An Adaptive Control Method and Device for UAV Mapping Based on Image Analysis

Abstract

An embodiment of the present invention provides a method and device for adaptive control of UAV mapping based on image analysis, belonging to the technical field of laser mapping. The method includes: controlling a UAV carrying a laser scanning and ranging module to fly to a target area and hover stably, and obtaining three-dimensional point cloud data of the target; analyzing the target features based on the obtained three-dimensional point cloud data to determine at least one target measurement point that needs to be measured for fine displacement; obtaining image data of the target measurement point, and judging whether the measurement point meets the laser interference measurement condition based on image processing; when the image judgment meets the measurement condition, collecting the laser interference signal of the target measurement point, and calculating the vibration or displacement data of the measurement point; performing fusion processing on the three-dimensional point cloud data and the vibration / displacement data of the measurement point, and outputting the geometric size and displacement measurement result of the target. The solution of the present invention effectively improves the stability and pertinence of micro displacement measurement in UAV mapping tasks.

Description

Technical Field

[0001] The present invention relates to the technical field of laser mapping, and particularly to an unmanned aerial vehicle (UAV) mapping adaptive control method based on image analysis and an unmanned aerial vehicle mapping adaptive control device based on image analysis. Background Art

[0002] In the fields of precision measurement and engineering structure detection, non-contact measurement technologies are widely used to obtain physical quantities such as length, displacement, and vibration due to their advantages of high efficiency and strong safety. With the gradual maturity of the flight control technology of the unmanned aerial vehicle (UAV) platform, carrying measurement devices to conduct mapping and structure monitoring of spatial scenes has become a current research and application hotspot. Among them, laser scanning (such as lidar) and photogrammetry technologies are widely used to obtain the three-dimensional structure data of the scene, realizing geometric reconstruction and deformation monitoring of ground objects.

[0003] However, existing solutions mostly focus on the acquisition of macroscopic structure contours, and there are still technical bottlenecks in the measurement of micro displacements, vibrations, or dynamic response characteristics on the structure surface. For example, large structures such as buildings, bridges, and towers may undergo small periodic or sudden displacements during operation. Existing measurement methods often rely on pre-installing contact sensors at the measurement point positions, or arranging reflectors and targets to cooperate with total stations or photographic devices for observation. Such methods not only require physical intervention on the structure body, are complex to operate and costly, but also have certain safety hazards when facing high altitudes or dangerous areas.

[0004] In addition, even if some existing devices attempt to combine UAVs with high-precision measurement equipment, such as using image recognition or laser scanning for deformation measurement, they still generally rely on arranging obvious feature objects or manual preprocessing steps, and it is difficult to achieve real-time measurement. Especially in scenarios where it is necessary to simultaneously obtain the overall geometric information and microscopic dynamic response characteristics of the target object, the existing technology still lacks a solution that takes into account both measurement accuracy and deployment flexibility.

[0005] Therefore, in the current technical path of non-contact high-precision measurement assisted by UAVs, there are still two key problems that need to be solved urgently: First, how to accurately obtain micro displacements without additional sensors or markers; second, how to achieve data linkage and unified expression between macroscopic three-dimensional mapping and microscopic dynamic monitoring. These problems restrict the further promotion and application of measurement devices in structural safety assessment and on-site emergency detection. Summary of the Invention

[0006] The purpose of the embodiments of the present invention is to provide an unmanned aerial vehicle mapping adaptive control method and device based on image analysis, so as to at least solve the problem in the prior art that the judgment of measurement point conditions is not accurate enough, affecting the stability of non-contact high-precision measurement.

[0007] To achieve the above object, a first aspect of the present invention provides an adaptive control method for UAV mapping based on image analysis. The method includes: controlling a UAV equipped with a laser scanning ranging module to fly to a target area and hover stably, performing laser scanning mapping on the target area to obtain three-dimensional point cloud data of the target area; analyzing target features based on the obtained three-dimensional point cloud data to determine at least one target measurement point that requires fine displacement measurement; obtaining image data of the target measurement point, and judging whether the measurement point meets the laser interference measurement condition based on image processing; when the image judgment meets the measurement condition, collecting the laser interference signal of the target measurement point and calculating the vibration or displacement data of the measurement point; performing fusion processing on the three-dimensional point cloud data and the vibration / displacement data of the measurement point, and outputting the geometric size and displacement measurement result of the target.

[0008] Optionally, controlling a UAV equipped with a laser scanning ranging module to fly to a target area and hover stably includes: collecting spatial distance data and obstacle boundary information of the surrounding environment, inputting the data into a path planning algorithm, and generating an executable path point sequence according to the three-dimensional coordinates of the target area, flight altitude constraints, and attitude adjustment strategies; during flight, comparing and adjusting the flight attitude and position in real time according to the current flight state and path points to avoid obstacles and enter the mapping hover point of the target area.

[0009] Optionally, performing laser scanning mapping on the target area includes: performing an initial low-density scan to obtain a point cloud skeleton of the target area; analyzing the edge gradient, local curvature, and geometric complexity in the point cloud, and marking the area where the significant degree of geometric change meets the expectation as the densification scan area; performing a high-density rescan on the densification scan area according to a preset scan resolution to obtain a three-dimensional point cloud data set with a greater density.

[0010] Optionally, obtaining image data of the target measurement point and judging whether the measurement point meets the laser interference measurement condition based on image processing includes: collecting visible light images and infrared images of the measurement point area, and performing pixel-level alignment processing on the images; identifying the area where the measurement point is located in the registered image, and analyzing the occlusion contour, infrared reflection intensity, and texture continuity index of the corresponding area; further collecting polarization images of the measurement point area, and analyzing the polarization rate and polarization angle distribution of the measurement point; judging the surface scattering characteristics of the measurement point according to the polarization characteristics, and combining the occlusion contour, infrared reflection intensity, and texture continuity index to determine whether the measurement point meets the laser interference measurement requirements.

[0011] Optionally, calculating the vibration or displacement data of the measurement point includes: emitting a coherent laser beam to a target measurement point and receiving a reflected signal to form an interference pattern; performing analog-to-digital conversion on the signal of the interference intensity varying with time and inputting it into a phase tracking algorithm to extract a continuous phase sequence; calculating the displacement change curve of the measurement point according to the relationship between the phase change and the laser wavelength; if there is periodic displacement within the measurement period, extracting vibration frequency and amplitude parameters through frequency domain analysis; establishing a motion compensation model based on the flight attitude data to eliminate non-target motion components in the phase change and obtain the actual displacement or vibration response.

[0012] Optionally, when there are multiple measurement points that meet the measurement conditions, the method further includes: calculating a priority ranking based on the spatial position of the measurement point, the structural risk level, and the lighting conditions; sequentially adjusting the position and attitude of the unmanned aerial vehicle according to the priority to make the measurement point located in the direction of the laser measurement optical axis; before each measurement, making the measurement beam perpendicular to the normal of the target surface through attitude adjustment and fine adjustment of the pan-tilt direction.

[0013] Optionally, performing fusion processing on the three-dimensional point cloud data and the vibration / displacement data of the measurement point includes: establishing a multi-field data structure containing three-dimensional coordinates, displacement values, vibration frequencies, and measurement times for each measured target measurement point; embedding the multi-field data structure into the original index position of the point cloud data to achieve attribute-level expansion; generating a composite model containing the spatial shape and measurement response.

[0014] Optionally, outputting the geometric dimensions and displacement measurement results of the target includes: generating a structured data file, where the structured data file contains a three-dimensional point cloud model, displacement / vibration parameters of key measurement points, and measurement element information; assigning a unique identifier to each key measurement point and attaching the corresponding measurement time, numerical result, and confidence level; exporting the measurement result in a parsable standard format.

[0015] In a second aspect of the present invention, there is provided an unmanned aerial vehicle mapping adaptive control device based on image analysis. The device includes: an acquisition unit for controlling the unmanned aerial vehicle equipped with a laser scanning and ranging module to fly to a target area and hover stably, performing laser scanning mapping on the target area to obtain three-dimensional point cloud data of the target area; a measurement point determination unit for analyzing target features based on the obtained three-dimensional point cloud data to determine at least one target measurement point that needs to perform fine displacement measurement; a judgment unit for acquiring image data of the target measurement point and judging whether the measurement point meets the laser interference measurement conditions based on image processing; a processing unit for collecting the laser interference signal of the target measurement point and calculating the vibration or displacement data of the measurement point when the image judgment meets the measurement conditions; an output unit for performing fusion processing on the three-dimensional point cloud data and the vibration / displacement data of the measurement point and outputting the geometric dimensions and displacement measurement results of the target.

[0016] On the other hand, the present invention provides a computer-readable storage medium, on which instructions are stored, and when the instructions run on a computer, the computer is caused to execute the above-mentioned adaptive control method for UAV mapping based on image analysis.

[0017] Through the above technical solution, the solution of the present invention obtains three-dimensional point cloud data of the target area through laser scanning, realizes the macroscopic geometric modeling of the target structure, and combines image processing technology to judge the measurability of key measurement points, so as to adaptively select areas suitable for carrying out fine interference measurement. Based on the image judgment result, laser interference measurement is only performed on the measurement points that meet the conditions, the vibration or displacement data is extracted, and the data is fused with the three-dimensional structure information. While realizing non-contact acquisition of the geometric dimensions of the target object, this method can also accurately extract the dynamic response characteristics of key parts, effectively improving the stability and pertinence of micro-displacement measurement in UAV mapping tasks.

[0018] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent specific embodiment part. Description of the Drawings

[0019] The drawings are used to provide a further understanding of the embodiments of the present invention, and constitute a part of the specification. They are used together with the following specific embodiments to explain the embodiments of the present invention, but do not constitute a limitation to the embodiments of the present invention. In the drawings:

[0020] Figure 1 is a flowchart of the steps of an adaptive control method for UAV mapping based on image analysis provided by an embodiment of the present invention;

[0021] Figure 2 is a structural diagram of an adaptive control device for UAV mapping based on image analysis provided by an embodiment of the present invention. Specific Embodiments

[0022] The following will describe in detail the specific embodiments of the present invention with reference to the drawings. It should be understood that the specific embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0023] Figure 1 is a flowchart of a method for an adaptive control method for UAV mapping based on image analysis provided by an embodiment of the present invention. As Figure 1 shown, an embodiment of the present invention provides an adaptive control method for UAV mapping based on image analysis, and the method includes:

[0024] Step S10: Control the drone equipped with a laser scanning ranging module to fly to the target area and hover stably, perform laser scanning mapping on the target area, and obtain the three-dimensional point cloud data of the target area.

[0025] Specifically, controlling the drone equipped with a laser scanning ranging module to fly to the target area and hover stably includes: collecting the spatial distance data and obstacle boundary information of the surrounding environment, inputting the data into a path planning algorithm, and generating an executable path point sequence according to the three-dimensional coordinates of the target area, flight altitude constraints, and attitude adjustment strategies; during the flight, compare and adjust the flight attitude and position in real time according to the current flight state and path points to avoid obstacles and enter the mapping hover point of the target area.

[0026] Further, performing laser scanning mapping on the target area includes: performing an initial low-density scan to obtain the point cloud skeleton of the target area; analyzing the edge gradient, local curvature, and geometric complexity in the point cloud, and marking the areas where the significant degree of geometric change meets the expectation as the densification scan areas; performing a high-density rescan on the densification scan areas according to the preset scan resolution to obtain a three-dimensional point cloud data set with a higher density.

[0027] In the embodiment of the present invention, to obtain high-precision three-dimensional data of the target area, first control the drone equipped with a laser scanning ranging module to fly above the target area and hover stably at the designated mapping position. During the flight, collect the spatial distance information and obstacle boundary data of the surrounding environment, and input them into a path planning algorithm. Combining the three-dimensional coordinate parameters of the target area, flight altitude limits, aircraft power model, and attitude control constraints, generate a continuous three-dimensional path point sequence. The path point sequence includes lateral and longitudinal control nodes, deceleration buffer segments, and attitude adjustment points, which are used to accurately guide the drone to complete the path transition.

[0028] During the flight control process, continuously monitor the state parameters such as the real-time position, speed, and attitude angle of the drone, and compare them with the path point sequence. Dynamically adjust the flight direction and attitude angle through a deviation correction algorithm, and gradually correct the heading, speed, and flight posture. When approaching the target area, gradually slow down the flight speed to the set threshold and enter the mapping hover mode. Through attitude holding and speed closed-loop control, control the spatial position and attitude fluctuation of the platform within the millimeter range to provide a stable flight platform for subsequent precise measurement operations.

[0029] After completing the hovering and positioning, start the laser scanning and ranging module to perform the laser scanning and mapping operation on the target area. In the initial stage, a low-density wide-area scanning mode is adopted, and the laser emission and echo reception operations are performed with a large angular step and a medium frequency to obtain the point cloud skeleton data covering the target area. The point cloud skeleton contains the general shape information, geometric boundaries, and spatial configuration of the target area, providing a reference basis for the subsequent local densification scanning. After the initial scan is completed, geometric feature analysis is performed on the point cloud skeleton data, and multiple feature indicators including edge gradient, local curvature, and inter-point distribution density are extracted. The edge gradient detects the structural mutation area through the normal change of adjacent points; the local curvature estimates the bending degree by fitting the surface of the local point cluster; and the point density reflects the distribution of the distances between the point sets in the area. Compare the above features with the preset densification determination threshold to determine the areas with significant geometric changes or high structural complexity as the subsequent high-density scanning areas.

[0030] In the target area determined as the densification area, further perform the high-density scanning operation. By adjusting the laser emission frequency, rotation angular velocity, and angular resolution, high-resolution laser scanning is performed in the target area. The scanning angle step size is reduced and the coverage frequency is increased in this stage, and the point cloud density collected is significantly higher than that in the initial scanning stage, thus complementing the detailed features such as the target edge, depression, and protrusion. After completing the fusion of the low-density skeleton data and the high-density local data, a complete three-dimensional point cloud model of the target area is constructed. This model has the characteristics of strong structural continuity, high spatial resolution, and good detail integrity, and can provide a reliable data basis for subsequent operations such as measuring point screening, image acquisition, and micro-displacement detection.

[0031] This process is executed in coordination with path planning, attitude control, and hierarchical scanning strategies, realizing multi-level laser mapping of the target area from coarse to fine. It not only avoids the efficiency loss caused by blind high-density sampling of the entire scene, but also enhances the data details of the key areas while ensuring the overall geometric restoration of the target, improving the usability and accuracy guarantee ability of the point cloud model in subsequent micro-measurement tasks.

[0032] Step S20: Analyze the target features based on the obtained three-dimensional point cloud data to determine at least one target measurement point that requires fine displacement measurement.

[0033] Specifically, after obtaining the three-dimensional point cloud data of the target area, perform structural feature analysis on the point cloud data to determine at least one target measurement point suitable for performing fine displacement measurement. The point cloud data is first subjected to preliminary filtering processing to remove floating points, isolated points, and noise points, and retain the high-quality point set related to the surface structure of the target object. Subsequently, surface reconstruction or normal vector estimation is performed on the point cloud to analyze the geometric shape and change trend of the surface of the target object.

[0034] In the reconstructed three-dimensional point cloud model, the entire target area is divided into several local area units according to a preset spatial partitioning strategy, and geometric feature indicators such as curvature, normal change rate, local point density, and edge gradient of each area are calculated respectively. The curvature value is used to determine whether the area is a concave-convex junction, a corner, an edge, or other high-stress concentration positions; the normal change rate reflects the continuity and roughness of the structure surface; the local density change can reveal surface occlusion or measurement dead angles; the edge gradient is used to detect structural mutations or crack clues.

[0035] Input the above indicators into the comprehensive judgment model. Combining the target structure type, the distribution law of deformation-sensitive areas, and the requirements of the measurement task, several key measurement point positions that meet the conditions of "high structural importance, high deformation possibility, and stable laser measurement response" are screened out. If there are multiple areas that meet the conditions, a weight scoring mechanism can be introduced for ranking, and one or more of them can be selected as the target points for precise displacement measurement. By analyzing the target features in the three-dimensional point cloud and intelligently screening the key measurement points, the high-precision measurement range can be effectively reduced, avoiding repeated interference measurements on irrelevant areas, thereby improving the measurement efficiency and reducing the device load. At the same time, giving priority to areas with significant geometric changes or structurally sensitive areas helps to identify small deformations or potential hazard signs at an early stage, enhancing the device's response ability and application value to changes in the structural state.

[0036] Step S30: Obtain the image data of the target measurement point, and judge whether the measurement point meets the laser interference measurement conditions based on image processing.

[0037] Specifically, collect the visible light image and infrared image of the measurement point area, and perform pixel-level alignment processing on the images; identify the area where the measurement point is located in the registered image, and analyze the occlusion contour, infrared reflection intensity, and texture continuity index of the corresponding area; further collect the polarization image of the measurement point area, and analyze the polarization rate and polarization angle distribution of the measurement point; judge the surface scattering characteristics of the measurement point according to the polarization characteristics, and combine the occlusion contour, infrared reflection intensity, and texture continuity index to determine whether the measurement point meets the laser interference measurement requirements.

[0038] In the embodiments of the present invention, visible light images and infrared images of the measurement point area are collected, and pixel-level alignment processing is performed on the images. Specifically, first, based on the mounted visible light imaging device and infrared imaging device, synchronous or near-synchronous image acquisition operations are respectively performed on the target measurement point area to obtain image data of the area in the visible light band and the infrared band. To ensure the consistency of subsequent analysis, pixel-level alignment processing needs to be performed on the two types of acquired image data. Methods based on feature point matching (such as algorithms like SIFT, SURF, ORB, etc.) can be used to extract key points in their respective images, and the infrared image is spatially transformed through a homography matrix or an affine transformation matrix to overlap the infrared image with the visible light image in the same coordinate system, with pixel positions corresponding one by one, thereby ensuring the spatial consistency and attribute relevance between the image data.

[0039] After image registration is completed, based on the registered visible light image, the specific area where the target measurement point is located is identified. The identification method can be based on the existing three-dimensional point cloud projection information or determine the pixel area corresponding to the measurement point through the known spatial coordinate mapping relationship in the image. For the identified measurement point area, the occlusion contour features are further extracted, that is, whether there are obvious occlusion object contours around, such as cables, vegetation, poles, etc., is identified through methods such as image gradient detection and edge extraction (such as the Canny operator). The occlusion area can be determined by indicators such as local gradient amplitude anomalies and boundary discontinuities. When there are obvious occlusions around the measurement point, the measurability of the measurement point is considered to be reduced.

[0040] At the same time, for the area corresponding to the measurement point in the registered infrared image, the infrared reflection intensity value is extracted. The infrared reflection intensity can be directly obtained through the gray value, which reflects the reflection ability of the target surface to the infrared band laser. If the reflection intensity is too low, it may mean that the target surface scatters or absorbs the laser energy too strongly, which is not conducive to forming a stable laser interference signal; if the reflection intensity is too high and the distribution is extremely concentrated, it may cause overexposure of the interference signal and generate measurement noise. Therefore, the preferred reflection intensity should be in the set intermediate range, and appropriate measurement points can be selected based on empirical thresholds (for example, the gray value is in the range of [80, 200]).

[0041] In addition, the texture continuity index of the measurement point area is analyzed from the registered visible light image. The texture continuity can obtain indicators such as energy, contrast, and correlation by calculating the gray-level co-occurrence matrix (GLCM), or the local binary pattern (LBP) descriptor can be used to characterize the local texture features. If there are high-frequency texture changes or chaotic texture directions in the measurement point area, it may cause distortion of the laser reflection signal and is not conducive to forming a regular interference pattern. The ideal measurement point area should exhibit higher texture energy and lower texture contrast changes to ensure interference stability.

[0042] Based on the above visible light and infrared image analysis, the polarization images of the measuring point area are further collected. The polarization images can be obtained by adding a rotating polarization filter in front of the visible light imaging device, or by using an off-the-shelf polarization imaging module. For the collected polarization images, the degree of polarization (DOP) and the angle of polarization distribution (AOP) of each pixel are extracted. The degree of polarization can be calculated from the image intensities collected at multiple angles (usually 0°, 45°, 90°, 135°) according to the formula:

[0043] ;

[0044] where 𝐼 𝑥 represents the image brightness at the polarization angle 𝑥. By analyzing the degree of polarization of the measuring point area, the scattering characteristics of the surface can be judged. If the degree of polarization is high (e.g., greater than 0.6), it indicates that there is a specular reflection component on the target surface, which is likely to cause saturation or instability of the laser interference signal and is not suitable to be selected as the measurement target; if the degree of polarization is low (e.g., less than 0.2), it means that the surface is mainly diffuse reflection and is suitable for stable interference measurement. The consistency of the angle of polarization distribution is also used as a reference index. If the angle of polarization distribution is disordered, it reflects a large surface roughness or a change in the microstructure, which can also reduce the measurement feasibility.

[0045] Finally, the occlusion contour features, infrared reflection intensity, texture continuity index and polarization characteristic results are comprehensively weighted and analyzed. Scoring rules can be set. For example, each item meets the expected weighted score, and a total score threshold is set (e.g., above 80 points). The measuring points exceeding the threshold are determined to meet the laser interference measurement conditions, and those below the threshold are excluded.

[0046] Step S40: When the image is judged to meet the measurement conditions, collect the laser interference signal of the target measuring point and calculate the vibration or displacement data of the measuring point.

[0047] Specifically, calculating the vibration or displacement data of the measuring point includes: emitting a coherent laser beam to the target measuring point and receiving the reflected signal to form an interference pattern; performing analog-to-digital conversion on the signal of the interference intensity varying with time and inputting it into a phase tracking algorithm to extract a continuous phase sequence; calculating the displacement change curve of the measuring point according to the relationship between the phase change and the laser wavelength; if there is a periodic displacement within the measurement period, extracting the vibration frequency and amplitude parameters through frequency domain analysis; establishing a motion compensation model based on the flight attitude data to eliminate the non-target motion components in the phase change and obtain the actual displacement or vibration response.

[0048] In the embodiments of the present invention, after determining that the target measurement point meets the conditions for laser interferometry, the acquisition of laser interference signals and the extraction of displacement / vibration parameters are carried out. This process is based on the principle of coherent light interference. By irradiating the target surface with a laser beam of high frequency and high stability, the phase change information carried in the reflected interference signal is analyzed, so as to obtain a tiny dynamic displacement or vibration response.

[0049] Control the laser to emit a continuous coherent laser beam towards the target measurement point. The laser wavelength can select a single-mode laser source with high stability (such as the 1550nm and 1064nm bands) to ensure that the interference pattern has good phase stability. After the laser irradiates the surface of the measurement point, part of the light beam is reflected back into the acquisition optical path and forms an interference signal with the reference beam. This interference signal generates a periodic intensity change with the tiny displacement of the target surface, manifested as the dynamic change of interference fringes over time. After receiving the interference signal, first perform analog-to-digital conversion processing on it. Convert the continuously changing light intensity signal into a digital signal through a high-speed analog-to-digital converter (ADC) to form an interference intensity sequence that changes over time. This sequence serves as the input basis for subsequent phase demodulation and requires a sampling frequency higher than twice the target vibration (meeting the Nyquist sampling theorem) to avoid spectral aliasing.

[0050] Perform a phase demodulation operation on the interference intensity signal. Input the acquired time series into a phase tracking demodulation algorithm, such as using the Hilbert transform, quadrature demodulation, or a phase extraction algorithm based on the phase-locked loop (PLL) structure, to obtain a continuous time-domain phase change sequence. Since the movement of the interference fringes directly corresponds to a tiny change in the optical path, the relationship between the phase change amount Δφ and the displacement ΔL is:

[0051] ;

[0052] where λ is the laser wavelength. Through this formula, the demodulated phase change value can be converted into the tiny displacement data of the target surface, thereby reconstructing the displacement change curve of the measurement point during the measurement period. This curve can present the instantaneous displacement amount and is applicable to the detection scenario of static deformation amounts.

[0053] If it is detected that the displacement change during the measurement period has an obvious periodic component, the frequency-domain analysis can be further performed on the displacement curve. Use the fast Fourier transform (FFT) or short-time Fourier transform (STFT) method to extract the spectral structure, and identify the main vibration frequency, harmonic distribution, and corresponding amplitude parameters from it. It can be used to identify the dynamic characteristics such as cable vibration of bridges, working frequencies of mechanical equipment, and environmental vibration effects.

[0054] To improve the reliability of measurement data, it is also necessary to strip the error of the non-target motion components that may be mixed in the laser interference signal. During the measurement process, due to factors such as the slight drift, hovering micro-vibration, and attitude jitter of the UAV platform itself, it may cause changes in the interference optical path that have nothing to do with the target, thereby interfering with the phase demodulation result. Therefore, during the laser interference measurement process, the real-time attitude information of the UAV is collected, including pitch angle, yaw angle, acceleration, and angular velocity data. The flight attitude data is input into the motion compensation model to construct the reference phase drift sequence caused by the motion of the flight platform itself.

[0055] Through signal alignment and phase difference analysis, the reference phase change is removed from the total phase sequence, and only the net change part caused by the real displacement or vibration of the target surface is retained. Finally, the high-confidence displacement / vibration response data after platform motion correction is obtained, which can be directly used for subsequent engineering analysis processes such as structural health assessment and deformation trend modeling.

[0056] Based on the solution of the present invention, without the premise of contacting the measurement point, the dynamic response measurement of the micron level is realized. Combining the real-time attitude compensation mechanism, the stability and reliability of the laser interference measurement on the UAV flight platform are effectively improved, enabling the precision interference measurement ability to be extended from the laboratory to the airborne mobile platform, meeting the actual needs of high-resolution, non-contact, and remote displacement monitoring in complex environments.

[0057] Preferably, when there are multiple measurement points that meet the measurement conditions, the method further includes: calculating the priority ranking according to the spatial position, structural risk level, and illumination conditions of the measurement points; adjusting the position and attitude of the UAV in turn according to the priority, so that the measurement point is located in the direction of the laser measurement optical axis; before each measurement, through attitude adjustment and fine-tuning of the pan-tilt pointing, making the measurement beam perpendicular to the normal of the target surface.

[0058] In the embodiment of the present invention, when there are multiple target measurement points that meet the laser interference measurement conditions, to ensure the measurement efficiency and data quality, the priority ranking of the measurement points based on multiple factors is performed, and the fine measurement operations of each measurement point are completed in turn according to the ranking order. The ranking basis includes factors such as the spatial position of the measurement point in the target structure, the structural risk level, and the current illumination conditions.

[0059] Based on the positions of each measurement point in the three-dimensional point cloud coordinate system and in combination with the structural design drawing or structural topological relationship, determine the engineering importance and structural sensitivity of the area where it is located. For example, structural nodes, connection parts, or regions with abrupt curvature changes are assigned a higher monitoring priority; areas far from the main load-bearing path of the structure or with low measurement value are assigned a lower priority. Secondly, combine external sensing information (such as light sensors, image brightness analysis) to judge the lighting conditions of the area where the current measurement point is located, and preferentially arrange the measurement tasks for areas with good lighting conditions and stable reflection signals to reduce the interference of ambient light on the interference signal. After quantifying and scoring the above factors, perform weighted calculation to form a measurement point priority queue. Subsequently, perform measurements in sequence according to the priority order. Before performing the measurement task of each measurement point, first adjust the flight position and overall attitude of the drone so that the target measurement point is located in the direction of the measurement optical axis of the laser, and control it within the laser measurement working distance range to ensure the beam coverage efficiency and signal echo quality.

[0060] After adjusting the position, further optimize the irradiation direction through a refined attitude control mechanism and the pointing fine-tuning function of the measurement gimbal. This adjustment process is calculated based on the normal direction of the surface where the measurement point is located, and automatically makes the incident direction of the measurement beam as close to perpendicular to this normal direction as possible. By controlling the incident angle of the laser beam to approach the normal direction, the coherence and stability of the interference signal can be effectively improved, and the optical path error and reflection offset caused by oblique incidence can be reduced.

[0061] Step S50: Fuse and process the three-dimensional point cloud data and the vibration / displacement data of the measurement point, and output the geometric size and displacement measurement results of the target.

[0062] Specifically, fusing and processing the three-dimensional point cloud data and the vibration / displacement data of the measurement point includes: establishing a multi-field data structure containing three-dimensional coordinates, displacement values, vibration frequencies, and measurement times for each measured target measurement point; embedding the multi-field data structure into the original index position of the point cloud data to achieve attribute-level expansion; generating a composite model containing spatial shapes and measurement responses.

[0063] Furthermore, outputting the geometric size and displacement measurement results of the target includes: generating a structured data file, where the structured data file contains a three-dimensional point cloud model, displacement / vibration parameters of key measurement points, and measurement element information; assigning a unique identifier to each key measurement point, and attaching the corresponding measurement time, numerical result, and confidence level; exporting the measurement results in a parsable standard format.

[0064] In the embodiments of the present invention, after non-contact laser interferometry measurements of multiple target measurement points are completed, further fusion processing is performed on the three-dimensional point cloud data and the vibration or displacement data corresponding to the measurement points, so as to output complete and analyzable target geometric dimension information and displacement response results. Preprocessing operations are performed on the three-dimensional point cloud data generated by the laser scanning module to ensure that the position data of each measurement point in the point cloud has a unique spatial index identifier. Methods such as spatial hashing encoding, KD-tree construction, or voxel grid division can be used to assign corresponding spatial index values to each point in the point cloud, so as to perform a one-to-one fusion operation with the displacement data subsequently. Subsequently, a multi-field data structure is established for each target measurement point that has completed laser interferometry measurements, and the field content includes at least the following:

[0065] 1) The three-dimensional spatial coordinate values (X, Y, Z) of the point position;

[0066] 2) The measured displacement change value (ΔL), which can be a single static displacement or time series data;

[0067] 3) If there is a periodic response in the measurement, it includes the main vibration frequency (f), amplitude (A), and possible harmonic components;

[0068] 4) The corresponding measurement timestamp (T), which is used for multi-temporal data management and comparative analysis;

[0069] 5) The confidence score of the interferometry measurement, which can be calculated based on the signal-to-noise ratio, phase demodulation continuity, or attitude compensation residuals.

[0070] After the data structure is established, the structure is embedded into the original index position of the three-dimensional point cloud data to perform attribute-level data fusion. During the fusion process, without changing the original geometric topology of the point cloud, the measurement point attributes are extended to the additional fields of the point cloud point object. For example, fields such as "displacement", "frequency", and "confidence" are added to each point record in the point cloud to achieve the extension from single coordinate information to spatial-physical composite information.

[0071] For the case where displacement measurement results are available for only some points, the original fields of the unmeasured points are retained, and the measured values are optionally extended to the neighboring areas through interpolation algorithms (such as KNN interpolation, weighted average diffusion) for constructing a displacement change field or vibration thermogram.

[0072] After the fusion is completed, a composite spatial model of the target area is generated based on the extended point cloud data. This model not only retains the geometric shape of the target structure but also contains the response information of the measurement points in the time domain. Further, local vertices on the model surface can be deformed and simulated according to the displacement direction and amplitude of the measurement points, so that the three-dimensional model visually reflects the structural response trends, such as morphological changes like bulging, denting, and offset. The vibration frequency or displacement amplitude can also be projected onto the model surface through color mapping to achieve heat map visualization. In the output result stage, a structured measurement data file is generated. This file includes at least:

[0073] 1) Three-dimensional point cloud model data, including spatial coordinates and topological structure;

[0074] 2) A complete set of attribute fields for the measurement points, including vibration / displacement parameters, measurement time, and confidence level;

[0075] 3) Meta-information such as model unit description, coordinate system identification, and measurement device parameters, which is convenient for subsequent data calling and compatibility between platforms.

[0076] For the convenience of data tracking and analysis, a unique identifier is generated for each measurement point. The unique label can be generated by means of measurement point coordinate hashing, task number splicing, or device unique coding. Each label is attached with the corresponding measurement result value, time stamp, and confidence level, and written into the structured data packet. The fused model and data results are exported in a parsable standard format, such as PLY (point cloud format with attributes), OBJ (supporting topology and attribute mapping), GeoJSON (format supported by structural monitoring devices), or a custom XML / JSON structure. The exported files can be directly used by the structural health assessment platform, building information modeling (BIM) devices, or post-processing software, or uploaded to the cloud monitoring device through a network interface to achieve remote synchronous viewing.

[0077] Based on the solution of the present invention, the deep fusion of structural geometric information and high-precision dynamic measurement data is realized, and the static spatial modeling and dynamic behavior recognition are unified in one data model. This fusion model can be used for tasks such as multi-temporal comparison analysis, deformation trend visualization, local risk positioning, and structural safety assessment. Compared with the traditional separate data acquisition method, this method effectively reduces data mismatch, synchronization error, and modeling complexity, while enhancing the readability, expandability, and engineering application value of the measurement results.

[0078] In another possible implementation manner, after the preliminary three-dimensional point cloud modeling of the target area is completed, combined with spatial model analysis and structural change pattern recognition, an additional interference precise measurement operation is automatically triggered for the abnormal areas in the model. The specific steps include:

[0079] After the point cloud modeling is completed, extract the surface feature change rate of the model, including indicators such as curvature change, normal continuity, and boundary deviation. By comparing the morphology with the standard structure model, identify the positions with obvious geometric anomalies, such as depressions, local protrusions, continuity breaks, and structural misalignments. Perform local enhanced point cloud analysis on the abnormal areas to determine whether there are structural anomaly risks (such as cracks, settlements, and offset trends). If the risk determination conditions are met, immediately automatically deploy a high-density sampling grid inside the area, and generate a list of new measurement points according to the fine measurement point layout standard. Perform image judgment and laser interferometry on the new measurement points in sequence to obtain their dynamic response parameters. Finally, fuse the data of the new measurement points with the original point cloud and the data of the existing measurement points, and output an updated composite model containing the response data of the automatically added measurement points.

[0080] This embodiment realizes active perception and response to unknown structural anomalies through an anomaly-driven measurement point addition mechanism, which not only reduces the number of invalid measurement points, but also enhances the targeted monitoring of the structural deformation trend. It has stronger intelligent features and is suitable for complex scenarios such as post-disaster assessment and high-risk area monitoring.

[0081] Figure 2 It is the device structure diagram of the unmanned aerial vehicle mapping adaptive control device based on image analysis provided by an embodiment of the present invention. As Figure 2 shown, the embodiment of the present invention provides an unmanned aerial vehicle mapping adaptive control device based on image analysis. The device includes: a collection unit for controlling the unmanned aerial vehicle equipped with a laser scanning and ranging module to fly to the target area and hover stably, and performing laser scanning mapping on the target area to obtain the three-dimensional point cloud data of the target area; a measurement point determination unit for analyzing the target features based on the obtained three-dimensional point cloud data and determining at least one target measurement point that needs to perform fine displacement measurement; a judgment unit for obtaining the image data of the target measurement point and judging whether the measurement point meets the laser interferometry condition based on image processing; a processing unit for collecting the laser interference signal of the target measurement point and calculating the vibration or displacement data of the measurement point when the image judgment meets the measurement condition; an output unit for performing fusion processing on the three-dimensional point cloud data and the vibration / displacement data of the measurement point, and outputting the geometric size and displacement measurement result of the target.

[0082] The embodiment of the present invention also provides a computer-readable storage medium, which stores instructions that cause a computer to execute the above-mentioned unmanned aerial vehicle mapping adaptive control method based on image analysis when running on the computer.

[0083] Those skilled in the art can understand that all or part of the steps in the methods of the above embodiments can be completed by instructing relevant hardware through a program. The program is stored in a storage medium and includes several instructions to enable a single-chip microcomputer, a chip, or a processor to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0084] The optional embodiments of the present invention have been described in detail above in conjunction with the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details in the above embodiments. Within the technical concept scope of the embodiments of the present invention, various simple variants can be made to the technical solutions of the embodiments of the present invention, and these simple variants all fall within the protection scope of the embodiments of the present invention. Additionally, it should be noted that in the above specific embodiments, the various specific technical features described can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the embodiments of the present invention will not separately describe various possible combination methods.

[0085] Furthermore, any combination can be made among the various different embodiments of the present invention as long as it does not violate the idea of the embodiments of the present invention, and it should also be regarded as the content disclosed in the embodiments of the present invention.

Claims

1. An adaptive control method for UAV mapping based on image analysis, characterized in that, The method includes: Controlling a drone equipped with a laser scanning ranging module to fly to a target area and hover stably, performing laser scanning mapping on the target area, and obtaining three-dimensional point cloud data of the target area; Analyzing target features based on the obtained three-dimensional point cloud data, and determining at least one target measuring point that requires precise displacement measurement; Collecting visible light images and infrared images of the measuring point area, and performing pixel-level alignment processing on the images; Identifying the area where the measuring point is located in the registered images, and analyzing the occlusion contour, infrared reflection intensity, and texture continuity index of the corresponding area; Further collecting polarization images of the area where the measuring point is located, and analyzing the polarization rate and polarization angle distribution of the measuring point; Judging the surface scattering characteristics of the measuring point according to the polarization characteristics, and combining the occlusion contour, infrared reflection intensity, and texture continuity index to determine whether the measuring point meets the requirements of laser interferometry; When the image is judged to meet the measurement conditions, emitting a coherent laser beam to the target measuring point, and receiving the reflected signal to form an interference pattern; Performing analog-to-digital conversion on the signal of the interference intensity varying with time, and inputting it into a phase tracking algorithm to extract a continuous phase sequence; Calculating the displacement change curve of the measuring point according to the relationship between the phase change and the laser wavelength; If there is periodic displacement during the measurement period, extracting vibration frequency and amplitude parameters through frequency domain analysis; Establishing a motion compensation model based on the flight attitude data, and eliminating the non-target motion components in the phase change to obtain the actual displacement or vibration response; Performing fusion processing on the three-dimensional point cloud data and the vibration / displacement data of the measuring point, and outputting the geometric size and displacement measurement results of the target.

2. The adaptive control method for UAV mapping based on image analysis according to claim 1, characterized in that, Controlling a drone equipped with a laser scanning ranging module to fly to a target area and hover stably includes: Collecting spatial distance data and obstacle boundary information of the surrounding environment, inputting the data into a path planning algorithm, and generating an executable path point sequence according to the three-dimensional coordinates of the target area, flight altitude constraints, and attitude adjustment strategies; During the flight, comparing and adjusting the flight attitude and position in real time according to the current flight state and path points, avoiding obstacles and entering the mapping hover point of the target area.

3. The adaptive control method for UAV mapping based on image analysis according to claim 1, wherein Performing laser scanning mapping on the target area includes: Performing an initial low-density scan to obtain the point cloud skeleton of the target area; Analyzing the edge gradient, local curvature, and geometric complexity in the point cloud, and marking the area where the geometric change significance meets the expectation as the densification scan area; Performing a high-density rescan on the densification scan area according to the preset scan resolution to obtain a three-dimensional point cloud data set with a greater density.

4. The adaptive control method for UAV mapping based on image analysis according to claim 1, characterized in that If there are multiple measuring points that meet the measurement conditions, the method further includes: Calculating the priority ranking according to the spatial position, structural risk level, and lighting conditions of the measuring point; Adjusting the position and attitude of the drone in sequence according to the priority, so that the measuring point is located in the direction of the laser measurement optical axis; Before each measurement, adjusting the attitude and fine-tuning the pan-tilt direction to make the measurement beam perpendicular to the normal of the target surface.

5. The adaptive control method for UAV mapping based on image analysis according to claim 1, wherein, Performing fusion processing on the three-dimensional point cloud data and the vibration / displacement data of the measuring point includes: Establishing a multi-field data structure including three-dimensional coordinates, displacement values, vibration frequencies, and measurement times for each measured target measuring point; Embed the multi-field data structure into the original index position of the point cloud data to achieve attribute-level expansion; Generate a composite model including spatial shape and measurement response.

6. The adaptive control method for UAV mapping based on image analysis according to claim 1, characterized in that Output the geometric dimensions and displacement measurement results of the target, including: Generate a structured data file, which includes a three-dimensional point cloud model, displacement / vibration parameters of key measurement points, and measurement element information; Assign a unique identifier to each key measurement point, and attach the corresponding measurement time, numerical result, and confidence level; Export the measurement results in an analyzable standard format.

7. An adaptive control device for UAV mapping based on image analysis, characterized in that, The device includes: A collection unit, which is used to control the UAV equipped with a laser scanning ranging module to fly to the target area and hover stably, perform laser scanning mapping on the target area, and obtain the three-dimensional point cloud data of the target area; A measurement point determination unit, which is used to analyze the target characteristics based on the obtained three-dimensional point cloud data and determine at least one target measurement point that needs to perform fine displacement measurement; A judgment unit, which is used to: Collect visible light images and infrared images of the measurement point area, and perform pixel-level alignment processing on the images; Identify the area where the measurement point is located in the registered image, and analyze the occlusion contour, infrared reflection intensity, and texture continuity index of the corresponding area; Further collect polarization images of the area where the measurement point is located, and analyze the polarization rate and polarization angle distribution of the measurement point; Judge the surface scattering characteristics of the measurement point according to the polarization characteristics, and combine the occlusion contour, infrared reflection intensity, and texture continuity index to determine whether the measurement point meets the requirements of laser interferometry; A processing unit, which is used to: When the image is judged to meet the measurement conditions, emit a coherent laser beam to the target measurement point, and receive the reflected signal to form an interference pattern; Perform analog-to-digital conversion on the signal of the interference intensity changing with time, and input the phase tracking algorithm to extract the continuous phase sequence; According to the relationship between the phase change and the laser wavelength, calculate the displacement change curve of the measurement point; If there is periodic displacement during the measurement period, extract the vibration frequency and amplitude parameters through frequency domain analysis; Establish a motion compensation model according to the flight attitude data, and eliminate the non-target motion components in the phase change to obtain the actual displacement or vibration response; An output unit, which is used to perform fusion processing on the three-dimensional point cloud data and the vibration / displacement data of the measurement point, and output the geometric dimensions and displacement measurement results of the target.

8. A computer-readable storage medium, characterized in that, Instructions are stored on the computer-readable storage medium, and when they run on a computer, they cause the computer to execute the image analysis-based UAV mapping adaptive control method according to any one of claims 1-6.

Images