camera lens to suppress the influence of ambient stray light on linear grayscale imaging.

10. A method for detecting defects in water conveyance tunnels based on a laser line scan camera array, using the apparatus described in any one of claims 1-9, characterized in that... The method includes the following steps: S1. Control the mobile detection platform to move continuously along the axis of the water conveyance tunnel at a preset speed; S2. During the movement, the laser line scan camera array synchronously acquires line scan grayscale image data of the tunnel inner wall under near-infrared laser active illumination, and records the event timestamp. S3. Obtain the displacement and attitude information of the mobile detection platform, perform time matching and binding of the line scan grayscale image data with the displacement and attitude information according to the event time stamp, and complete motion compensation and spatial registration. S4. The line scan grayscale image data after motion compensation and spatial registration is stitched together to generate two-dimensional unfolded image information of the tunnel inner wall. S5. Based on the two-dimensional unfolded image information, perform automatic identification and analysis of cracks and / or leakage defects.

Images