Bridge inspection method and system, computer readable storage medium and program product

Abstract

The invention relates to a bridge inspection method and system and a computer readable storage medium. The inspection method of the bridge comprises the following steps: receiving an inspection task, and driving the inspection device according to the inspection task, so that the inspection device travels along the running guide rail of the bridge; obtaining a sequence image of the bridge surface acquired by the image acquisition assembly; identifying the disease type of the bridge based on the sequence image; and determining geometric feature parameters corresponding to the disease type according to the disease type of the bridge, and determining the severity level of the disease type based on the geometric feature parameters. According to the bridge inspection method, automatic grading evaluation of bridge apparent diseases is facilitated, errors caused by manual judgment are reduced, and the accuracy degree of disease detection is improved.

Description

Technical Field

[0001] This application relates to the technical field of bridge inspection and patrol, and in particular to bridge inspection methods, systems, computer-readable storage media, and program products. Background Technology

[0002] Currently, with the continuous development of transportation systems, bridge construction is becoming increasingly common. The bottom of a bridge beam is a critical load-bearing component, playing a vital role in bridge safety. Being exposed to the elements for extended periods, the bottom of the beam is prone to defects such as cracks, spalling, rust, and water seepage. Therefore, it is necessary to conduct regular inspections of the beam bottom and sides, and address any defects promptly.

[0003] Currently, the inspection of bridge beam bottoms and sides is conducted manually, which presents challenges such as working at heights, high safety risks in windy conditions, and low defect detection rates. Specifically, the inspection of bridge beam bottoms mainly relies on the following methods: First, manual inspection with telescopes, but this method is inefficient and requires subjective judgment of whether defects exist, making it prone to missed detections. Second, using bridge inspection vehicles or suspended platforms to transport personnel for close-range inspection, but this method involves high-altitude work risks, subjective defect inspection, and the possibility of missed or misjudged defects. Summary of the Invention

[0004] Therefore, it is necessary to provide a bridge inspection method, system, computer-readable storage medium, and program product to address the issue of how to improve the automation level of bridge inspection processes in order to enhance the accuracy of defect detection.

[0005] According to a first aspect of this application, a bridge inspection method is provided. The bridge is equipped with a running guide rail, and an inspection device is slidably mounted on the running guide rail. The inspection device is equipped with an image acquisition component. The inspection method includes:

[0006] Receive inspection tasks, drive the inspection device according to the inspection tasks, so that the inspection device travels along the running guide rail of the bridge;

[0007] Acquire a sequence of images of the bridge surface captured by the image acquisition component;

[0008] The type of damage to the bridge was identified based on the sequence of images;

[0009] Based on the type of bridge damage, determine the geometric feature parameters corresponding to the type of damage, and determine the severity level of the type of damage based on the geometric feature parameters.

[0010] In one embodiment, identifying the bridge's damage type based on the sequence of images includes:

[0011] Identify diseased regions in the image sequence; these diseased regions are used to characterize the diseases present in the image sequence.

[0012] Calculate the geometry corresponding to the diseased area;

[0013] The disease type is determined based on the geometry.

[0014] In one embodiment, identifying the bridge's damage type based on the sequence of images further includes:

[0015] Extract the image appearance parameters of the diseased area;

[0016] The geometric shape and the image appearance parameters are input into a preset deep learning model, and the deep learning model outputs at least one disease type.

[0017] In one embodiment, determining the severity level of the disease type based on the geometric feature parameters includes:

[0018] The geometric feature parameters are input into the disease assessment model, and each preset parameter interval corresponds to a severity level.

[0019] Determine the preset parameter range to which the geometric feature parameter corresponding to the disease type belongs, and output the severity level corresponding to the preset parameter range.

[0020] In one embodiment, when the disease type is crack disease, the geometric feature parameter includes at least one of disease length and disease width;

[0021] And / or, when the type of damage is coating damage, the geometric feature parameter includes the area of ​​damage;

[0022] And / or, when the defect type is bolt missing defect, the geometric feature parameters include the defect area.

[0023] In one embodiment, the bridge inspection method further includes:

[0024] Confirm the location of the diseased area.

[0025] In one embodiment, determining the diseased area in the sequence of images includes:

[0026] Confirm the driving parameters required to obtain the diseased area;

[0027] The location of the lesion is confirmed based on the driving parameters.

[0028] In one embodiment, the inspection task includes an inspection mode and historical disease locations, and the inspection mode includes a comprehensive inspection mode and a key inspection mode.

[0029] The step of driving the inspection device according to the inspection task, causing the inspection device to travel along the running guide rail of the bridge, includes:

[0030] When the inspection mode is in the full inspection mode, the inspection device is controlled to travel along the first path; the first path covers all preset detection areas on the bridge.

[0031] When the inspection mode is in the key inspection mode, the inspection device is controlled to travel along the second path, which covers one or more of the historical disease locations.

[0032] In one embodiment, the bridge inspection method further includes:

[0033] Retrieve historical geometric feature parameters stored at the location of the disease in this inspection task from the historical disease database;

[0034] The geometric characteristic parameters of the disease obtained in this inspection are compared with the historical geometric characteristic parameters to calculate at least one quantitative indicator of disease development.

[0035] When the quantitative indicator of disease development is less than the first preset threshold, it can be determined that the disease is developing slowly; when the quantitative indicator of disease development is greater than or equal to the first preset threshold and less than the second preset threshold, it can be determined that the disease is developing stably; when the quantitative indicator of disease development is greater than or equal to the second preset threshold, it can be determined that the disease is developing rapidly.

[0036] Wherein, the first preset threshold is less than the second preset threshold.

[0037] According to a second aspect of this application, a bridge inspection system is provided, comprising:

[0038] The inspection device is mounted on a guide rail on the bridge; the inspection device is slidably mounted on the guide rail.

[0039] An image acquisition component is installed on the inspection device; the image acquisition component is used to acquire a sequence of images of the bridge surface;

[0040] A processor is communicatively connected to the image acquisition component; the processor is used to acquire the sequence of images and identify the type of damage to the bridge based on the sequence of images; the processor is also used to determine geometric feature parameters corresponding to the type of damage to the bridge, and determine the severity level of the type of damage based on the geometric feature parameters.

[0041] The controller is communicatively connected to the inspection device; the controller is used to receive inspection tasks and drive the inspection device to travel along the running guide rail of the bridge according to the inspection tasks.

[0042] In one embodiment, the processor includes a mobile processor and a cloud processor; the mobile processor and the cloud processor are communicatively connected.

[0043] The mobile processor is used to acquire the sequence of images and identify the type of damage to the bridge based on the sequence of images; the cloud processor is also used to determine the geometric feature parameters corresponding to the type of damage to the bridge, and determine the severity level of the type of damage based on the geometric feature parameters.

[0044] In one embodiment, the bridge inspection system further includes an environmental sensor; the environmental sensor is used to acquire environmental data at the location of the inspection device; the environmental sensor is communicatively connected to the controller, and the controller is used to control the inspection device to change its speed based on the environmental data.

[0045] According to a third aspect of this application, a computer-readable storage medium is provided, characterized in that, when the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is able to perform the bridge inspection method described in the above embodiments.

[0046] According to a fourth aspect of this application, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the steps of the bridge inspection method described in the above embodiments.

[0047] The aforementioned bridge inspection methods, systems, computer-readable storage media, and program products identify defective areas in sequential images, extract their geometric features to determine the type of defect, and then use a defect type assessment model to automatically classify and evaluate the apparent defects of the bridge, providing a good reference for bridge repair. Simultaneously, they reduce errors caused by manual judgment and improve the accuracy of defect detection. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the structure of a bridge inspection system shown in one embodiment.

[0049] Figure 2 This is a schematic diagram of the software electrical connections of a bridge inspection system shown in one embodiment.

[0050] Figure 3This is a schematic diagram of the connection structure between the cloud processor and the mobile processor of the bridge inspection system shown in one embodiment.

[0051] Figure 4 This is a flowchart illustrating a bridge inspection method in one embodiment.

[0052] Figure 5 for Figure 4 The diagram shows a detailed process flow for step S300.

[0053] Figure 6 for Figure 5 The diagram shows a detailed process flow for step S310.

[0054] Figure 7 for Figure 5 The diagram shows a detailed process flow diagram of step S330 in one embodiment.

[0055] Figure 8 for Figure 5 A schematic diagram of the specific process of step S330 in another embodiment shown.

[0056] Figure 9 This is a schematic diagram of some steps in a bridge inspection method shown in another embodiment.

[0057] Figure 10 for Figure 9 The diagram shows a detailed process flow for step S600.

[0058] Figure 11 As shown in another embodiment Figure 4 A detailed flowchart of step S100 is shown.

[0059] Figure 12 for Figure 11 The diagram shows a detailed process flow for step S120.

[0060] Figure 13 This is a flowchart illustrating the bridge inspection method shown in other embodiments.

[0061] Explanation of reference numerals in the attached figures:

[0062] 100. Bridge inspection system; 101. Environmental sensor; 1011. Wind sensor; 1012. Obstacle sensor; 110. Inspection device; 111. Vehicle body; 112. Drive motor; 113. Encoding component; 120. Image acquisition component; 121. Camera; 122. Switch; 130. Processor; 131. Mobile processor; 132. Cloud processor; 140. Controller; 150. Communication module; 151. Data processing unit; 152. Domain control unit; 160. Power supply; 161. Power detection sensor; 162. Battery; 163. Charging device; 200. Running rail. Detailed Implementation

[0063] To make the above-mentioned objectives, features, and advantages of this application more apparent and understandable, the specific embodiments of this application are described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of this application. Therefore, this application is not limited to the specific embodiments disclosed below.

[0064] like Figures 1 to 3 As shown, this application provides a bridge inspection system 100, including an inspection device 110, an image acquisition component 120, a processor 130, and a controller 140.

[0065] A running guide rail 200 is provided at the bottom or side of the bridge. An inspection device 110 is slidably mounted on the running guide rail 200. An image acquisition component 120 is mounted on the inspection device 110. The image acquisition component 120 is used to acquire sequential images of the bridge surface.

[0066] The processor 130 is communicatively connected to the image acquisition component 120. The processor 130 is used to acquire a sequence of images and identify the types of bridge defects based on the sequence of images. The processor 130 is also used to determine geometric feature parameters corresponding to the types of bridge defects and to determine the severity level of the defects based on the geometric feature parameters.

[0067] The controller 140 is communicatively connected to the inspection device 110. The controller 140 is used to receive inspection tasks and drive the inspection device 110 to travel along the bridge's running guide rail 200 according to the inspection tasks.

[0068] Specifically, such as Figure 4 As shown, the bridge inspection system 100 described above can be used to perform the following bridge inspection methods, specifically including:

[0069] S100: Receive the inspection task and drive the inspection device 110 according to the inspection task, so that the inspection device 110 travels along the bridge's running guide rail 200.

[0070] It is understood that in this step, external devices or personnel input instructions into the controller 140 in the above embodiment to generate an inspection task. The controller 140 drives the inspection device 110 through the inspection task. During this process, the controller 140 can obtain the driving parameters for driving the inspection device 110 from the inspection task, and the inspection device 110 will travel along the running guide rail 200 according to the driving parameters.

[0071] In reality, this driving parameter can be a fixed value or a non-fixed value, and can be designed according to different scenarios.

[0072] In one example scenario, the inspection task includes an inspection mode and a bridge type. When the controller 140 drives the inspection device 110 to inspect bridges of the same bridge type, the controller 140 can output driving parameters corresponding to the inspection mode according to that bridge type. Specifically, for example, the inspection mode may include a comprehensive inspection mode; then all bridges of the same bridge type can execute the same driving parameters when performing the comprehensive inspection mode. This driving parameter can be related to the setting range of the running guide 200 to maximize the coverage of the detection of all positions on the bridge, ensuring that the image acquisition component 120 can completely acquire the sequence of images of the bridge deck, beams, supports, and piers. For another example, the inspection mode may include a key inspection mode; the key inspection mode is to inspect key locations for defects. In this case, the driving parameter is related to the distance between the starting point of the running guide 200 and the key location, so that the inspection device can reach the key location and perform defect inspection on the key location.

[0073] The running guide 200 can be a straight path or a circular path, adapting to different bridge structures, such as circular overpasses, cable-stayed bridges, arch bridges, or continuous beam bridges, to meet the requirement of covering the key structural areas of the bridge.

[0074] S200: Acquire a sequence of images of the bridge surface acquired by the image acquisition component 120.

[0075] Here, a sequence of images refers to the sequentially arranged images of the bridge surface within a predetermined path area during the journey along that path. These images can be acquired through continuous shooting or by obtaining a sequence of image frames from video.

[0076] In one embodiment, the sequence of images may be denoised, enhanced, and geometrically corrected to improve the accuracy of subsequent disease detection.

[0077] Specifically, in one example, the denoising of a sequence of images may include the following process: establishing multiple matrices for the sequence of image frames at set time intervals or key frame changes, determining the matrix that has the most similar interior points to the standard matrix based on the standard matrix, and using the sequence image corresponding to the matrix as a subsequent disease detection.

[0078] In this way, unlike identifying diseases on each image frame, it avoids the possibility of different image frames acquiring the same disease and recording it as multiple diseases, resulting in inconsistent disease detection. This process can denoise sequential image frames, which helps reduce the amount of subsequent processing of ordered images, achieves filtering of ordered images corresponding to the same disease, improves the accuracy of subsequent disease detection, and also improves computational efficiency.

[0079] S300. Identify the types of defects in the bridge based on sequence images.

[0080] It is understandable that the processor stores a pre-trained deep learning model, which stores a large number of known disease images. Therefore, when a sequence of images is input into the deep learning model, it can go through image analysis, comparison and other processes to determine the known disease image that is most similar to or even the same as the sequence of images. Then it can be output into the deep learning model. The disease type corresponding to the known disease image is the same as the disease type corresponding to the sequence image.

[0081] Specifically, in one example, the comparison between a sequence of images and known disease images can be done by comparing the edges of geometric shapes or by comparing differences in color contrast. It is understandable that different disease types will have different features such as shape and color; therefore, features such as shape and color can be used as a reliable basis for identifying disease types.

[0082] For example, if a known defect image contains a slit-shaped pattern, the corresponding defect is a crack. Similarly, if a sequence of images is identified with an elongated pattern, this sequence of images can be associated with the defect image corresponding to the slit shape, and the defect type corresponding to this sequence of images can be identified as a crack. As another example, if a known defect image includes irregular polygons or curved shapes, the corresponding defect type could be a coating defect such as paint peeling or corrosion. In this case, the sequence of images can be associated with the defect image corresponding to the irregular polygon or curved shape, and the defect type corresponding to this sequence of images can be identified as a coating defect.

[0083] In the examples above, the type of disease is determined by identifying the geometric shapes in the image sequence. This geometric shape can be used as a basis for judgment. In this process, object detection algorithms (such as YOLO, Canny edge detection, or Faster R-CNN) can be used to obtain the geometric shapes corresponding to the diseased areas in the image sequence.

[0084] S400. Determine the geometric characteristic parameters corresponding to the type of bridge damage based on the type of damage.

[0085] Based on the above analysis, the geometric shapes of diseases differ across different disease types. Therefore, the geometric characteristic parameters of different disease types also differ. Even within the same disease type, variations in identical geometric characteristic parameters allow for comparison along the same dimension, providing greater reference value. Identifying geometric characteristic parameters by disease type ensures that the same disease category outputs the same geometric characteristic parameters, providing a scientific reference for subsequent comparisons along the same dimension for classification.

[0086] Among them, geometric characteristic parameters include disease length, disease width, and disease area.

[0087] For example, when the defect is a crack, at least one of the crack length and crack width can be used as a geometric feature parameter; in another example, when the defect is a missing bolt, the missing area can be used as a geometric feature parameter; in yet another example, when the defect is a type of defect such as paint corrosion, paint aging, or paint chalking, the defect area can be used as a geometric feature parameter.

[0088] Geometric feature parameters can be obtained through deep learning models, such as semantic segmentation algorithms (e.g., U-Net, DeepLab). Specifically, in one example, geometric feature parameters are obtained through a semantic segmentation algorithm, enabling pixel-level identification and localization of lesions. This improves the computational accuracy of the geometric feature parameters, thereby enhancing the precision of the obtained lesion severity level.

[0089] S500: Determine the severity level of disease type based on geometric characteristic parameters.

[0090] Based on the above analysis, within the same disease type, identical geometric characteristic parameters of the disease may differ and can be compared along the same dimension. A larger geometric characteristic parameter indicates a higher severity level, while a smaller parameter indicates a lower severity level. Therefore, the severity level of a disease type can be determined based on its geometric characteristic parameters.

[0091] In one example, determining the severity level based on geometric feature parameters can be achieved by comparing the geometric feature parameters with different interval ranges. The interval in which the geometric feature parameters fall determines the severity level of the disease type in that area. This method has a relatively simple logical flow and helps improve the efficiency of disease detection.

[0092] Of course, severity levels can also be determined by inputting geometric feature parameters into a model to dynamically obtain the level. In another example, geometric feature parameters can be input into a stress analysis model to fit the stress state of the bridge, and output a severity parameter based on the bridge's material and structural characteristics, which is then used as the severity level. This method of determining severity levels is more accurate and helps improve the detection precision of defects.

[0093] Understandably, by identifying the type of damage based on sequential images, and then by using the type of damage and the corresponding geometric feature parameters, it is possible to conduct an automated classification assessment of the apparent damage to bridges, thereby providing a good reference for the repair of bridge damage.

[0094] In this way, on the one hand, the absence of subjective manual inspection effectively improves the accuracy of inspection and assessment, and reduces the risk of manual high-altitude inspection. On the other hand, the procedure of identifying, classifying and grading diseases helps to improve the efficiency of calculation, avoids the time increase caused by using a uniform model for the assessment of each disease, and improves the pertinence and accuracy of disease assessment, ensuring that different types of diseases can be reasonably graded according to their unique evaluation standards.

[0095] In one embodiment, such as Figure 5 As shown, step S300 further includes:

[0096] S310. Determine the diseased areas in the sequence images. The diseased areas are used to characterize the diseases present in the sequence images.

[0097] It is understandable that, when a disease exists, there will be differences in color, grayscale, brightness, etc. between the disease and the surrounding area. Based on the deep learning model in the above embodiment, grayscale calculation or brightness calculation can be performed on the image to obtain the diseased area and non-disease area in the sequence image. The diseased area refers to the part of the sequence image where the disease exists, while the non-disease area refers to the part of the sequence image where the disease does not exist.

[0098] Among them, the identification of diseased areas in the sequence images can be done by algorithms such as the Canny edge detection algorithm, or by YOLO, Faster R-CNN, etc., or by using the shadow position that appears within the preset box as the diseased area.

[0099] S320 Calculate the geometry of the affected area.

[0100] The calculation of geometric shapes can be performed using algorithms such as the Canny edge detection algorithm mentioned in the above embodiments, or YOLO, Faster R-CNN, etc.

[0101] S330. Determine the type of disease based on its geometric shape.

[0102] Based on the above analysis, different types of diseases will have different geometric shapes, so geometric shapes can be used as a basis for judging the type of disease.

[0103] For example, when the defect is a crack, the geometry is elongated. When the defect is a paint defect, such as paint peeling or paint corrosion, the geometry is an irregular closed image formed by curves.

[0104] When the defect is a missing bolt, the geometry will be a circular hole structure. The principle for identifying the defect as a missing bolt is as follows: Bolts are usually regular polygonal structures, not circular hole structures. Therefore, when the geometry is a circular hole structure, it can be concluded that the defect of missing bolts does exist.

[0105] It should be noted that in step S310 above, the disease area in the sequence image is determined by comparing the sequence image with known disease images. This process can be done by comparing the sequence images one by one, or by generating a two-dimensional image from the sequence image for comparison.

[0106] In one embodiment, step S310 includes:

[0107] The sequence of images is stitched together to generate a panoramic image of the underside of the bridge or a panoramic image of the side of the beam.

[0108] It is understandable that the image sequence consists of multiple image frames taken sequentially along the bridge. By stitching together these multiple image frames, all image frames can be mapped onto the canvas of the panoramic image, thereby outputting a panoramic image of the area under the bridge.

[0109] Furthermore, by comparing the panoramic image of the bridge underside with multiple known defect images, the geometric shapes corresponding to different defect areas in the panoramic image of the bridge underside are obtained. Then, the defects of the known defect images with the most similar regions in the known defect images are obtained, thereby outputting the defect type.

[0110] In this way, the output of panoramic images under the bridge helps to avoid excessive overlap between adjacent image frames, thereby avoiding interference with the statistical quantity of defects, preventing excessive subsequent errors, and improving the accuracy of defect type identification.

[0111] In addition, before stitching the sequence of images to form a panoramic image under the bridge, the processor 130 can also perform feature detection and extraction on the sequence of images to obtain the position, size, and orientation of each key point; then, feature matching is performed, which essentially involves finding feature point A in one image and finding feature point B in adjacent images that is most similar to feature A; finally, geometric verification is performed, specifically: multiple test matrices are established for feature point A and multiple feature points B, and then the standard matrix is ​​compared one by one with the test matrices to obtain the test matrix with the most points that are similar to the standard matrix. The feature points of this test matrix are output and determined as the final feature points. After repeating the above steps, the images corresponding to the multiple final feature points can be stitched together to obtain the panoramic image under the bridge.

[0112] In this way, these operations avoid ghosting in the panoramic image under the bridge, ensuring the accuracy of the identification of the type of damage.

[0113] In another embodiment, the bridge inspection system 100 further includes a pose adjustment component. The pose adjustment component is used to adjust at least one of the installation height and installation angle of the image acquisition component relative to the inspection device 110. The controller 140 is communicatively connected to the pose adjustment component to drive it to adjust the relative pose of the image acquisition component 120. Figure 6 As shown, step S310 above includes:

[0114] S311. Generate a three-dimensional model of the bridge based on a sequence of images from multiple perspectives.

[0115] It is understandable that when the image acquisition component 120 is in the same pose as the inspection device, it can acquire a series of images. However, by using the pose adjustment component to drive the image acquisition component 120 to adjust its pose, the image acquisition component 120 can be in multiple poses relative to the inspection device 110 for inspection. Therefore, the bridge can be photographed from different perspectives, and multiple series of images can be acquired from multiple perspectives.

[0116] Specifically, the processor can automatically select the pair of images with the most matching points and the most stable geometric relationship as the starting point, and generate the first batch of 3D points through triangulation. Then, the second, third, fourth...Nth image is gradually added; these images are called newly registered images. For each newly registered image, its camera pose is solved using the known 2D-3D point correspondence. New matching points are generated between the newly registered image and existing images. Using these matching points and the known camera pose, new 3D points can be obtained.

[0117] All three-dimensional points are aggregated and arranged according to their corresponding three-dimensional coordinates to form a point cloud, representing the features of the bridge surface. Finally, a three-dimensional model of the bridge can be built using the point cloud and the three-dimensional coordinates.

[0118] S312. Output panoramic images of the bottom or sides of the bridge beam based on the three-dimensional model of the bridge.

[0119] Based on the aforementioned 3D model of the bridge, a panoramic image of the bridge's bottom or sides can be obtained simply by projecting the 3D model vertically. In this process, the 3D to 2D conversion can be achieved through methods such as screenshotting or planar output.

[0120] S313. Based on a preset deep learning model, determine the defect area in the panoramic image of the bottom of the beam or the panoramic image of the side of the beam.

[0121] Based on the above analysis, by comparing the panoramic image of the bridge bottom or the panoramic image of the beam side with multiple known defect images, the geometric shapes corresponding to different defect areas can be obtained from the panoramic image of the bridge bottom or the panoramic image of the beam side. Then, the defect of the known defect image with the most similarity to different areas in the known defect images can be obtained, so that the defect type can be output.

[0122] Outputting panoramic images of the bridge underside or the beam side helps avoid excessive overlap between adjacent image frames, which could interfere with the statistical count of defects and prevent excessive subsequent errors.

[0123] Thus, unlike panoramic images generated by stitching together a sequence of images, which are limited by the shooting angle, panoramic images cannot capture a complete range of images.

[0124] By first establishing a 3D model of the bridge using a sequence of images, and then obtaining a panoramic image of the bridge bottom (or side) from the 3D model, the 3D model of the bridge can be constructed using a sequence of images from different perspectives. This allows for the coverage of various perspectives of the bridge, resulting in higher accuracy in obtaining panoramic images of the bridge bottom and side, and making the images more closely resemble the actual bridge bottom and sides.

[0125] In another embodiment, such as Figure 7As shown, step S330 above may further include:

[0126] S331. Extract the image appearance parameters of the diseased area.

[0127] It is understood that, unlike the geometric feature parameters in the above embodiments, geometric feature parameters are used to characterize the geometric properties of the disease. They are parameters inherent to the object itself and do not change with the environment. Geometric feature parameters can be such as length, width, area, or depth, and can be two-dimensional or three-dimensional data. They can generally be obtained through coordinate methods, calculus methods, etc.

[0128] Image appearance parameters are used to characterize the visual and physical features of diseases. These appearance features are determined by imaging conditions and change with factors such as lighting and viewing angle. Image appearance parameters are two-dimensional data. Appearance features can be color features such as average grayscale value, hue, or grayscale uniformity, or texture features such as graininess and roughness. Appearance features are generally obtained through image processing and other methods.

[0129] Because different types of defects may exhibit similar geometric shapes and geometric feature parameters—such as in coating aging and coating corrosion, where these defects present irregular geometric images—it's difficult to directly determine the defect type by combining feature parameters alone. However, by extracting image appearance parameters and subsequently using both geometric feature parameters and image appearance parameters as the basis for defect type determination, the accuracy of defect type identification can be improved. Since it's almost impossible for different defect types to have completely identical or similar geometric feature parameters and image appearance parameters (meaning that different defect types will not have identical geometric shapes and image appearance parameters), using geometric shape and image appearance parameters as the basis for defect type determination is beneficial for improving the accuracy of defect type identification.

[0130] For example, taking the aforementioned coating corrosion as an example, if the ratio of the roughness of the affected area to that of other non-affected areas is greater than a preset roughness ratio, then the affected area can be inferred to be coating corrosion. Conversely, if the ratio of the roughness of the affected area to that of other non-affected areas is less than a preset roughness ratio, then the affected area can be inferred to be coating peeling.

[0131] For example, taking paint aging as an example, if the brightness ratio between the affected area and other non-affected areas is greater than a preset ratio, it can be inferred that the affected area is paint peeling or paint corrosion. When the brightness ratio between the affected area and other non-affected areas is less than a preset ratio, it can be inferred that the affected area is paint aging (such as blistering of the wall due to moisture).

[0132] S332. Input the geometric shape and image appearance parameters into the preset deep learning model, and the deep learning model outputs at least one disease type.

[0133] Understandably, deep learning models store known disease images, along with the corresponding geometric shapes and image appearance parameters of the diseases in those images. When the geometric shapes and image appearance parameters are output to the deep learning model, the model automatically matches the known disease image that is most similar to the geometric shapes and image appearance parameters, and outputs the disease type corresponding to that known disease image as the corresponding disease type in the sequence of images.

[0134] Therefore, using deep learning models to identify disease types through geometric shapes and image appearance parameters helps improve the accuracy of disease type identification. For example, if the disease type is rust, aging and powdering, or flaking, these types of diseases are irregularly shaped, making it difficult to further subdivide the disease type directly by shape. However, by combining geometric shapes and image appearance parameters, the disease type can be subdivided. In a practical example detection scenario, the presence of a large area of ​​red tones can preliminarily identify the disease type as rust. The presence of a large area of ​​gray tones or other dark tones (refer to the color of cement surfaces) can preliminarily identify the disease type as flaking. When the roughness of the diseased area is significant, it can be preliminarily identified as powdering or rust, which helps improve the accuracy of disease type identification.

[0135] In addition, when a deep learning model stores a sufficient amount of reference data, the following methods can be used to improve the accuracy of disease type identification.

[0136] In one embodiment, such as Figure 8 As shown, step S330 above also includes:

[0137] S333, Based on deep learning model output, the confidence level corresponds to the disease type.

[0138] Confidence level refers to the degree of similarity between at least one of the geometric parameters and image appearance parameters of the disease in the sequence images described in the above embodiments and at least one of the geometric parameters and image appearance parameters of the disease in a known disease image. A higher confidence level indicates a higher degree of similarity and a higher accuracy in disease type identification. Conversely, a lower confidence level indicates a lower degree of similarity and a lower accuracy in disease type identification.

[0139] S334. Is the confidence level greater than a preset threshold? If the confidence level is greater than the preset threshold, proceed to step S3341. If the confidence level is less than or equal to the preset threshold, proceed to step S3342.

[0140] S3341. Adopt this disease type as the identification result.

[0141] S3342. Discard this disease type as the identification result.

[0142] In some embodiments, such as Figure 9 As shown, step S500 above includes:

[0143] S510. Input the geometric feature parameters into the disease assessment model. Based on the multiple preset parameter intervals in the disease assessment model, determine the preset parameter interval to which the geometric feature parameters corresponding to the disease type belong, and output the severity level corresponding to the preset parameter interval.

[0144] Each preset parameter range corresponds one-to-one with a severity level.

[0145] Understandably, the severity level of a disease is output by defining a preset parameter range to which the geometric feature parameters corresponding to the disease type belong. Specifically, the severity level of a disease corresponds to a given preset parameter range, determined by which the geometric feature parameters fall within that range. Thus, by comparing the geometric feature parameters with the preset parameter range, the severity level of the disease is confirmed without needing to calculate the severity level specifically for the geometric feature parameters. This reduces the demands on the algorithm and computing power, thereby improving computational efficiency.

[0146] like Figure 9 As shown, in addition to the bridge inspection methods described above, the following are also included:

[0147] S600. Confirm the location of the diseased area.

[0148] This facilitates the generation of a complete disease report, which reflects the location of the disease, making it easier for manual verification and subsequent repairs.

[0149] Specifically, in one implementation scenario, step S600 above further includes:

[0150] S610, The driving parameters required to confirm the location of the diseased area.

[0151] Understandably, the driving parameters are used to reflect the inspection distance when the inspection device 110 reaches the defect area. The inspection distance is scaled proportionally and positioned on any of the preset bridge side two-dimensional module, beam bottom two-dimensional model, or bridge three-dimensional model, so that the location of the defect area can be recorded on these models.

[0152] Among them, the driving parameters can be the number of rotations of the drive motor when reaching the diseased area, the running time of the inspection device 110, etc.

[0153] S620. Confirm the location of the disease based on the driving parameters.

[0154] Thus, this method of identifying the location of the disease is simple, easy to operate, and does not require complex algorithms.

[0155] In one example scenario, by acquiring the bridge model and the speed of the inspection device 110, and simultaneously obtaining the time period of the defect area, the corresponding defect location is obtained within a preset mapping relationship. In another example scenario, by obtaining the number of rotations of the drive motor that drives the inspection device 110 (which can be obtained through an encoding component, which can be used to obtain the number of rotations of the drive motor), the inspection distance is obtained through the number of rotations and the rotation period, thereby confirming the location of the defect area.

[0156] To facilitate understanding, the calculation process will be explained below using a pre-defined mapping relationship as an example.

[0157] Suppose that the bridge length obtained from the bridge model is L1, and that the inspection device 110 starts from the starting point of the bridge, and the speed of the inspection device 110 is V. If the inspection vehicle obtains a defect area at a certain location after time t.

[0158] Therefore, we can obtain the distance from the location of the defect to the starting point of the bridge as L2 = Vt. Here, we need to determine whether V2 > V1.

[0159] If V2 > V1, then the location of the disease S = If L1 = 1km and L2 = 1.5km, it means that the inspection device 110 is performing a back-and-forth inspection. At this time, it can be determined that the location of the defect is at the midpoint of the preset path.

[0160] If V2 ≤ V1, then the location of the disease S = If L1 = 1km and L2 = 0.5km, it means that the inspection device 110 is performing a back-and-forth inspection. At this time, it can be determined that the location of the defect is at the midpoint of the preset path.

[0161] In some embodiments, such as Figure 11 As shown, the inspection task includes inspection modes and historical disease locations. Inspection modes include a comprehensive inspection mode and a focused inspection mode. Step S100 also includes:

[0162] S110. When the inspection mode is in the full inspection mode, the inspection device 110 is controlled to travel along the first path, which covers all the preset inspection areas on the bridge.

[0163] The preset detection area refers to all areas on the bridge to be inspected. That is, when the inspection mode is in full inspection mode, the inspection device 110 can drive the image acquisition component 120 to perform a full inspection, thereby achieving a comprehensive investigation of all defects on the bridge. Specifically, the first path can be, but is not limited to, a straight path, a back-and-forth path, a circular path, etc.

[0164] In one example, the first path is the path from the starting point to the ending point of the bridge. This allows for a comprehensive inspection of the bridge, increasing the scope of bridge defect detection.

[0165] S120. When the inspection mode is in the key inspection mode, the inspection device 110 is controlled to travel along the second path, which covers one or more historical defect locations.

[0166] In other words, when the inspection mode is in the key inspection mode, it can re-inspect and confirm the locations where diseases have occurred in the past, without the need for a full inspection, which helps to improve the detection efficiency of diseases.

[0167] In one example, the second path is a route from the bridge's starting point to the furthest historical defect location and back. In scenarios where the second path is a round trip, it can be formed between the bridge's starting point and the furthest historical defect location, and this second path can cover other historical defect locations, thus minimizing the second path and improving inspection efficiency.

[0168] Thus, by driving the inspection device 110 in different inspection modes, on the one hand, the accuracy of the drive is improved. When the inspection mode is in the key inspection mode, the inspection device 110 does not need to travel along the first path, reducing the amount of travel and improving inspection efficiency. On the other hand, when the inspection mode is in the comprehensive inspection mode, it can ensure that the inspection range is large enough to fully detect defects.

[0169] Other examples, such as Figure 12 As shown, step S120 may further include:

[0170] S121. Obtain multiple different historical disease locations.

[0171] Historical disease locations refer to the locations of diseased areas detected after each inspection task is completed. These locations are stored by the processor as historical disease locations.

[0172] S122. Based on the coordinates of multiple historical defects along the length of the bridge, sort them to generate a location sequence.

[0173] S130. Based on the position sequence, plan the second path.

[0174] It is understandable that when the bridge is a type such as a ring bridge, and there are no two identical defect areas in the same location, the defect locations can be sorted sequentially to form a location sequence. Then, a second path can be planned based on the location sequence. This ensures that the second path covers the historical defect locations to be detected, while also minimizing the travel path of the inspection device 110, which helps to improve the efficiency of inspection.

[0175] In one embodiment, such as Figure 13 As shown, in addition to the bridge inspection methods described above, the following are also included:

[0176] S810. Retrieve the historical geometric feature parameters stored at the location of the disease in this inspection task from the historical disease database.

[0177] Historical geometric feature parameters refer to the historical geometric feature parameters of the disease area corresponding to the historical disease location stored by the processor. Within the same inspection task, there is a one-to-one correspondence between historical disease locations and historical geometric feature parameters.

[0178] S820. Compare the geometric characteristic parameters of the disease obtained in this inspection with the historical geometric characteristic parameters, and calculate at least one quantitative indicator of disease development.

[0179] It is understandable that at the same disease location, there exist historical geometric characteristic parameters and updated geometric characteristic parameters obtained from the current inspection. Since they both belong to the same disease type, there is a time-related change parameter between the historical and updated geometric characteristic parameters. This change parameter is the quantitative indicator of disease development. One disease area corresponds to one quantitative indicator of disease development, and multiple disease areas correspond to multiple quantitative indicators of disease development.

[0180] Among them, the quantitative indicators for disease development include at least one of the following: disease size growth rate, disease grade change, and disease development rate obtained by fitting geometric features based on multiple consecutive inspection tasks.

[0181] S830. Determine whether the quantitative indicator of disease development is less than the first preset threshold. If yes, proceed to step S831. If not, proceed to step S832.

[0182] S831. Determine if the disease is developing slowly.

[0183] S832. Determine whether the quantitative indicator of disease development is less than the second preset threshold. Wherein, the first preset threshold is less than the second preset threshold. If yes, proceed to step S8321. If not, proceed to step S8322.

[0184] S8321. Determine if the disease is developing stably.

[0185] S8322, Judging the accelerated development of the disease.

[0186] In this way, the development trend of defects can be predicted, providing an effective reference for whether emergency repairs are needed, which is conducive to further optimizing the application scenarios of the bridge inspection system 100.

[0187] In conjunction with any of the above-described embodiments of the processor, see back Figure 3 as well as Figure 3 The processor 130 includes a mobile processor 131 and a cloud processor 132. The mobile processor 131 and the cloud processor 132 are communicatively connected.

[0188] The mobile processor 131 is used to acquire a sequence of images and identify the types of bridge defects based on the sequence of images. The cloud processor 132 is also used to determine geometric feature parameters corresponding to the types of bridge defects and to determine the severity level of the defects based on the geometric feature parameters.

[0189] In this way, by having the mobile processor 131 and the cloud processor 132 perform different functions, it is beneficial to reduce the computational burden on the cloud processor 132, thereby reducing the requirements for computing power and algorithms, reducing the operational requirements of the cloud processor 132, and consequently reducing costs and the risk of lag in the cloud processor 132.

[0190] It should be noted that the mobile processor 131 can be a programmable logic controller (PLC) or an embedded controller, etc. The cloud processor 132 can be an Internet of Things platform or a cloud resource management platform, etc.

[0191] In one embodiment, see back Figure 1 as well as Figure 2 The bridge inspection system 100 also includes an environmental sensor 101, which is used to acquire environmental data at the location of the inspection device 110. The environmental sensor 101 is communicatively connected to a controller 140, which is used to control the inspection device 110 to change its speed based on the environmental data.

[0192] The environmental data can be any one or a combination of liquid volume, air volume, and light intensity. This allows the inspection device 110 to adjust its speed based on the environmental data, adapting to different inspection environments and making inspections safer.

[0193] In one implementation, see back Figure 1 as well as Figure 2The environmental sensor 101 includes a wind sensor 1011. The wind sensor 1011 is used to detect wind speed data. The controller 140 is configured to control the drive motor 112 to reduce the actual speed of the inspection device 110 to a first speed value or below when the wind speed exceeds a first preset threshold.

[0194] Thus, when extreme weather occurs, such as when the wind is too strong, the inspection device 110 can be slowed down or stopped, which helps to ensure the safety of the operation during the inspection process.

[0195] In another embodiment, refer to Figure 1 As shown, the environmental sensor 101 includes a light detection sensor for detecting visibility data. The controller 140 is configured to control the drive motor 112 to reduce the actual speed of the inspection device 110 to the second speed value or below when the visibility is below a second preset threshold.

[0196] It is understandable that the air is the medium through which the light emitted by the light detection sensor propagates. However, in extreme weather conditions such as fog, haze, or rain, the speed of light propagation slows down and light loss becomes severe. In such cases, the visibility speed feedback is the speed of light propagation or light intensity data. Thus, when extreme weather conditions such as fog, haze, or rain cause excessively low visibility, the actual speed of the inspection device 110 is reduced to the second speed value or below, thereby improving the safety of the inspection process.

[0197] In other implementations, see back. Figure 1 as well as Figure 2 The environmental sensor 101 includes an obstacle sensor 1012, such as an optical sensor or an acoustic sensor. The obstacle detection sensor 1012 is used to detect the distance between the obstacle in front and the inspection device 110. When the distance is greater than a third preset threshold, the actual speed of the inspection device 110 is reduced to the third speed value or below, and a warning command is issued to allow maintenance personnel or robots to clear the obstacle.

[0198] In some implementations, see back Figure 2 The inspection device 110 includes a vehicle body 111, a drive motor 112, and an encoding component 113. The drive motor 112 is driven by rollers on the vehicle body 111 to move the vehicle body 111. The encoding component 113 is electrically connected to the drive motor 112 and is used to adjust the drive parameters of the drive motor 112 in real time according to the actual position of the inspection device 110.

[0199] The coding component 113 can be configured such that the positioning accuracy of the inspection device 110 to the target defect location is less than or equal to a preset distance value, such as 5cm. That is, the actual distance between the inspection device 110 and the bridge starting point and the target distance between the target defect location and the bridge starting point are less than or equal to the preset distance value, such as 5cm.

[0200] This improves the alignment between the location captured by the image acquisition component 120 and the target lesion location, thereby enhancing the quality and accuracy of lesion identification.

[0201] In one example, see back Figure 2 The inspection device 110 also includes a data read / write terminal. The distance data detected by the obstacle sensor 1012 in the above embodiment is transmitted to the controller 140 through the data read / write terminal, which can convert the distance into a digital signal. In this way, by processing the data from the obstacle sensor 1012 through the data read / write terminal, the signal conversion difficulty of the controller 140 is reduced, which is beneficial to improving the driving efficiency.

[0202] In conjunction with any of the embodiments of the mobile processor 131 described above, see back Figure 1 as well as Figure 2 The mobile processor 131 also includes a communication module 150. Specifically, the controller 140 is communicatively connected to the cloud processor 132 via the communication module 150.

[0203] In one example, the communication module 150 includes a data processing unit 151 and a domain control unit 152. The data processing unit 151 facilitates data interaction between the cloud processor 132 and the domain control unit 152. The domain control unit 152 interacts with the controller 140, enabling the inspection task to be parsed into different instructions, which are then distributed to the corresponding modules by the controller 140. This improves the speed of control interaction while ensuring the quality of data transmission and processing.

[0204] In conjunction with any embodiment of the image acquisition component 120 described above, see back Figure 2 The image acquisition component 120 includes a camera 121 and a switch 122. The camera 121 is communicatively connected to the processor 130 via the switch 122. Thus, the switch 122 enables the camera 121 to be networked, allowing the processor 130 and other electronic devices to access the camera 121 to transmit commands and receive data. Furthermore, the switch 122 provides a stable data channel for the camera 121, which helps ensure the transmission quality of sequential images.

[0205] In one embodiment, see back Figure 1 as well as Figure 2The bridge inspection system 100 includes a power supply 160. The power supply 160 includes a power detection sensor 161 and a battery 162. The power detection sensor 161 is used to detect the power level of the battery 162 and is communicatively connected to a controller 140. The controller 140 is also configured to control the drive motor 112 to stop driving when the power level is less than a preset value, thus pausing the inspection device 110.

[0206] Furthermore, in one embodiment, such as Figure 2 As shown, the power supply 160 also includes a charging device 163. The charging device 163 is used to charge the battery 162. In one example, the charging device 163 can be communicatively connected to a controller 140, which is also configured to control the drive motor 112 to stop driving and control the charging device 163 to charge the battery when the battery level is less than a preset value.

[0207] In another aspect, this application also provides a computer-readable storage medium that, when the instructions in the computer-readable storage medium are executed by a processor of an electronic device, enables the electronic device to perform the bridge inspection method of the above embodiments.

[0208] Other aspects of this application also provide a computer program product, including a computer program. When executed by a processor, the computer program implements the steps of the bridge inspection method described in the above embodiments.

[0209] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, image processors, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0210] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method of inspecting a bridge, characterized by, The bridge is provided with a running rail, a patrol device is slidingly arranged on the running rail, the patrol device is provided with an image acquisition assembly, and the patrol method comprises the following steps: A patrol task is received, and the patrol device is driven according to the patrol task, so that the patrol device travels along the running rail of the bridge; Sequence images of the surface of the bridge collected by the image acquisition assembly are obtained; The disease type of the bridge is identified based on the sequence images; The geometric feature parameters corresponding to the disease type of the bridge are determined according to the disease type of the bridge, and the severity level of the disease type is determined based on the geometric feature parameters.

2. The method of claim 1, wherein, The disease type of the bridge is identified based on the sequence images, which comprises the following steps: A disease area in the sequence images is determined; the disease area is used to represent the disease existing in the sequence images; The geometric shape corresponding to the disease area is calculated; The disease type is determined according to the geometric shape.

3. The method of claim 2, wherein, The disease type of the bridge is identified based on the sequence images, which further comprises the following steps: An image appearance parameter of the disease area is extracted; The geometric shape and the image appearance parameter are input into a preset deep learning model, and at least one disease type is output by the deep learning model.

4. The method of claim 1, wherein, The severity level of the disease type is determined based on the geometric feature parameters, which comprises the following steps: The geometric feature parameters are input into a disease evaluation model, and a plurality of preset parameter intervals are preset in the disease evaluation model, one of which corresponds to one severity level; The preset parameter interval to which the geometric feature parameters corresponding to the disease type belong is determined, and the severity level corresponding to the preset parameter interval is output.

5. The method of claim 1, wherein, When the disease type is a crack disease, the geometric feature parameters include at least one of a disease length and a disease width; And / or, when the disease type is a painting disease, the geometric feature parameters include a disease area; And / or, when the disease type is a bolt missing disease, the geometric feature parameters include a disease area.

6. The method of claim 2, wherein, The bridge patrol method further comprises the following steps: The disease position where the disease area is located is confirmed.

7. The method of claim 6, wherein, The disease position where the disease area is located is confirmed, which comprises the following steps: A driving parameter required when the disease area is confirmed is obtained; the driving parameter is used to reflect the patrol distance of the patrol device when the disease area is reached; The disease position is confirmed based on the driving parameter.

8. The method of claim 6, wherein, The patrol task comprises a patrol mode and a historical disease position, and the patrol mode comprises a comprehensive patrol mode and a key patrol mode; The patrol device is driven according to the patrol task, so that the patrol device travels along the running rail of the bridge, which comprises the following steps: When the patrol mode is in the comprehensive patrol mode, the patrol device is controlled to travel along a first path; the first path covers all preset detection areas on the bridge; When the patrol mode is in the key patrol mode, the patrol device is controlled to travel along a second path, and the second path covers one or more historical disease positions.

9. The method of claim 6, wherein, The bridge patrol method further comprises the following steps: From the historical disease database, the historical geometric feature parameters stored at the disease location of the current inspection task are called; The geometric feature parameters of the disease obtained by the current inspection are compared with the historical geometric feature parameters, and at least one disease development quantitative index is calculated; When the disease development quantitative index is less than a first preset threshold, it can be judged that the disease develops slowly; when the disease development quantitative index is greater than or equal to the first preset threshold and less than a second preset threshold, it can be judged that the disease develops stably; when the disease development quantitative index is greater than or equal to the second preset threshold, it can be judged that the disease develops rapidly. The first preset threshold is less than the second preset threshold.

10. A bridge inspection system, comprising: It comprises: An inspection device is provided with a running guide rail; the inspection device is slidingly arranged on the running guide rail; An image acquisition component is installed on the inspection device; the image acquisition component is used to acquire sequence images of the surface of the bridge; A processor is in communication connection with the image acquisition component; the processor is used to acquire the sequence images and identify the disease type of the bridge based on the sequence images; the processor is also used to determine the geometric feature parameters corresponding to the disease type according to the disease type of the bridge, and determine the severity level of the disease type based on the geometric feature parameters; A controller is in communication connection with the inspection device; the controller is used to receive an inspection task and drive the inspection device to travel along the running guide rail of the bridge according to the inspection task.

11. The bridge inspection system of claim 10, wherein, The processor comprises a mobile processor and a cloud processor; the mobile processor and the cloud processor are in communication connection; The mobile processor is used to acquire the sequence images and identify the disease type of the bridge based on the sequence images; the cloud processor is also used to determine the geometric feature parameters corresponding to the disease type according to the disease type of the bridge, and determine the severity level of the disease type based on the geometric feature parameters.

12. The bridge inspection system of claim 11, wherein, The bridge inspection system further comprises an environment sensor; the environment sensor is used to acquire environmental data of the location of the inspection device; the environment sensor is in communication connection with the controller, and the controller is used to control the inspection device to change speed according to the environmental data.

13. A computer-readable storage medium, characterized in that, When the instructions in the computer readable storage medium are executed by the processor of the electronic device, the electronic device can execute the bridge inspection method of any one of claims 1 to 9.

14. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the bridge inspection method of any one of claims 1 to 9.

Images