Bridge detection method and system based on unmanned aerial vehicle

Abstract

The invention discloses a bridge detection method and system based on an unmanned aerial vehicle, and belongs to the technical field of bridge detection. The objective of the invention is to improve the quality and recognition accuracy of bridge detection images. The method comprises the steps of constructing an image definition scoring function of a single-frame image acquired by an unmanned aerial vehicle; constructing a speed-attitude fuzzy coupling risk scoring function of a single-frame image acquired by the unmanned aerial vehicle; constructing an image transmission time domain continuity scoring function of continuous images acquired by the unmanned aerial vehicle; constructing an image comprehensive value scoring function; constructing an image main direction deviation vector; in combination with the image main direction deviation vector and the image comprehensive value scoring function, constructing an identification suitability weight factor of the image; and based on the image main direction deviation vector and the identification suitability weight factor, designing a dynamic optimization strategy with a suggested flight speed, a suggested attitude angular speed and a suggested flight height of the unmanned aerial vehicle. According to the invention, the quality and recognition accuracy of the bridge detection image are improved to a great extent.

Description

Technical Field

[0001] This invention belongs to the field of bridge inspection technology, specifically relating to a bridge inspection method and system based on unmanned aerial vehicles (UAVs). Background Technology

[0002] With the continuous increase in the number of highway and railway bridges, regular inspection and condition assessment have become crucial guarantees for the safe operation and maintenance of infrastructure. Traditional bridge inspection relies on manual close-range inspections and high-altitude operations, which pose significant safety risks, are inefficient, and have limited coverage. In recent years, drones, with their advantages of flexibility, efficiency, and low cost, have gradually become an emerging tool for bridge inspection. They can perform close-range photography and video inspections across rivers, high piers, and in high-altitude environments, significantly improving operational efficiency.

[0003] While drones demonstrate significant application potential in many scenarios, their application in actual bridge inspection tasks still faces numerous challenges. Specifically, these include image blurring due to flight vibrations, defocusing caused by perspective deviations, and the impact of discontinuous image frame sequences on structural recognition results. Furthermore, drones lack image quality feedback during flight, making it difficult to re-capture images of poor quality during inspection. Therefore, it is necessary to construct a flight control adjustment mechanism specifically designed for image recognition targets. This mechanism should dynamically integrate image quality, recognition value, and flight control to guide the drone in adjusting its speed, altitude, and attitude, thereby acquiring more stable, clearer, and more easily identifiable bridge images. This will provide fundamental support for bridge defect detection and assessment. Summary of the Invention

[0004] The problem this invention aims to solve is to improve the quality and recognition accuracy of bridge inspection images, and proposes a bridge inspection method and system based on unmanned aerial vehicles (UAVs).

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A bridge inspection method based on unmanned aerial vehicles (UAVs) includes the following steps:

[0007] S1. Construct an image sharpness scoring function for a single frame of image captured by a drone;

[0008] S2. Construct a velocity-attitude ambiguity coupling risk scoring function for single-frame images acquired by a UAV;

[0009] S3. Construct a temporal continuity scoring function for image transmission of continuous images acquired by UAVs;

[0010] S4. Based on the image sharpness scoring function obtained in step S1, the velocity-attitude fuzzy coupling risk scoring function obtained in step S2, and the image transmission temporal continuity scoring function obtained in step S3, construct the comprehensive image value scoring function.

[0011] S5. Construct the image principal direction deviation vector;

[0012] S6. Combine the image principal direction deviation vector and the image comprehensive value scoring function to construct the image recognition suitability weight factor;

[0013] S7. Based on the image main direction deviation vector obtained in step S5 and the recognition suitability weight factor obtained in step S6, design a dynamic optimization strategy with the suggested flight speed, suggested attitude angular velocity and suggested flight altitude of the UAV.

[0014] Furthermore, in step S1, by fusing the proportion of high-frequency energy and spatial edge geometric information in the image frequency domain signal, an image sharpness scoring function is obtained, the expression of which is:

[0015]

[0016] in, Rate the image sharpness; These are Fourier weighting coefficients, determined by expert experience; The residual inverse value weighting coefficient is determined by expert experience; The critical threshold is determined by expert experience; Represents the high-frequency range; u is the two-dimensional frequency coordinate of the image in the frequency domain; I t Image frames acquired at time t For I t Fourier transform; For I t The inverse value of the local edge linear fitting residual is obtained by calculating the Sobel gradient.

[0017] Furthermore, the velocity-attitude ambiguity coupling risk scoring function constructed in step S2 for acquiring a single frame image by the UAV evaluates the impact of UAV translational velocity and yaw attitude changes on imaging. The expression is:

[0018]

[0019] in, Risk value of velocity-attitude ambiguity coupling; The speed-sensitive weighting coefficient is determined by expert experience. The angular velocity fuzziness sensitivity weighting coefficient is determined by expert experience; The XY plane velocity components are obtained from the UAV monitoring platform; The angular velocity of the drone around the Z-axis is obtained by the drone monitoring platform; This is a speed reference value, determined by expert experience; This is a reference value for angular velocity, determined by expert experience.

[0020] Furthermore, in step S3, the structural similarity between two frames and the amplitude of the inter-frame time interval change are combined to construct an image transmission temporal continuity scoring function, the expression of which is:

[0021]

[0022] in, Scoring the temporal continuity of image transmission; The SSIM structural similarity is obtained by calling the SSIM calculation module; The time interval between the current frame and the previous frame is obtained from the camera parameters; , These are the mean and standard deviation of the Slide window, respectively, calculated from the local cache. For a fixed constant, the dimensions are the same as... Maintain consistency, determined by expert experience.

[0023] Furthermore, the image comprehensive value scoring function expression constructed in step S4 is as follows:

[0024]

[0025] in, Score the overall value of the image.

[0026] Furthermore, in step S5, to evaluate the degree of matching between the current camera angle and the structural layout, a principal direction deviation vector for the image is constructed, expressed as:

[0027]

[0028] in, The image's principal direction deviation vector; The orientation unit vector of the camera lens in the panorama is obtained using a drone monitoring platform; The average value for the direction of the straight line in the image is obtained from the drone monitoring platform.

[0029] Furthermore, in step S6, the square of the image's principal direction deviation vector is used as a value penalty term. An exponential function is introduced for non-linear scaling, and this is combined with the image's comprehensive value score to calculate the image's recognition suitability weight factor, expressed as:

[0030]

[0031] in, This is a weighting factor for the suitability of the image for recognition. The main directional deviation weighting factor is determined by expert experience.

[0032] Furthermore, the specific implementation method of step S7 includes the following steps:

[0033] S7.1. Using a linear response model, design the recommended flight speed for the UAV, expressed as:

[0034]

[0035] in, Recommended flight speed for drones; The current speed of the drone is obtained from the drone monitoring platform. The directional gradient of the image recognition suitability weight factor; The directional gradient weighting factor is determined by expert experience;

[0036] S7.2. Based on the image principal direction deviation vector, the recommended attitude angular velocity of the UAV is designed using a proportional control method, expressed as:

[0037]

[0038] in, Recommended attitude angular velocity for the UAV; The angular velocity weighting factor is determined by expert experience;

[0039] S7.3. The overall recognition status is reflected by the image recognition suitability weight factor, and then converted into a proportional adjustment coefficient to design the recommended flight altitude of the UAV. The expression is:

[0040]

[0041] in, Recommended flight altitude for drones; This is the current flight altitude of the drone, obtained from the drone monitoring platform. This is the flight altitude adjustment factor.

[0042] A system for a bridge inspection method based on unmanned aerial vehicles (UAVs) includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed, it implements the steps of the UAV-based bridge inspection method.

[0043] The beneficial effects of this invention are:

[0044] The present invention discloses a bridge detection method based on unmanned aerial vehicles (UAVs). First, the sharpness and motion blur risk of a single frame image are evaluated. Then, based on this, the differences between inter-frame continuity and the main direction of the components are extracted and fused into a unified recognition value score. Next, based on the scoring function, a dynamic optimization strategy for flight speed, attitude angular velocity, and flight altitude is designed, so that the UAV can determine flight parameters according to environmental conditions and image feedback, thereby improving the effectiveness of image acquisition and the reliability of bridge risk detection.

[0045] The bridge detection method based on UAV described in this invention integrates multiple factors such as image clarity, stability, and structural angles into a unified recognition value score for flight control. Compared with traditional fixed strategies, it has the advantages of strong adaptability, complete coverage, and dynamic optimization, which can greatly improve the quality and recognition accuracy of bridge detection images. Attached Figure Description

[0046] Figure 1 This is a flowchart of a bridge inspection method based on unmanned aerial vehicles (UAVs) according to the present invention.

[0047] Figure 2 This is a comparison chart of the structural similarity index (SSIM) and confidence level before and after the adjustment of this invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention; that is, the described specific embodiments are merely a part of the embodiments of the invention, and not all of them. The components of the specific embodiments of the invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations, and the invention may also have other embodiments.

[0049] Therefore, the following detailed description of specific embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected specific embodiments of the invention. All other specific embodiments obtained by those skilled in the art based on these specific embodiments without inventive effort are within the scope of protection of this invention.

[0050] To further understand the invention's content, features, and effects, the following specific embodiments are provided, along with accompanying drawings. Figure 1 and attached Figure 2 Detailed explanation is as follows:

[0051] Example 1:

[0052] In the complex and ever-changing task of bridge inspection, high-quality, highly recognizable, and stable images are crucial for the effectiveness of UAV inspection. However, due to factors such as UAV flight status, target angle errors, and image transmission jitter, problems such as image blurring, perspective deviation, and loss of continuity often occur, severely impacting the identification and analysis of bridge structural defects. Considering the above, this embodiment first assesses the sharpness and motion blur risk of a single frame image. Then, based on this, it extracts the differences in inter-frame continuity and the main direction of components, fusing them into a unified recognition value score. Next, based on this scoring function, a dynamic optimization strategy for flight speed, attitude angular velocity, and flight altitude is designed, enabling the UAV to determine flight parameters based on environmental conditions and image feedback, improving the effectiveness of image acquisition and the reliability of bridge risk detection.

[0053] A bridge inspection method based on unmanned aerial vehicles (UAVs) includes the following steps:

[0054] S1. Construct an image sharpness scoring function for a single frame of image captured by a drone;

[0055] Furthermore, in step S1, by fusing the proportion of high-frequency energy and spatial edge geometric information in the image frequency domain signal, an image sharpness scoring function is obtained, the expression of which is:

[0056]

[0057] in, Rate the image sharpness; These are Fourier weighting coefficients, determined by expert experience; The residual inverse value weighting coefficient is determined by expert experience; The critical threshold is determined by expert experience; Represents the high-frequency range; u is the two-dimensional frequency coordinate of the image in the frequency domain, used to traverse each frequency component in the frequency domain, and is an index variable in the calculation process; I t Image frames acquired at time t For I t Fourier transform; For I t The inverse value of the local edge linear fitting residual is obtained by calculating the Sobel gradient.

[0058] Bridge structures have minute cracks and linear features like stay cables, making it difficult to fully represent image quality using only traditional frequency domain energy or single edge responses. This function fuses the proportion of high-frequency energy in the image's frequency domain signal with spatial edge geometry information, considering both crack details and structural contours. High-frequency energy measures the overall sharpness of the image, while edge residuals reflect the sharpness of the geometric contours. Weighting both factors results in stable performance under varying lighting conditions and material reflections, allowing for a more accurate evaluation of the image's ability to preserve detail in accordance with bridge defect detection requirements.

[0059] S2. Construct a velocity-attitude ambiguity coupling risk scoring function for single-frame images acquired by a UAV;

[0060] Furthermore, the velocity-attitude ambiguity coupling risk scoring function constructed in step S2 for acquiring a single frame image by the UAV evaluates the impact of UAV translational velocity and yaw attitude changes on imaging. The expression is:

[0061]

[0062] in, Risk value of velocity-attitude ambiguity coupling; The speed-sensitive weighting coefficient is determined by expert experience. The angular velocity fuzziness sensitivity weighting coefficient is determined by expert experience; The XY plane velocity components are obtained from the UAV monitoring platform; The angular velocity of the drone around the Z-axis is obtained by the drone monitoring platform; This is a speed reference value, determined by expert experience; This is a reference value for angular velocity, determined by expert experience.

[0063] When a drone performs bridge inspection, its motion along the flight path can easily affect image sharpness. During low-altitude inspection, airflow disturbances or rapid yaw can blur the image. This function evaluates the impact of drone translational velocity and yaw attitude changes on imaging. It introduces a nonlinear fourth-power term to improve the representation of gyroscopic disturbance signals in the model and avoids local extremum leakage through its exponential form. This form can more accurately characterize the impact of stability on image sharpness during actual flight.

[0064] S3. Construct a temporal continuity scoring function for image transmission of continuous images acquired by UAVs;

[0065] Furthermore, in step S3, the structural similarity between two frames and the amplitude of the inter-frame time interval change are combined to construct an image transmission temporal continuity scoring function, the expression of which is:

[0066]

[0067] in, Scoring the temporal continuity of image transmission; The SSIM structural similarity is obtained by calling the SSIM calculation module; The time interval between the current frame and the previous frame is obtained from the camera parameters; , These are the mean and standard deviation of the Slide window, respectively, calculated from the local cache. For a fixed constant, the dimensions are the same as... Maintain consistency, determined by expert experience.

[0068] In bridge inspection, continuous image sequences are fundamental for judging the development trend of defects and the splicing effect. To ensure the continuity of the structural recognition process, this scoring function combines the structural similarity between two frames with the amplitude of the inter-frame time interval variation, thus comprehensively reflecting the stability of the image stream. Here, image structural similarity can summarize the changes in image content, while the fluctuation of the time interval reflects whether the image acquisition is uniform and stable. By multiplying the similarity by a time adjustment term to obtain a stability score, the location of risks such as image frame breaks and stuttering during UAV flight can be determined.

[0069] S4. Based on the image sharpness scoring function obtained in step S1, the velocity-attitude fuzzy coupling risk scoring function obtained in step S2, and the image transmission temporal continuity scoring function obtained in step S3, construct the comprehensive image value scoring function.

[0070] Furthermore, the image comprehensive value scoring function expression constructed in step S4 is as follows:

[0071]

[0072] in, Score the overall value of the image.

[0073] Whether an image is meaningful for recognition depends not only on the sharpness of a single frame, but also on the device's motion state and the continuity of the entire frame sequence. This function combines sharpness score, continuity score, and fuzzy coupling risk value to calculate the comprehensive value score of the current frame image, thus showing whether the image can be used for accurate information recognition. The denominator of the score introduces the velocity-attitude fuzzy coupling risk value, which plays a role in attenuation. In this way, when a serious fuzziness is observed, even if the image content is clear or continuous, its overall recognition value will drop significantly, which is more in line with the requirements of both quality and stability in actual bridge inspection applications.

[0074] S5. Construct the image principal direction deviation vector;

[0075] Furthermore, in step S5, to evaluate the degree of matching between the current camera angle and the structural layout, a principal direction deviation vector for the image is constructed, expressed as:

[0076]

[0077] in, The image's principal direction deviation vector; The orientation unit vector of the camera lens in the panorama is obtained using a drone monitoring platform; The average value for the direction of the straight line in the image is obtained from the drone monitoring platform.

[0078] In bridge inspection, components such as cables, columns, and railings generally exhibit stable orientation. Post-processing algorithms often align with these structural orientations to improve recognition accuracy. This model, to assess the match between the current camera angle and the structural layout, calculates a deviation vector consisting of the difference between the camera's optical axis and the main edge directions extracted from the image. This vector guides the drone to adjust towards a more suitable viewpoint. Furthermore, this deviation has a crucial impact on the stability of subsequent component detection and texture matching, and it also aligns with the structure's orientation during attitude adjustment.

[0079] S6. Combine the image principal direction deviation vector and the image comprehensive value scoring function to construct the image recognition suitability weight factor;

[0080] Furthermore, in step S6, the square of the image's principal direction deviation vector is used as a value penalty term. An exponential function is introduced for non-linear scaling, and this is combined with the image's comprehensive value score to calculate the image's recognition suitability weight factor, expressed as:

[0081]

[0082] in, This is a weighting factor for the suitability of the image for recognition. The main directional deviation weighting factor is determined by expert experience.

[0083] For an image to be recognized accurately, it must be clear and stable, and the viewing angle must be appropriate. If the angle between the camera's viewing angle and the bridge structure is large, even if the image resolution is high and the content is complete, it will be difficult to correctly identify the structure and make the quantification accurate. This function uses the square of the image's main direction deviation as a value penalty term, introduces an exponential function for non-linear scaling, and combines it with the previously obtained comprehensive image value to calculate the weight of the overall image performance evaluation.

[0084] S7. Based on the image main direction deviation vector obtained in step S5 and the recognition suitability weight factor obtained in step S6, design a dynamic optimization strategy with the suggested flight speed, suggested attitude angular velocity and suggested flight altitude of the UAV.

[0085] Furthermore, the specific implementation method of step S7 includes the following steps:

[0086] S7.1. Using a linear response model, design the recommended flight speed for the UAV, expressed as:

[0087]

[0088] in, Recommended flight speed for drones; The current speed of the drone is obtained from the drone monitoring platform. The directional gradient of the image recognition suitability weight factor; The directional gradient weighting factor is determined by expert experience;

[0089] In bridge detection, the flight speed needs to be able to adapt to the difficulty of target area recognition. If the image has low recognition, it needs to be slowed down to slow down the shooting pace. In areas with high recognition evaluation, the speed can be increased. Moreover, the gradient magnitude will affect the current speed. With the help of the linear response model, a suggested speed value can be obtained. The speed can be guided by the spatial distribution of visual quality to improve the detection coverage and reduce the probability of omission.

[0090] S7.2. Based on the image principal direction deviation vector, the recommended attitude angular velocity of the UAV is designed using a proportional control method, expressed as:

[0091]

[0092] in, Recommended attitude angular velocity for the UAV; The angular velocity weighting factor is determined by expert experience;

[0093] The camera's alignment with the bridge structure is fundamental to improving recognition performance and structural accuracy measurement. However, manual adjustments are often limited by attitude feedback delays and error accumulation. The proposed model uses the image's principal orientation deviation as input and employs a proportional control method to drive the UAV's attitude angular velocity, gradually guiding the UAV closer to the principal orientation.

[0094] S7.3. The overall recognition status is reflected by the image recognition suitability weight factor, and then converted into a proportional adjustment coefficient to design the recommended flight altitude of the UAV. The expression is:

[0095]

[0096] in, Recommended flight altitude for drones; This is the current flight altitude of the drone, obtained from the drone monitoring platform. This is the flight altitude adjustment factor.

[0097] Image resolution is closely related to the spatial distance to the target. If the flight distance is too large, even if the image function parameters are set correctly, it will be difficult to obtain accurate details of bridge defects. This function uses an image recognition suitability weight factor to reflect the overall recognition status and converts it into a proportional adjustment coefficient to control the flight altitude. In cases where the recognition result weakens or becomes blurred / enlarged, the altitude will be appropriately reduced to shorten the shooting distance, improve image quality, and compensate for the disadvantages caused by complex scene conditions.

[0098] The following is an example of its application: A drone equipped with a fixed-focus camera flies at an altitude of 18.2m, with a horizontal flight speed of 2.40 m / s and a yaw rate of 14.2° / s. The image frame interval is 41ms, the local SSIM is 0.54, and the target structure confidence score is 0.42 (bridge crack area). This frame is currently determined to be a quality degradation frame. Using this method, the drone's horizontal flight speed, yaw rate, and flight altitude are adjusted to 1.9m / s, 11.6° / s, and 16.2m, respectively, resulting in an increase in the corresponding SSIM and target structure confidence scores to 0.69 and 0.63, respectively.

[0099] Example 2:

[0100] A system for a bridge inspection method based on unmanned aerial vehicles (UAVs) includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed, it implements the steps of the bridge inspection method based on UAVs as described in Embodiment 1.

[0101] It should be noted that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0102] Although this application has been described above with reference to specific embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of this application. In particular, as long as there is no structural conflict, the features in the specific embodiments disclosed in this application can be combined with each other in any way. The lack of an exhaustive description of these combinations in this specification is merely for the sake of brevity and resource conservation. Therefore, this application is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A bridge inspection method based on unmanned aerial vehicles (UAVs), characterized in that, Includes the following steps: S1. Construct an image sharpness scoring function for a single frame of image captured by a drone; S2. Construct a velocity-attitude ambiguity coupling risk scoring function for single-frame images acquired by a UAV; S3. Construct a temporal continuity scoring function for image transmission of continuous images acquired by UAVs; S4. Based on the image sharpness scoring function obtained in step S1, the velocity-attitude fuzzy coupling risk scoring function obtained in step S2, and the image transmission temporal continuity scoring function obtained in step S3, construct the comprehensive image value scoring function. S5. Construct the image principal direction deviation vector; S6. Combine the image principal direction deviation vector and the image comprehensive value scoring function to construct the image recognition suitability weight factor; S7. Based on the image main direction deviation vector obtained in step S5 and the recognition suitability weight factor obtained in step S6, design a dynamic optimization strategy with the suggested flight speed, suggested attitude angular velocity and suggested flight altitude of the UAV.

2. The bridge inspection method based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, In step S1, the image sharpness scoring function is obtained by fusing the high-frequency energy ratio and spatial edge geometric information in the image frequency domain signal. The expression is as follows: ； in, Rate the image sharpness; These are Fourier weighting coefficients, determined by expert experience; The residual inverse value weighting coefficient is determined by expert experience; The critical threshold is determined by expert experience; Represents the high-frequency range; u is the two-dimensional frequency coordinate of the image in the frequency domain; I t Image frames acquired at time t For I t Fourier transform; For I t The inverse value of the local edge linear fitting residual is obtained by calculating the Sobel gradient.

3. The bridge inspection method based on unmanned aerial vehicles (UAVs) according to claim 2, characterized in that, The velocity-attitude ambiguity coupling risk scoring function for single-frame images acquired by the UAV, constructed in step S2, evaluates the impact of UAV translational velocity and yaw attitude changes on imaging. The expression is: ； in, Risk value of velocity-attitude ambiguity coupling; The speed-sensitive weighting coefficient is determined by expert experience. The angular velocity fuzziness sensitivity weighting coefficient is determined by expert experience; The XY plane velocity components are obtained from the UAV monitoring platform; The angular velocity of the drone around the Z-axis is obtained by the drone monitoring platform; This is a speed reference value, determined by expert experience; This is a reference value for angular velocity, determined by expert experience.

4. The bridge inspection method based on unmanned aerial vehicles (UAVs) according to claim 3, characterized in that, In step S3, the structural similarity between two frames and the magnitude of the inter-frame time interval change are combined to construct an image transmission temporal continuity scoring function, the expression of which is: ； in, Scoring the temporal continuity of image transmission; The SSIM structural similarity is obtained by calling the SSIM calculation module; The time interval between the current frame and the previous frame is obtained from the camera parameters; , These are the mean and standard deviation of the Slide window, respectively, calculated from the local cache. For a fixed constant, the dimensions are the same as... Maintain consistency, determined by expert experience.

5. A bridge inspection method based on unmanned aerial vehicles (UAVs) according to claim 4, characterized in that, The expression for the comprehensive image value scoring function constructed in step S4 is as follows: ； in, Score the overall value of the image.

6. The bridge inspection method based on unmanned aerial vehicles (UAVs) according to claim 5, characterized in that, Step S5: To evaluate the degree of matching between the current camera angle and the structural layout, construct the image principal direction deviation vector, expressed as: ； in, The image's principal direction deviation vector; The orientation unit vector of the camera lens in the panorama is obtained using a drone monitoring platform; The average value for the direction of the straight line in the image is obtained from the drone monitoring platform.

7. A bridge inspection method based on unmanned aerial vehicles (UAVs) according to claim 6, characterized in that, Step S6 uses the square of the image's principal direction deviation vector as a value penalty term, introduces an exponential function for non-linear scaling, and combines it with the image's comprehensive value score to calculate the image's recognition suitability weight factor, expressed as: ； in, This is a weighting factor for the suitability of the image for recognition. The main directional deviation weighting factor is determined by expert experience.

8. A bridge inspection method based on unmanned aerial vehicles (UAVs) according to claim 7, characterized in that, The specific implementation method of step S7 includes the following steps: S7.

1. Using a linear response model, design the recommended flight speed for the UAV, expressed as: ； in, Recommended flight speed for drones; The current speed of the drone is obtained from the drone monitoring platform. The directional gradient of the image recognition suitability weight factor; The directional gradient weighting factor is determined by expert experience; S7.

2. Based on the image principal direction deviation vector, the recommended attitude angular velocity of the UAV is designed using a proportional control method, expressed as: ； in, Recommended attitude angular velocity for the UAV; The angular velocity weighting factor is determined by expert experience; S7.

3. The overall recognition status is reflected by the image recognition suitability weight factor, and then converted into a proportional adjustment coefficient to design the recommended flight altitude of the UAV. The expression is: ； in, Recommended flight altitude for drones; This is the current flight altitude of the drone, obtained from the drone monitoring platform. This is the flight altitude adjustment factor.

9. A system for bridge inspection based on unmanned aerial vehicles (UAVs), characterized in that, It includes a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed, implements the steps of a UAV-based bridge detection method as described in any one of claims 1-8.

Images