Bridge crack detection system and equipment thereof

Abstract

The invention provides a bridge crack detection system and equipment thereof, and relates to the field of bridge crack detection robots. The system comprises a robot body, a negative pressure adsorption system, a wheel leg type driving device, a crack detection sensor module, an image acquisition system and a power supply and control system. The robot body adopts a lightweight design, the three groups of negative pressure adsorption devices form a closed cavity to realize stable attachment, and the wheel-leg type driving device has driving and obstacle avoidance capabilities and adapts to complex bridge surfaces. The crack detection module realizes accurate crack detection and parameter extraction through a matrix type visual camera array, a main camera and a laser ranging module in combination with a CNN algorithm. The equipment is controlled by an upper computer, transmits data through a composite cable, and supports path planning, online learning and structured report generation. The equipment replaces manual high-altitude operation, the detection automation and intelligence level is improved, the problems that traditional detection is low in efficiency, high in risk, insufficient in precision and the like are solved, and detection safety and reliability are guaranteed.

Description

Technical Field

[0001] This invention relates to the field of bridge crack detection robots, specifically to a bridge crack detection system and equipment. Background Technology

[0002] As bridges age, structural safety issues become increasingly prominent. Cracks, a common form of bridge damage, can seriously affect a bridge's safety and lifespan if not detected and repaired in a timely manner. Traditional bridge crack detection typically relies on manual inspections, which is not only time-consuming and labor-intensive but also limited by the complexity of the working environment and the dangers of working at heights, making it difficult to guarantee detection efficiency and safety.

[0003] Currently, while existing bridge-climbing robots can perform surface inspections on bridges, they generally suffer from insufficient adhesion stability, poor mobility, and low inspection accuracy, making it difficult to meet the demands for refined and efficient inspections. Therefore, there is an urgent need for a more stable, safe, and efficient bridge-climbing robot to improve the accuracy and efficiency of bridge crack detection.

[0004] Therefore, this invention provides a bridge crack detection system and equipment, including a wall-climbing robot body, a negative pressure adsorption system, a wheel-leg drive device, a crack detection sensor module, an image acquisition system, and a power supply and control system. The robot body features a lightweight design and is equipped with three sets of high-efficiency negative pressure adsorption devices. A fan forms a closed adsorption cavity, enabling stable attachment to the vertical or inverted surfaces of the bridge. The wheel-leg drive mechanism integrates drive wheels and deformable support wheels, providing both driving and obstacle avoidance capabilities. It can adapt to bridge surfaces with different inclination angles and complex textures, enabling the robot to move smoothly in steep slopes and obstacle environments. The entire system is controlled and data processed by a host computer platform, and high-bandwidth images and sensor signals are transmitted via cable, ensuring stability during long-distance operation. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a bridge crack detection system and equipment, which solves the problems of low efficiency, high operation risk, strong data subjectivity and insufficient accuracy of traditional manual bridge crack detection methods. It addresses the challenges of high-altitude operation on bridge structures and high safety protection requirements, replaces manual operation in high-risk areas, and improves the automation and intelligence level of bridge inspection.

[0006] The technical solution adopted in this invention is to provide a bridge crack detection system and equipment, including a robot body, a negative pressure adsorption system, a wheel-leg drive device, a crack detection sensor module, an image acquisition system, and a power supply and control system. The robot body adopts a lightweight design and is equipped with three sets of high-efficiency negative pressure adsorption devices. A fan forms a closed adsorption cavity, achieving stable attachment to the vertical or inverted surface of the bridge. A matrix-type vision inspection system and a main camera are installed at the front of the robot, enabling multi-angle and multi-dimensional crack feature extraction. Combined with high-definition images and intelligent recognition algorithms, it obtains key parameters such as crack geometry, length, width, and depth. The system supports 220V AC power supply, has a standard 5kW regulated power supply, and is equipped with wireless image projection functionality to improve the collaborative efficiency of on-site inspection.

[0007] In a preferred embodiment, this invention provides a negative pressure adsorption system. Three sets of negative pressure adsorption chambers are arranged at the bottom of the robot, forming a stable three-point adsorption support structure. Each set of suction cups includes a suction fan and an adsorption blower. The suction blower provides adsorption force to ensure reliable adhesion of the robot to surfaces of different materials and angles, such as vertical concrete walls, the inner walls of steel box beams, and overhead base plates. The adsorption status is monitored in real time by an embedded pressure sensor and a negative pressure safety controller. When the adsorption force decreases, the system will issue an alarm and stop movement to prevent the robot from falling.

[0008] In a preferred embodiment, the present invention also provides a crack detection system. The crack detection unit consists of a matrix visual camera array, a main camera, and an auxiliary laser ranging module, and is mounted on a shock-resistant bracket with an adjustable angle at the front end of a robot. The main camera is a 12-megapixel CMOS image sensor that can cover the entire width of the pier surface. The matrix camera array is distributed horizontally and uses a multi-view stitching algorithm to reconstruct the apparent crack information. The laser ranging module is used to correct the image size and, in conjunction with visual depth measurement, assists in estimating the crack depth. Image processing employs edge extraction and a CNN algorithm framework, supporting crack level classification, automatic numbering, and real-time display of detection results on a host computer terminal. Attached Figure Description

[0009] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a front view structural diagram of the wall-climbing robot of the detection equipment of this invention; Figure 2 This is a diagram of the bottom structure of the wall-climbing robot used in the detection equipment of this invention; Figure 3 This is a flowchart of the operation of the wall-climbing robot system of the detection equipment of this invention; Figure 4 This is a flowchart of the crack detection process of the wall-climbing robot system of the present invention. Detailed Implementation

[0010] To better understand the purpose, system architecture, and functional implementation of this embodiment, the embodiments and features described herein can be combined with each other without conflict. The exemplary embodiments disclosed herein will be described below with reference to the accompanying drawings, including specific technical details disclosed to aid understanding; however, these details should be considered exemplary rather than restrictive. Therefore, those skilled in the art should understand that various improvements and adjustments can be made to the embodiments described herein without departing from the scope and core ideas of the invention. Similarly, for clarity, detailed descriptions of well-known technologies, functions, and structures are omitted in the following description.

[0011] Example 1 Figure 1 This is a front view structural diagram of the wall-climbing robot detection device of the present invention.

[0012] Figure 2 This is a diagram of the bottom structure of the wall-climbing robot used in the detection equipment of this invention.

[0013] like Figure 1 and Figure 2 As shown, the wall-climbing robot includes a defect detection sensor 1, a mechanical rear device 2, a robot body shell 3, a suction fan 4, a cable 5, wheel legs 6, a drive wheel 7, a main camera 8, a matrix camera module 9, an adsorption fan 10, and an integrated suction cup 11. The modules are closely integrated and work together to achieve stable attachment, path movement, and automatic crack detection on the surface of complex bridge structures.

[0014] The wall-climbing robot features a box-like structure, with the main shell constructed from aluminum alloy and carbon fiber composite materials, balancing strength, rigidity, and lightweight. Measuring approximately 810mm × 475mm × 365mm, the entire robot weighs under 18kg, facilitating single-person transport and deployment. The shell surface is lined with soft rubber-coated cushioning pads and impact-resistant plates, effectively reducing damage or adhesion failure caused by collisions with the wall. The internal structure employs a layered layout: the upper layer houses the power supply and control unit, the middle layer contains the image and sensing system, and the lower layer houses the drive and adhesion modules. Modules are connected via quick-connect interfaces for easy on-site maintenance and replacement.

[0015] The wall-climbing robot has three sets of negative pressure adsorption chambers at its bottom, forming a stable three-point adsorption support structure. Each set of suction cups consists of an exhaust fan and an adsorption fan. The exhaust fan is powered by 24V DC, with a single fan power of approximately 2400W, a negative pressure value of -70kPa, and a unit adsorption force of ≥350N, ensuring reliable adhesion and adsorption of the robot on various materials and angles, such as vertical concrete walls, the inner walls of steel box beams, and overhead base plates. Real-time monitoring is achieved through embedded pressure sensors and a negative pressure safety controller. When the robot's bottom edge lifts or the surface peels off, causing a decrease in adsorption force, the system will issue an alarm and stop movement to prevent the robot from falling.

[0016] The wall-climbing robot's driving system employs a four-wheel independent drive and auxiliary wheel leg design. The four main drive wheels are symmetrically positioned at the front and rear of the robot, encased in rubber, and house a built-in 300W DC motor, providing independent speed and direction adjustment capabilities. Combined with the servo-controlled auxiliary wheel leg structure, it can achieve a certain range of vertical posture adjustment and lateral obstacle crossing. The wall-climbing robot connects to the ground power supply and control box via a composite cable. The standard cable length is 100m, expandable to 400m. The cable includes a 24V power line, video signal line, CAN control signal line, and gigabit network cable, and is designed to be anti-torsion, anti-wear, and water-resistant. The power supply uses an industrial regulated power supply, supporting 220V AC / 380V AC input and 24V / 5kW DC output, with built-in overcurrent, overvoltage, short circuit, and leakage protection functions. During robot operation, data such as images, adsorption status, posture, wheel speed, and voltage are uploaded to the ground control terminal in real time via gigabit wired communication, with system communication latency controlled within 50ms.

[0017] Figure 3 This is a flowchart of the operation of the wall-climbing robot system of the detection equipment of this invention.

[0018] like Figure 3 As shown, the wall-climbing robot achieves bridge crack detection through deep coupling between its main body structure, negative pressure adsorption, wheel and leg drive, crack detection, power supply, communication and control system. After the operator lifts the robot to the starting position of the bridge detection, the mechanical auxiliary support needs to be activated. The support needs to be connected to the M8 threaded interface reserved in the robot body to ensure that the angle between the working surface of the main body and the bridge surface is less than 5° before the adsorption system is activated. At this time, the negative pressure fan is in standby mode, the IMU starts to record the initial attitude angle, and the power supply box outputs 24V DC standby voltage.

[0019] The adsorption system startup and safety verification mainly involve three levels of safety interlocks. These include: a soft-start phase where the fan PWM duty cycle gradually increases from 0% to 30%, eventually reaching 60% to prevent instantaneous high-current surges; a dynamic sealing phase where the embedded spring steel frame deforms under pressure, forming a seal with the micro-undulations of the rough concrete surface; a differential pressure sensor with a sampling frequency of 100Hz monitoring the pressure difference inside and outside the cavity; and finally, calculation and verification of the adsorption force, with the system calculating the effective adsorption force in real time. As shown in equation (1) below.

[0020] (1) in, Atmospheric pressure, Negative pressure inside the cavity. To effectively adsorb the mask, Seal leakage loss force, To determine the theoretical maximum adsorption force, the net adsorption force generated by the robot's three sets of negative pressure adsorption units was calculated to ensure the robot's safe and stable attachment to the bridge surface. Furthermore, to guarantee effective adsorption even under extreme conditions, a dynamic safety factor was incorporated into the design. As shown in equation (2) below.

[0021] (2) in, Design values ​​for practical applications, dynamic safety factor It is calculated from the long-term operational reliability margin coefficient and the redundancy efficiency coefficient, as shown in equation (3) below.

[0022] (3) in, The long-term operational reliability margin factor, taking into account factors such as wind turbine performance degradation, voltage fluctuations, surface contamination, maintenance delays, and temperature effects, is set at 1.67. The redundancy efficiency coefficient is rounded to 0.70 considering incomplete failure scenarios. Therefore, the dynamic safety factor of this invention is... The calculation yields a value of 2.39 ≈ 2.4. By combining the failure redundancy capability and the long-term operational reliability margin into a composite safety factor, the cumulative failure probability of the robot under extreme conditions is ensured to be low throughout its entire service life.

[0023] Once the adsorption force is confirmed to be operating normally, the host computer generates the A* optimal path based on the preset 3D model of the bridge. For expansion joints wider than 5mm, the system automatically switches to leg-type obstacle-crossing mode, with the front drive wheels decelerating and the legs vertically raised to overcome the obstacle. When the robot's moving speed reaches 0.1m / s, the matrix camera array begins synchronous exposure in soft-trigger mode, with the main camera maintaining 30FPS to record a high-definition video stream. The laser ranging module continuously outputs the distance between the robot and the wall for visual dimension calibration. Manual annotation is performed by transmitting visual stream, laser stream, and state stream data. The host computer software provides annotation tools. For cases where the CNN algorithm misses or falsely detects, the operator selects an area and inputs "true / false" crack labels. The labels are fed back to the robot system, triggering online learning and updating the model weights. After the task is completed, the host computer sends a return command, and the robot executes reverse path tracking, returning to the starting point using the wheel speed integral trajectory recorded by the encoder. If a new obstacle is encountered during the return journey, local replanning is initiated. During the retrieval process, the negative pressure fan is first slowly shut down, then the main controller is shut down, and finally the power is disconnected to complete the robot retrieval. After the recovery is completed, the host computer software automatically generates a structured report, including metadata, crack data table, statistical charts and raw data packets. The report is then uploaded to the detection cloud platform and the local database.

[0024] Figure 4 This is a flowchart of the crack detection process of the wall-climbing robot system of the present invention.

[0025] like Figure 4 As shown, the main camera for image data acquisition is a 12-megapixel CMOS image sensor equipped with a 179° wide-angle lens, capable of covering the entire width of the web / pier surface. The matrix camera array consists of 3-5 small wide-angle cameras distributed horizontally. A multi-view stitching algorithm is used to reconstruct the surface crack information. The robot performs image acquisition at 150ms intervals to avoid repeated acquisition and data redundancy. The laser ranging module includes a single-line laser displacement sensor that retains effective points within a 500±10mm range between the robot and the wall surface, filtering out abnormal jump points caused by surface corrosion, thereby assisting in the estimation of crack depth in conjunction with visual depth measurement.

[0026] In the image data preprocessing stage, a multi-scale Retinex algorithm is employed to enhance image contrast in low-light areas of the bridge, widening the grayscale difference between cracks and the background, making crack images more clearly visible. Laser data preprocessing addresses speckle noise caused by laser scattering due to pores on the concrete surface by fusing multiple measurements using Kalman filtering to improve the signal-to-noise ratio.

[0027] After image preprocessing, the crack image is segmented by CNN algorithm. An attention gating module is added to the original algorithm framework to focus on crack edge details. The loss function in the CNN algorithm is shown in equation (4).

[0028] (4) in, Predict the probability of the crack category output by the model. The crack is a true label. The intersection of the prediction and the ground truth represents the predicted probability of the real crack pixel accumulation model. To approximate the union of the predicted and true values, when the prediction of crack pixels is insufficient, the denominator... It won't be small for no reason. To balance the weighting coefficients across categories, positive samples are given higher weights for crack pixels. To focus on dynamically adjusting the weights of easy and difficult samples, the model can concentrate on learning difficult-to-classify samples, effectively alleviating the class imbalance problem when the proportion of crack pixels is less than 5%.

[0029] Laser measurement is calculated using the principle of laser triangulation, as shown in equation (5) below.

[0030] (5) in, For depth, The angle between the laser emitter and the receiver. Focal length For pixel displacement, Using the baseline distance, the same crack location is scanned three times, and the median is taken as the final value to eliminate random errors in a single measurement. The laser parameter width is calculated by directly measuring the distance between the laser point clouds on both sides of the crack, while the depth is calculated by fitting the cross-sectional profile of the crack. The maximum indentation depth is the crack depth.

[0031] By employing a bimodal confidence assessment fusion weight allocation, visual confidence is evaluated. Image sharpness is evaluated based on the Tenengrad gradient function, as shown in equation (6).

[0032] (6) in, and pixel coordinates Axis and pixel coordinates The number of pixels in the image along the axis. and This is the Sobel gradient.

[0033] Laser confidence The point cloud quality is evaluated based on the signal-to-noise ratio, as shown in equation (7).

[0034] (7) in, The laser ranging value represents the actual crack profile. The laser ranging value is the value measured on the surface of the bridge. The noise amplitude is represented by the multimodal fusion weighting algorithm shown in equation (8).

[0035] (8) in, and The weights for visual and laser fusion are shown in equation (9) below.

[0036] (9) The dynamic weight adjustment is determined based on factors such as lighting, shadows, roughness, and smoothness in the environment. When the lighting is sufficient and the surface is rough, the weight is increased. ,reduce When the environment is a shaded area but the surface is reflective, increase ,reduce When both the image and laser fail, the system triggers an early warning to pause detection, and the robot adjusts its pose or performs supplemental lighting. The final output is a JSON structured data packet. The original image is archived named by crack ID, and the laser point cloud is stored in .las format for BIM model inversion.

[0037] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0038] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A bridge crack detection system, characterized in that, Includes a negative pressure adsorption system, a wheel-leg drive device, a crack detection sensor module, and an image acquisition system; The negative pressure adsorption system, located at the bottom of the robot, includes three sets of negative pressure adsorption chambers and a suction monitoring unit; The wheel-leg drive system includes four main drive wheels and an auxiliary wheel-leg structure; The crack detection sensor module is mounted on an adjustable-angle shockproof bracket at the front of the robot. An image acquisition system, a matrix-style visual inspection system, and a main camera are used to extract crack features.

2. The bridge crack detection system according to claim 1, characterized in that, The negative pressure adsorption system includes three sets of negative pressure adsorption chambers, forming a stable three-point adsorption support structure. Each adsorption chamber includes a suction fan and an adsorption blower; The suction monitoring unit includes an embedded pressure sensor and a negative pressure safety controller, which are used to monitor the adsorption status in real time, issue an alarm and stop movement when the adsorption force decreases.

3. The bridge crack detection system according to claim 1, characterized in that, The crack detection sensor module includes a matrix vision camera array, a main camera, and an auxiliary laser ranging module, and is mounted on a shockproof bracket. The main camera is a CMOS image sensor that covers the entire width of the pier surface; The laser ranging module is used to correct the image size and, in conjunction with visual depth measurement, assists in estimating the crack depth.

4. The bridge crack detection system according to claim 1, characterized in that, The matrix-type visual camera array uses a multi-view stitching algorithm to restore the apparent crack information.

5. The bridge crack detection system according to claim 1, characterized in that, The image acquisition system adopts edge extraction and CNN algorithm framework, supports crack level classification, automatic numbering and real-time display of detection results on the host computer terminal. The loss function in the CNN algorithm includes an attention gating module to focus on crack edge details.

6. The bridge crack detection system according to claim 1, characterized in that, The wheel-leg drive device includes four main drive wheels and auxiliary wheel leg structures. The main drive wheels are distributed symmetrically at the front and rear of the robot, are covered with rubber material, and have built-in DC motors.

7. The bridge crack detection system according to claim 6, characterized in that, The auxiliary wheel legs of the wheel-leg drive device are connected to the servo control system, enabling vertical posture adjustment and lateral obstacle crossing. When an obstacle with a width exceeding a preset threshold is detected, the front drive wheel decelerates and the wheel legs are raised vertically to cross the obstacle.

8. The bridge crack detection system according to claim 1, characterized in that, The system employs a three-level safety interlock mechanism during the start-up and safety verification process of the adsorption system: During the soft start phase, the PWM duty cycle of the fan increases in stages. During the dynamic sealing stage, the embedded spring steel frame is deformed under pressure. The differential pressure sensor monitors the pressure difference inside and outside the cavity, and the system calculates the effective adsorption force in real time to ensure the robot's safe and stable attachment.

9. The bridge crack detection system according to claim 1, characterized in that, The system employs a dual-modal confidence assessment fusion weight allocation, including visual confidence and laser confidence. The weights are dynamically adjusted based on factors such as lighting, shadows, roughness, and smoothness in the environment. When both the image and the laser fail, the system triggers an early warning to pause detection, and the robot adjusts its pose or performs supplementary lighting.

10. A bridge crack detection device, characterized in that, It includes a box-type robot body shell, inside which power supply and control unit, image and sensing system, drive and adsorption module are arranged in layers. The modules are connected by quick-connect interface. The shell surface is equipped with soft rubber-coated cushioning pads and anti-collision plates.

Images