A bridge crack image recognition method based on drones and AI algorithms

Abstract

The present invention discloses a bridge crack image recognition method based on a drone and an AI algorithm, and relates to the technical field of bridge crack image recognition. The method comprises: acquiring image information data of a target bridge to obtain an image information set, wherein the image information set includes at least one image information item consisting of a target bridge picture; acquiring a judgment area of ​​the image information item to obtain a judgment target item, wherein the judgment target item is used to represent the crack judgment area of ​​the target bridge picture. The present invention optimizes image quality through a two-stage illumination compensation mechanism. In the first stage, basic compensation adopts channel splitting and global mean calculation, and linearly corrects the original pixels in combination with the calibrated grayscale value to eliminate the overall deviation of the ambient light. In the second stage, shadow compensation of the sun's incident angle is introduced, and the shadow interference is dynamically corrected through the attenuation coefficient in combination with the sky segmentation area distance parameter and the incident angle, thereby reducing the deviation rate of the compensated pixel value and providing high-quality input for subsequent recognition.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge crack image recognition, and specifically to a bridge crack image recognition method based on a drone and an AI algorithm. Background Art

[0002] Bridge crack image recognition refers to the use of image processing and machine learning technology to analyze photos or scanned images taken of the bridge surface to detect and identify whether there are cracks on the bridge surface, as well as the type, location and extent of the cracks. Drones and AI algorithms are a technical solution that combines drone technology and artificial intelligence algorithms to automatically detect and identify cracks on the bridge surface. Its main idea is to use drones to collect images of bridges, and then use AI algorithms to analyze the collected images and identify cracks.

[0003] Patent publication number CN119540240A is a bridge crack detection method based on image recognition. The method splits the bridge image into three monochrome channels of red, green, and blue to obtain three spectral images. The spectral image corresponding to each channel is processed separately, and the suspected points on each spectral image are extracted. Then, based on each suspected point on each spectral image, the channel contrast value and regional trajectory distortion value of the suspected point are calculated to reflect the subtle texture and structural characteristics of the crack. Then, the red, green, and blue channel weight matrices are constructed to characterize the weight of the crack on the three channels. The weight of the pixel points on the crack is quantified by the red, green, and blue channel weight matrices, which can more accurately distinguish between real cracks and background noise, significantly reduce the false detection rate, and improve the reliability of crack detection.

[0004] When the above-mentioned and similar technical solutions analyze the acquired bridge images, the images will be affected by sunlight during the image acquisition process, and the intensity, color temperature, and incident angle of sunlight will change in real time with the weather, time, and season, which will lead to certain differences in the acquired image data. As a result, omissions or incorrect markings are prone to occur during image recognition and analysis. At the same time, the texture and background interference of the bridge body due to the material or age of the bridge itself will also interfere with the image recognition and analysis, thereby increasing the risk of recognition errors. Summary of the Invention

[0005] The purpose of the present invention is to provide a bridge crack image recognition method based on drones and AI algorithms to solve the problems raised in the above background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a bridge crack image recognition method based on drones and AI algorithms, comprising:

[0007] Acquire image information data of a target bridge to obtain an image information set, wherein the image information set includes at least one image information item consisting of a picture of the target bridge;

[0008] Acquire a determination area of ​​the image information item to obtain a determination target item, where the determination target item is used to represent a crack determination area of ​​the target bridge image;

[0009] Based on the determination target item, the original image pixel value of the determination target item is obtained to obtain the initial pixel item; based on the image information item, the channel mean of the image information item is obtained; a basic illumination compensation formula is created as the first stage compensation to obtain the first compensation item;

[0010] Acquire the time series information of the image information item to obtain a time series information set corresponding to the image information set; based on the time series information set, obtain the illumination incident angle information when the image information item interacts with the target bridge in a state corresponding to the determination target item to obtain the incident information item; simultaneously obtain the distance between the pixel coordinates in the image information item and the sky segmentation area and the maximum sky distance value in the image information item; and create a solar incident angle shadow compensation formula as the second compensation stage;

[0011] Based on the first-stage compensation and the second-stage compensation, the initial pixel items are segmentedly compensated, and the compensated pixel values ​​of the target items are obtained to obtain the processed pixel items, thereby achieving compensation for the comprehensive basic illumination and the shadow of the sun's incidence angle;

[0012] The identification features of the processed pixel items are obtained to obtain the identification feature items, and the identification feature items are judged, thereby realizing the recognition and judgment of the bridge crack image.

[0013] Furthermore, the method for creating the basic illumination compensation formula includes:

[0014] Based on the image information item, split the image information item according to the number of channels to obtain at least two split channel items;

[0015] Get the global average of the split channel items respectively to get the channel mean item;

[0016] Create a basic lighting compensation formula:

[0017] ;

[0018] in is the image pixel value after basic lighting compensation, is the initial pixel term, To calibrate the grayscale value, is the channel mean term;

[0019] The first stage pixel compensation data of the initial pixel item is obtained based on the basic illumination compensation formula to obtain a first compensation item.

[0020] Furthermore, the method for obtaining the channel mean term includes:

[0021] Based on the split channel item, create the channel mean acquisition formula:

[0022] ;

[0023] in is the total number of pixels, The value of the split channel item with coordinates (x, y) in the image information item.

[0024] Furthermore, the method for obtaining the image information set includes:

[0025] Set the acquisition module, which is a UAV module, to obtain the overall information of the target bridge and obtain the target information item;

[0026] An acquisition path is created based on the target information item, and the acquisition module is path-limited based on the acquisition path. The height of the acquisition module is limited based on the initial height information of the acquisition module and the target bridge. The drone module is equipped with a camera component, and image information data of the target bridge is acquired based on the camera component, thereby obtaining an image information set.

[0027] Furthermore, the acquisition module further includes a photosensitive component, and the method for acquiring the incident information item includes:

[0028] Obtaining the image information item where the determination target item is located, and obtaining the target information item;

[0029] Based on the time series information set, obtaining a time series information item corresponding to the target information item to obtain a target time series item;

[0030] Based on the target timing item, the acquisition module obtains the incident angle information of the light through the photosensitive component to obtain the incident information item.

[0031] Furthermore, the method for creating the solar incident angle shadow compensation formula includes:

[0032] Identify and determine the sky pixel area in the image information item to obtain a divided area item;

[0033] Obtaining the maximum value information of the pixel distance division area item in the image information item to obtain the maximum distance value;

[0034] Obtain the distance information between the determination target item and the divided area item to obtain the determination distance value;

[0035] Create the shadow compensation formula for the sun incidence angle:

[0036] ;

[0037] in is the image pixel value after the sun’s incident angle shadow compensation, is the image pixel value after basic lighting compensation, is the intensity attenuation coefficient, is the incident information item, To determine the distance value, is the maximum distance value.

[0038] Furthermore, the method for determining the identification feature item includes:

[0039] Based on the processed pixel items, color features, shape features, and range features are respectively obtained to obtain color recognition items, shape recognition items, and range recognition items, and the color recognition items, shape recognition items, and range recognition items are combined to obtain recognition feature items;

[0040] An impact information library is created, which includes background impact data and impurity impact data. Based on the comparison results between the impact information library and the identification feature items, it is determined whether the identification feature items are in the impact information library, thereby realizing the determination of the identification feature items.

[0041] Furthermore, the method for creating the impact information database includes:

[0042] Obtaining initial state data of the target bridge, the initial state data including the background color and background texture of the bridge;

[0043] Obtaining initial impurity data of the target bridge, the initial impurity data including residual impurity color, residual impurity size, and residual impurity shape;

[0044] An impact information library is created based on the initial state data and the initial impurity data as data filling.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] This bridge crack image recognition method based on drones and AI algorithms optimizes image quality through a two-stage illumination compensation mechanism. In the first stage, basic compensation uses channel splitting and global mean calculation, combined with calibrated grayscale values ​​to perform linear correction on the original pixels to eliminate the overall deviation of ambient light. In the second stage, shadow compensation for the sun's incident angle is introduced. Combined with the sky segmentation area distance parameter and the incident angle, shadow interference is dynamically corrected through the attenuation coefficient to solve the problem of insufficient crack contrast under complex lighting conditions such as backlight and cloudy weather, reduce the deviation rate of the compensated pixel value, and provide high-quality input for subsequent recognition.

[0047] At the same time, a flight path is generated based on the size of the bridge, which constrains the drone's linear scanning to avoid shooting blind spots. The light incident angle is obtained in real time by associating the photosensitive components with the time series information set to ensure the timeliness of the shadow compensation parameters. The drone is equipped with a high-definition camera and photosensitive plate to achieve synchronous "image-light angle" acquisition, completing the acquisition of image and light data in a short time. The efficiency is significantly improved compared to traditional manual inspection, and the data has high temporal and spatial consistency.

[0048] Finally, the color of the compensated pixel values ​​is analyzed, such as dark crack patterns, linear extension shapes, and local concentration range features. By influencing the construction of the information database, the background color and texture of the bridge, such as the grayscale texture of concrete, and common impurity data, such as the color of bird droppings and the shape of rust spots, are integrated to establish a false positive factor database. The identification feature items are matched with the impurity data in the library for similarity, and non-crack interference is eliminated. The false detection rate of bridge background color differences and attached impurities is reduced, and real cracks and residual stains are effectively distinguished, thereby improving the reliability of AI recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic diagram of the overall process of the present invention;

[0050] Figure 2 A schematic diagram of the image information item acquisition process of the present invention;

[0051] Figure 3 This is a schematic diagram of the image data acquisition process of the present invention;

[0052] Figure 4 This is a schematic diagram showing image b of the present invention;

[0053] Figure 5 This is a schematic diagram of the target information item acquisition process of the present invention. DETAILED DESCRIPTION

[0054] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0055] In the bridge crack image recognition based on drones and artificial intelligence algorithms, the quality of image data is crucial and directly affects the accuracy and reliability of crack recognition. However, in practical applications, the bridge image acquisition process will inevitably be interfered by multiple factors, resulting in certain differences and complexity in the image data, which in turn increases the difficulty of crack recognition. First of all, sunlight is one of the important factors affecting the quality of bridge images. In the process of drones acquiring bridge images, the intensity, color temperature, and incident angle of sunlight will change in real time with weather, time, and season. These changes cause the acquired image data to present different lighting conditions. For example, in clear weather, the light intensity is high, and strong light spots and shadows may appear on the bridge surface, which reduces the contrast between the cracks and the background and increases the difficulty of crack recognition. At the same time, changes in the incident angle will also affect the image quality. When light shines on the bridge surface at an oblique angle, it will produce longer shadows, causing some cracks to be shadowed. Occlusion makes it impossible to detect. Therefore, changes in sunlight bring great challenges to bridge crack image recognition, making it difficult for the algorithm to accurately identify cracks. The bridge crack image recognition method based on drones and AI algorithms provided in this application optimizes image quality through a two-stage illumination compensation mechanism. In the first stage, basic compensation adopts channel splitting and global mean calculation, and linearly corrects the original pixels in combination with calibrated grayscale values ​​to eliminate the overall deviation of ambient light. In the second stage, shadow compensation for solar incident angle is introduced, and combined with the distance parameter and incident angle of the sky segmentation area, shadow interference is dynamically corrected through the attenuation coefficient to solve the problem of insufficient contrast of cracks under complex lighting conditions such as backlight and cloudy weather. The flight path is generated based on the bridge size, the drone is constrained to scan linearly, and the light incident angle is obtained in real time by associating the photosensitive components with the time series information set to ensure the timeliness of the shadow compensation parameters. The drone is equipped with a high-definition camera and a photosensitive plate to realize the synchronous acquisition of "image-light angle", completing the acquisition of image and light data in a short time. Figure 1 As shown, steps S100-S700 are included.

[0056] Step S100: Acquire image information data of a target bridge to obtain an image information set.

[0057] It should be noted that if Figure 2 As shown, the image information set includes at least one image information item consisting of a target bridge picture, and the method for acquiring the image information set includes: setting an acquisition module, the acquisition module is a drone module, acquiring overall information of the target bridge, and obtaining a target information item; creating an acquisition path based on the target information item, limiting the path of the acquisition module based on the acquisition path, limiting the height of the acquisition module based on the initial height information of the acquisition module and the target bridge, the drone module is equipped with a camera component, and image information data of the target bridge is acquired based on the camera component, thereby obtaining an image information set.

[0058] In the specific implementation process, Figure 3 As shown, it is necessary to perform crack detection on a bridge A. First, the width of the bridge is obtained to be 30m and the length is 90m. The overall information of Bridge A is obtained, and then the target information item is obtained. The image information of Bridge A needs to be obtained and subsequently analyzed. By setting the UAV module, the UAV module is equipped with a camera component, which is a high-definition camera. According to the target information item of Bridge A, the acquisition path is set. First, the two ends of Bridge A are set as the acquisition starting point and the acquisition end point respectively. Based on the connection line between the acquisition starting point and the acquisition end point as the acquisition path, the path of the UAV module is limited so that the UAV module moves along the acquisition path. When the image information of Bridge A is started to be acquired, the height information of the UAV module from Bridge A is obtained. The height of the UAV module is limited based on the height information, and then the image information data of Bridge A is acquired based on the camera component on the UAV module to obtain an image information set. The image information set includes three bridge pictures of Bridge A.

[0059] Step S200: Acquire a determination area of ​​an image information item to obtain a determination target item.

[0060] It should be noted that the target item is used to represent the crack determination area of ​​the target bridge image. After obtaining the image information set, image recognition judgment is required. When the image shows that the entire bridge has suspected cracks, the area with suspected cracks is determined as the target item.

[0061] During the specific implementation process, image information of a certain bridge is now obtained. When identifying and judging the image information, it is found that there is an irregular shadow part at a certain position in one of the images, while the shadow part does not exist in other positions and other images. At this time, this area is determined to be the judgment target item.

[0062] Step S300: Create a basic illumination compensation formula as the first stage compensation to obtain a first compensation item.

[0063] It should be noted that the method for creating the basic illumination compensation formula includes: based on the image information item, splitting the image information item according to the number of channels to obtain at least two split channel items; obtaining the global average value of the split channel items respectively to obtain the channel mean item;

[0064] Create a basic lighting compensation formula:

[0065] ;

[0066] in is the image pixel value after basic lighting compensation, is the initial pixel term, To calibrate the grayscale value, To calibrate the grayscale value, the calibrated grayscale value is 128, and the first stage pixel compensation data of the initial pixel item is obtained based on the basic illumination compensation formula to obtain the first compensation item.

[0067] It should be noted that based on the determination target item, the original image pixel value of the determination target item is obtained to obtain the initial pixel item, and based on the image information item, the channel mean of the image information item is obtained. The method for obtaining the channel mean item includes:

[0068] Based on the split channel item, create the channel mean acquisition formula:

[0069] ;

[0070] in is the total number of pixels, The value of the split channel item with coordinates (x, y) in the image information item.

[0071] In the specific implementation process, Figure 4 As shown in the figure, when crack image recognition is performed on a bridge B, an image b of the bridge B is obtained. There is a position c in the image b that is suspected to be a crack. At this time, the original image pixel value of position c is (50, 70, 90). At this time, the image b is first split into three channels: red, green, and blue, which are set as channel 1, channel 2, and channel 3 respectively. The mean value information of these three channels in image b is obtained respectively. It is obtained that there are ten pixels in image b. The channel values ​​of each pixel are shown in Table 1:

[0072] Table 1

[0073]

[0074] At this time, the formula is obtained based on the channel mean:

[0075] ;

[0076] The calculation shows that:

[0077] ;

[0078] ;

[0079] ;

[0080] That is, the channel means of channel 1, channel 2, and channel 3 are 85, 90, and 60 respectively;

[0081] At this time, according to the basic lighting compensation formula:

[0082] ;

[0083] Since the original image pixel values ​​are (50, 70, 90), the initial pixel items are 50, 70, and 90 respectively;

[0084] The corresponding calculation results are:

[0085] ;

[0086] ;

[0087] ;

[0088] 75.5 is rounded to 76, that is, the image pixel value after basic illumination compensation is (76, 99, 192).

[0089] Step S400: Obtaining the illumination incident angle information when the image information item interacts with the target bridge in a state corresponding to the determination target item, and obtaining the incident information item.

[0090] It should be noted that the timing information of the image information item is obtained to obtain a timing information set corresponding to the image information set. Based on the timing information set, the acquisition module also includes a photosensitive component, and the method for obtaining the incident information item includes: obtaining the image information item where the target item is located to obtain the target information item; based on the timing information set, obtaining the timing information item corresponding to the target information item to obtain the target timing item; based on the target timing item, the acquisition module obtains the incident angle information of the light through the photosensitive component to obtain the incident information item.

[0091] In the specific implementation process, Figure 5 As shown, it is necessary to perform crack detection on a bridge C. Three image information items of the bridge are obtained through the acquisition module composed of the UAV module, namely Image 1, Image 2 and Image 3, and the corresponding time sequence information is 8:00, 8:10 and 8:20 respectively, thereby obtaining a time sequence information set. At this time, a certain area in Image 2 is determined to be the determination area item, and the target information item is obtained as Image 2. The time sequence information corresponding to Image 2 is obtained as 8:10. Since a photosensitive component is also provided on the acquisition module, which is a photosensitive plate, the incident angle of light at 8:10 is obtained as 45° through the photosensitive plate, thereby obtaining the incident information item.

[0092] Step S500: Create a solar incidence angle shadow compensation formula as the second compensation stage.

[0093] It should be noted that the method for obtaining the distance between the pixel coordinates in the image information item and the sky segmentation area and the maximum sky distance value in the image information item, and the method for creating the solar incidence angle shadow compensation formula include: identifying and determining the sky pixel area in the image information item to obtain the segmentation area item; obtaining the maximum value information of the distance between the pixel point in the image information item and the segmentation area item to obtain the maximum distance value; obtaining the distance information between the determination target item and the segmentation area item to obtain the determination distance value;

[0094] Create the shadow compensation formula for the sun incidence angle:

[0095] ;

[0096] in is the image pixel value after the sun’s incident angle shadow compensation, is the image pixel value after basic lighting compensation, is the intensity attenuation coefficient, which is set to 0.5. is the incident information item, To determine the distance value, is the maximum distance value;

[0097] In the specific implementation process, when crack image recognition is performed on a bridge B, an image b of the bridge B is obtained. There is a position c in the image b that is suspected to be a crack. At this time, the original image pixel value of position c is (50, 70, 90). After basic illumination compensation, the image pixel value of position c is (76, 99, 192). At the same time, when the image b is obtained, the incident angle of the light is 45°, that is, is 45°, and after obtaining the sky pixel area in image b, the maximum value information of the distance between the pixel point in image b and the sky pixel area is 500 pixels, so is 500, and the distance between the area where position c is located and the sky pixel area is 200 pixels, so is 200, so according to the formula:

[0098] ;

[0099] The calculation results are:

[0100] ;

[0101] That is, after the shadow compensation of the solar incidence angle, the pixel value of position c is (103, 126, 219).

[0102] Step S600: performing segmented compensation on the initial pixel items based on the first stage compensation and the second stage compensation.

[0103] It should be noted that the compensated pixel value of the target item is obtained to obtain the processed pixel item, thereby realizing the compensation of comprehensive basic illumination and solar incident angle shadow. The image quality is optimized through a two-stage illumination compensation mechanism. In the first stage, the basic compensation adopts channel splitting and global mean calculation, and combines the calibrated grayscale value to perform linear correction on the original pixel to eliminate the overall deviation of the ambient light. In the second stage, solar incident angle shadow compensation is introduced, combined with the sky segmentation area distance parameter and the incident angle, and the shadow interference is dynamically corrected through the attenuation coefficient to solve the problem of insufficient crack contrast under complex lighting conditions such as backlight and cloudy weather, and reduce the deviation rate of the compensated pixel value.

[0104] Step S700: Obtain identification features of the processed pixel items to obtain identification feature items.

[0105] It should be noted that the identification feature items are judged, thereby realizing the identification and judgment of the bridge crack image. The method for judging the identification feature items includes: based on processing pixel items, respectively obtaining color features, shape features and range features to obtain color recognition items, shape recognition items and range recognition items, and combining the color recognition items, shape recognition items and range recognition items to obtain identification feature items; creating an impact information library, the impact information library includes background impact data and impurity impact data, and based on the comparison results between the impact information library and the identification feature items, judging whether the identification feature items are in the impact information library, thereby realizing the judgment of the identification feature items.

[0106] Specifically, by creating an influencing information database, factors that are prone to misjudgment in the process of identifying bridge crack images, such as the bridge background and impurities attached to the bridge, are obtained. By collecting these factors and comparing the color features, shape features, and range features extracted from the processed pixel items with the factor features in the influencing information database, the determination of the identification feature items is achieved.

[0107] It should be noted that the method for creating the impact information library includes: obtaining the initial state data of the target bridge, which includes the background color and texture of the bridge; obtaining the initial impurity data of the target bridge, which includes the residual impurity color, residual impurity size, and residual impurity shape; and creating the impact information library based on the initial state data and the initial impurity data as data filling.

[0108] Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is limited by the accompanying embodiments and their equivalents.

Claims

1. A bridge crack image recognition method based on drones and AI algorithms, comprising: Acquire image information data of a target bridge to obtain an image information set, wherein the image information set includes at least one image information item consisting of a picture of the target bridge; Its characteristics are: Acquire a determination area of ​​the image information item to obtain a determination target item, where the determination target item is used to represent a crack determination area of ​​the target bridge image; Based on the determination target item, the original image pixel value of the determination target item is obtained to obtain the initial pixel item; based on the image information item, the channel mean of the image information item is obtained; a basic illumination compensation formula is created as the first stage compensation to obtain the first compensation item; Acquire the time series information of the image information item to obtain a time series information set corresponding to the image information set; based on the time series information set, obtain the illumination incident angle information when the image information item interacts with the target bridge in a state corresponding to the determination target item to obtain the incident information item; simultaneously obtain the distance between the pixel coordinates in the image information item and the sky segmentation area and the maximum sky distance value in the image information item; and create a solar incident angle shadow compensation formula as the second compensation stage; Based on the first-stage compensation and the second-stage compensation, the initial pixel items are segmentedly compensated, and the compensated pixel values ​​of the target items are obtained to obtain the processed pixel items, thereby achieving compensation for the comprehensive basic illumination and the shadow of the sun's incidence angle; Obtaining identification features of processed pixel items, obtaining identification feature items, and determining the identification feature items, thereby achieving identification and determination of bridge crack images; The method for creating the basic illumination compensation formula includes: Based on the image information item, split the image information item according to the number of channels to obtain at least two split channel items; Get the global average of the split channel items respectively to get the channel mean item; Create a basic lighting compensation formula: ； in is the image pixel value after basic lighting compensation, is the initial pixel term, To calibrate the grayscale value, is the channel mean term; Obtaining first-stage pixel compensation data for the initial pixel item based on a basic illumination compensation formula to obtain a first compensation item; The method for creating the solar incident angle shadow compensation formula includes: Identify and determine the sky pixel area in the image information item to obtain a divided area item; Obtaining the maximum value information of the pixel distance division area item in the image information item to obtain the maximum distance value; Obtain the distance information between the determination target item and the divided area item to obtain the determination distance value; Create the shadow compensation formula for the sun incidence angle: ； in is the image pixel value after the sun’s incident angle shadow compensation, is the image pixel value after basic lighting compensation, is the intensity attenuation coefficient, is the incident information item, To determine the distance value, is the maximum distance value.

2. The bridge crack image recognition method based on drone and AI algorithm according to claim 1 is characterized by: The method for obtaining the channel mean term includes: Based on the split channel item, create the channel mean acquisition formula: ； in is the total number of pixels, The value of the split channel item with coordinates (x, y) in the image information item.

3. The bridge crack image recognition method based on drone and AI algorithm according to claim 1 is characterized by: The method for obtaining the image information set includes: Set the acquisition module, which is a UAV module, to obtain the overall information of the target bridge and obtain the target information item; An acquisition path is created based on the target information item, and the acquisition module is path-limited based on the acquisition path. The height of the acquisition module is limited based on the initial height information of the acquisition module and the target bridge. The drone module is equipped with a camera component, and image information data of the target bridge is acquired based on the camera component, thereby obtaining an image information set.

4. The bridge crack image recognition method based on drone and AI algorithm according to claim 3 is characterized by: The acquisition module further includes a photosensitive component, and the method for acquiring the incident information item includes: Obtaining the image information item where the determination target item is located, and obtaining the target information item; Based on the time series information set, obtaining a time series information item corresponding to the target information item to obtain a target time series item; Based on the target timing item, the acquisition module obtains the incident angle information of the light through the photosensitive component to obtain the incident information item.

5. The bridge crack image recognition method based on drone and AI algorithm according to claim 1 is characterized by: The method for determining the identification feature item includes: Based on the processed pixel items, color features, shape features, and range features are respectively obtained to obtain color recognition items, shape recognition items, and range recognition items, and the color recognition items, shape recognition items, and range recognition items are combined to obtain recognition feature items; An impact information library is created, which includes background impact data and impurity impact data. Based on the comparison results between the impact information library and the identification feature items, it is determined whether the identification feature items are in the impact information library, thereby realizing the determination of the identification feature items.

6. The bridge crack image recognition method based on drone and AI algorithm according to claim 5 is characterized by: The method for creating the impact information database includes: Obtaining initial state data of the target bridge, the initial state data including the background color and background texture of the bridge; Obtaining initial impurity data of the target bridge, the initial impurity data including residual impurity color, residual impurity size, and residual impurity shape; An impact information library is created based on the initial state data and the initial impurity data as data filling.

Images