A construction site drone intelligent inspection method

Abstract

The present invention belongs to the technical field of drone inspection and relates to an intelligent drone inspection method for construction sites. The method plans the drone inspection path based on the construction progress map, loads the preset intersection position of the completed section and the current operation section, controls the drone to fly along the inspection path, drives the drone to respond to the first acquisition mode of the completed section and the second acquisition mode of the current operation section, and determines the validity of the preset intersection position by verifying the ditch boundary structure characteristics to confirm the actual intersection position, realizes adaptive control of the drone's multiple acquisition modes, then fuses the multiple acquisition modes to generate data, generates alarm instructions and hidden danger location marking information based on the ditch bottom state abnormality level or slope stability risk level, and simultaneously pushes them to the construction terminal, thereby promoting the engineering decision-making closed loop of drone water conservancy ditch segmented excavation construction site inspection, and improving the safety and management efficiency of water conservancy project construction.

Description

Technical Field

[0001] The present invention belongs to the technical field of unmanned aerial vehicle inspection, and relates to an intelligent unmanned aerial vehicle inspection method for a construction site. Background Art

[0002] Modern construction projects are becoming increasingly large-scale and complex, posing significant challenges to safety, quality, and progress control on construction sites. Traditional, manual inspection methods, with their inherent flaws such as limited coverage, low efficiency, difficulty reaching high-risk areas, and strong subjectivity, struggle to meet the demands of refined, real-time management of large, complex projects. Drone technology, with its significant advantages of high mobility, flexible deployment, multi-angle observation, and rapid acquisition of large-scale spatial information, is rapidly becoming a key tool for intelligent monitoring and efficient management of construction sites.

[0003] In the field of water conservancy project construction, water conservancy ditches, as core infrastructure of water conservancy projects, have a construction quality that is directly related to the efficiency of regional water resource allocation and the safety of people's livelihoods. In actual construction, the segmented excavation construction method is widely adopted due to its enhanced construction flexibility. However, this method also leads to scattered work areas, complex environments, and frequent cross-construction on the project site.

[0004] Drone inspections, with their advantages of high maneuverability, multi-angle observation, and real-time data collection, have demonstrated significant value in construction progress monitoring, quality inspection, and safety management. However, current drone inspections at water conservancy and canal construction sites still face technical bottlenecks such as insufficient positioning accuracy, low data processing efficiency, and poor adaptability to complex environments. Therefore, exploring intelligent inspection methods is crucial for improving the scientific and reliable management of water conservancy project construction.

[0005] However, it should be noted that the existing drone inspection solutions for water conservancy ditch segmented excavation construction sites still have significant limitations when dealing with the above-mentioned complex working conditions and technical challenges, which are mainly manifested in the following aspects: 1. Existing technologies mostly focus on the optimal planning of drone inspection paths, but fail to fully consider the differences in construction progress of different construction sections in the water conservancy ditch segmented excavation construction site, and lack precise matching analysis of the drone's acquisition modes under different inspection tasks, resulting in insufficient fit between the inspection data and actual needs, which in turn causes the problem that the acquisition accuracy is difficult to meet engineering requirements.

[0006] 2. Existing technologies lack an automatic switching and analysis mechanism for drone acquisition modes based on construction site environmental perception. Currently, they rely more on manually preset time nodes or spatial locations to trigger switching. Such fixed trigger conditions are difficult to match the dynamic changes of segmented excavation construction sites, which can easily lead to data redundancy or omission of key information, thereby affecting the integrity and accuracy of inspection data. Summary of the Invention

[0007] In view of this, in order to solve the problems raised in the above background technology, a construction site drone intelligent inspection method is proposed.

[0008] The purpose of the present invention can be achieved through the following technical solutions: The present invention provides a construction site drone intelligent inspection method, including: planning the drone inspection path based on the construction progress map, and loading the preset intersection position of the completed section and the current operation section.

[0009] The drone is controlled to fly along the inspection path, and in response to entering a completed section, the drone is switched to a first acquisition mode, and ditch bottom abnormality feature data of the completed section is generated in real time during the point cloud acquisition process.

[0010] When the UAV arrives at the airspace near the preset intersection position, it verifies the ditch boundary structure characteristics to determine the validity of the preset intersection position, and confirms the actual intersection position based on the determination result.

[0011] When the UAV crosses the actual boundary position and enters the current operation section, the UAV is switched to the second acquisition mode, and the slope risk characteristic data of the current operation section is generated in real time through edge waveform variation recognition.

[0012] The inspection report is generated by integrating the abnormal characteristic data of the ditch bottom and the risk characteristic data of the slope. Based on the abnormal level of the ditch bottom state or the slope stability risk level in the report, an alarm instruction and hidden danger location marking information are generated and pushed to the construction terminal simultaneously.

[0013] Compared with the existing technology, the beneficial effects of the present invention are as follows: (1) The present invention dynamically couples the construction progress map with the drone acquisition mode, drives the drone to match the first acquisition mode of the completed section and the second acquisition mode of the current operation section, ensures that the drone payload resources are accurately adapted to the project requirements, and effectively solves the problem of insufficient data validity caused by mode mismatch or mismatch during the inspection process.

[0014] (2) The present invention constructs a ditch boundary structure feature verification mechanism, analyzes the elevation gradient, reflection intensity and image texture change rate in real time in the airspace near the preset intersection position, determines the actual intersection position through comprehensive boundary feature indicators, promotes the perception closed loop of dynamic changes in construction progress, and further realizes the adaptive switching control of multiple acquisition modes of drones.

[0015] (3) The present invention effectively establishes a cross-modal data fusion architecture for ditch bottom anomalies and slope risks, generates alarm instructions and hidden danger location marking information based on the ditch bottom state anomaly level or slope stability risk level, and conducts slope instability probability analysis, which is pushed to the construction terminal, promoting the engineering decision-making closed loop of drone water conservancy ditch segmented excavation construction site inspection, and improving the safety and management efficiency of water conservancy project construction. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 The present invention is a flowchart of the steps for implementing the method.

[0018] Figure 2 This is a flowchart of the execution operation of the first acquisition mode of the drone of the present invention.

[0019] Figure 3 This is a flowchart of the execution operation of the second acquisition mode of the drone of the present invention. DETAILED DESCRIPTION

[0020] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0021] See also Figure 1 As shown, the present invention provides a construction site drone intelligent inspection method, including: S1. planning the drone inspection path based on the construction progress map, and loading the preset intersection position of the completed section and the current operation section.

[0022] In a preferred embodiment of the present invention, the planning of the drone inspection path based on the construction progress map includes: calling the ditch centerline coordinate sequence of the construction progress map, projecting the centerline to the geographic coordinate system, binding the path starting point to the starting point coordinate of the completed section, and the end point to the latest updated point coordinate of the current construction section, and generating the initial reference path of the drone.

[0023] Control the UAV to perform a pre-flight scan at the starting point of the initial reference path to identify obstacles that conflict with the planned trajectory.

[0024] It should be noted that the judgment criterion for the above-mentioned spatial conflict is that the minimum three-dimensional distance between the outer contour of the obstacle and the planned trajectory is less than or equal to the drone's preset safety margin threshold, where the drone's preset safety margin threshold specifically refers to the minimum safety buffer distance required for the drone to perform obstacle avoidance actions in three-dimensional space. Its value can be found in the drone's factory technical manual.

[0025] Horizontal obstacle avoidance offset correction is performed on the path nodes of the obstacle-affected section to generate the final planned inspection path, wherein the horizontal obstacle avoidance offset correction requires that the relative spatial relationship between the corrected section and the target detection area meets the preset acquisition constraint conditions.

[0026] It should be added that the specific contents of the above-mentioned preset collection constraints include: if the road section affected by the obstacle is located in a completed section, the projection of the road section after correction is required to cover the ditch width range of a preset multiple of the ditch centerline.

[0027] If the road section affected by the obstacle is located in the current construction section, the angle between the corrected road section and the slope normal direction is required to be less than the preset angle threshold.

[0028] Among them, the relevant standards, specifications or technical guidelines for the construction of segmented excavation of water conservancy ditches will provide reference values ​​or recommended values ​​for the preset multiple ditch width range and the preset angle threshold. For example, according to the anti-seepage requirements regulations of the "Irrigation and Drainage Engineering Design Standards", the coverage on both sides of the ditch centerline is times the design ditch width. The slope design specification document for water conservancy and hydropower projects stipulates that the angle between the optical axis of the slope deformation monitoring equipment and the slope normal should not be greater than To avoid distortion of elevation projection, the preset multiple ditch width range and the preset angle threshold can be exemplified as: times or .

[0029] It should also be added that the loading of the preset boundary position between the above-mentioned completed section and the current operation section is specifically derived from the progress-related marking point authorized by the supervision terminal in the construction progress map.

[0030] S2. Control the drone to fly along the inspection path. In response to entering a completed section, switch the drone to the first acquisition mode, and generate ditch bottom abnormality feature data of the completed section in real time during the point cloud acquisition process.

[0031] See also Figure 2 As shown, in a preferred embodiment of the present invention, the first acquisition mode performs the following operations in sequence: determining the flight altitude to be maintained by the UAV based on the trench depth data detected in the completed section.

[0032] It should be noted that the above-mentioned UAV needs to maintain its flight altitude so that the effective detection depth of the laser scanning device covers the entire cross-section of the ditch bottom, and the point cloud density meets the preset accuracy requirements. The determination process includes: the geometric coefficient calibrated by the laser divergence angle and the receiving field of view angle of the laser scanning device is multiplied by the ditch depth data detected in the completed section to obtain the reference height item.

[0033] Obtain the point cloud detection density of the ditch bottom in the completed section. When the density is lower than the preset density threshold, apply negative compensation to the reference height item according to the preset compensation height of the unit excess density to determine the maintained flight altitude.

[0034] Control the UAV to maintain a low-disturbance flight attitude so that the emission direction of the laser scanning device it carries is perpendicular to the ditch bottom plane.

[0035] Perform reciprocating zigzag flight along the ditch axis to cover the full width of the ditch bottom, and simultaneously activate the laser scanning device to perform continuous panoramic scanning vertically downward.

[0036] Panoramic optical image acquisition is triggered at a turning point of the path of the reciprocating zigzag flight, and the central area of ​​the optical image is spatially aligned with the central area of ​​the laser scanning.

[0037] In a preferred embodiment of the present invention, the process of generating abnormal characteristic data of the trench bottom of the completed section in real time includes: in the process of scanning the trench bottom of the completed section using a laser scanning device, detecting in real time the elevation deviation area, the abnormal linear area of ​​reflection intensity, and the moisture-sensitive color gamut concentration area on the trench bottom surface based on optical image recognition;

[0038] It should be noted that the above-mentioned elevation deviation area detection process of the ditch bottom surface includes: quantifying the absolute deviation between the elevation of each point cloud of the ditch bottom surface in the completed section and the ditch design elevation, comparing the absolute deviation with the preset elevation allowable error threshold of the drone, identifying each abnormal point cloud with an absolute deviation exceeding the threshold, marking the point set that continuously exceeds the preset construction error threshold, and clustering to generate the elevation deviation area boundary.

[0039] The above-mentioned reflection intensity abnormal linear area detection process includes: rasterizing the reflection intensity data collected by the laser scanning device into a two-dimensional spatial distribution matrix, calculating the reflection intensity gradient field through the spatial gradient operator, and identifying the mutation area where the gradient amplitude is significantly higher than the local background value.

[0040] Extract linear gradient mutation features with continuous extension characteristics, and their extension scale must exceed the preset geometric continuity threshold.

[0041] Generate a buffer zone with adaptive width along the linear feature axis to form the boundary of the abnormal reflection intensity area.

[0042] The humidity-sensitive color gamut cluster area identification process includes: converting the optical image synchronously collected by the drone from the RGB color space to the HSV color space.

[0043] The preset humidity-sensitive color gamut threshold is applied to perform pixel-level segmentation on the HSV image to generate a binary mask.

[0044] Through the registration process of point cloud and image, the target pixels in the mask image are mapped to three-dimensional space coordinates.

[0045] Density clustering analysis is performed on the set of spatial coordinate points to identify continuous clustered areas that meet preset spatial density constraints, where the spatial density constraints can specifically refer to a sphere below a preset diameter threshold containing more than a preset number of target pixels.

[0046] When any abnormal area is detected, the full-index correlation analysis of the abnormal area is automatically triggered to extract the normalized parameters of the elevation deviation, linear extension scale and humidity area coverage of the abnormal area.

[0047] A weight distribution rule is configured according to the ditch material type, and the normalized parameters are linearly weighted fused to generate a ditch bottom anomaly evaluation index for the abnormal area.

[0048] It should be noted that the basis for configuring the weight distribution rule based on the ditch material type is the difference in sensitivity of different materials to abnormal indicators. For example, when the ditch material type is concrete, the linear weights of the normalized parameters of the elevation deviation, linear extension scale, and humidity area coverage can be assigned as 0.6, 0.3, and 0.1, respectively. The key point is that the rigid structure of concrete is most sensitive to elevation changes. When the ditch material type is soil, the linear weights can be assigned in sequence as 0.2, 0.3, and 0.5. The key point is that humidity accumulation indicates soil saturation and instability.

[0049] The present invention can construct a mapping relationship table between predefined material types and allocation weight vectors based on industry experience or through preliminary experimental calibration. During the analysis process, the corresponding allocation weight vector of the mapping relationship table can be directly extracted according to the ditch material attribute index.

[0050] According to the abnormal level of each ditch bottom state and its preset evaluation index range stored in the WEB cloud, the abnormal level of the ditch bottom state in the abnormal area is matched and the corresponding level label is output.

[0051] The level label is bound to the spatial coordinates of the abnormal area to form the ditch bottom abnormality feature data.

[0052] S3. When the UAV arrives at the airspace near the preset intersection position, verify the ditch boundary structure characteristics to determine the validity of the preset intersection position, and confirm the actual intersection position based on the determination result.

[0053] In a preferred embodiment of the present invention, the adjacent airspace of the preset intersection position is defined as: a rectangular airspace defined by taking the preset intersection position as a reference, extending the preset intervals in both forward and reverse directions along the flight trajectory to determine the length, determining the width by the real-time ditch bottom detection width, and determining the height by the current flight maintenance altitude.

[0054] In a preferred embodiment of the present invention, the verification of the ditch boundary structural features includes: dividing the ditch bottom surface into continuous strip-shaped units along a direction perpendicular to the ditch center axis in the airspace adjacent to the preset intersection position.

[0055] The elevation gradient change rate, reflection intensity change rate and image texture change rate between adjacent strip units are quantified and fused to generate a comprehensive boundary feature index at the junction of each adjacent unit.

[0056] It should be noted that the above-mentioned elevation gradient change rate is obtained by evaluating the significant change degree of the surface flatness characteristics of the ditch bottom between adjacent strip-shaped units. The specific quantification process is: the elevation distribution fluctuation index of the ditch bottom point cloud of the front and rear adjacent units is obtained respectively by calculating the regional elevation standard deviation, and the absolute difference between the elevation distribution fluctuation index of the ditch bottom point cloud of the front and rear adjacent units is ratioed with the elevation distribution fluctuation index of the ditch bottom point cloud of the front unit to obtain the elevation gradient change rate between adjacent strip-shaped units.

[0057] The reflection intensity change rate is obtained by evaluating the degree of significant change in the stability of the surface material reflection characteristics between adjacent strip-shaped units. The specific quantification process is as follows: the reflection intensity values ​​of all laser points in the current strip-shaped unit are obtained to form a reflection intensity distribution sequence. Based on the preset minimum point group size threshold, the reflection intensity core interval that meets the aggregation density condition in the distribution sequence is determined. With the median reflection intensity of the core interval as the center, a symmetrical reflection intensity characteristic representation interval is constructed. The ratio of the number of laser points in the characteristic representation interval to the total number of laser points in the unit is calculated as the reflection intensity concentration characteristic value of the strip-shaped unit.

[0058] The reflection intensity concentration feature quantities of the front and rear adjacent units are extracted respectively, and the absolute deviation ratio of the rear unit relative to the front unit in the reflection intensity concentration feature quantity is quantified to obtain the reflection intensity change rate between adjacent strip-shaped units.

[0059] The image texture change rate is obtained by evaluating the degree of significant change in the complexity of the surface visual structure between adjacent strip-shaped units. The specific quantification process is as follows: the texture regularity index of the optical images of the front and rear adjacent units is extracted respectively, and the absolute deviation ratio of the rear unit relative to the front unit in the texture regularity index is quantified to obtain the image texture change rate between adjacent strip-shaped units. The texture regularity index can be obtained by constructing a texture feature matrix, extracting the contrast value, entropy value and energy represented by the matrix, and performing a ratio operation on the sum of the contrast value and the energy and the entropy value.

[0060] It should also be noted that the above-mentioned comprehensive boundary characteristic index is the cumulative calculation result of the elevation gradient change rate, reflection intensity change rate and image texture change rate between adjacent strip-shaped units.

[0061] The unit boundary position where the comprehensive boundary characteristic index exceeds the preset verification threshold is determined to be the mutation boundary position from the continuous hardened surface to the exposed soil.

[0062] In a preferred embodiment of the present invention, determining the validity of the preset boundary position includes: quantifying the coordinate distance between the sudden change boundary position and the preset boundary position.

[0063] It should be noted that the above coordinate spacing can be quantified using the Euclidean distance calculation formula.

[0064] If the coordinate spacing is less than or equal to the preset tolerance spacing threshold, the preset intersection position is determined to be valid and the preset intersection position is confirmed to be the actual intersection position; otherwise, it is determined to be invalid and the sudden change intersection position is confirmed to be the actual intersection position.

[0065] The embodiment of the present invention constructs a ditch boundary structure feature verification mechanism, analyzes the elevation gradient, reflection intensity and image texture change rate in real time in the airspace near the preset intersection position, determines the actual intersection position through comprehensive boundary feature indicators, promotes the perception closed loop of dynamic changes in construction progress, and further realizes adaptive switching control of multiple acquisition modes of drones.

[0066] S4. When the UAV crosses the actual boundary position and enters the current operation section, the UAV is switched to the second acquisition mode, and slope risk characteristic data of the current operation section is generated in real time through edge waveform variation recognition.

[0067] See also Figure 3 As shown, in a preferred embodiment of the present invention, the second acquisition mode performs the following operations in sequence: adjusting the yaw angle of the UAV so that its fuselage axis is parallel to the slope direction, synchronously controlling the gimbal pitch angle to match the slope inclination angle, and keeping the synthetic aperture radar beam perpendicular to the slope.

[0068] Control the UAV to perform variable pitch spiral flight, including: i. Real-time calculation of slope inclination and dynamic adjustment of spiral pitch.

[0069] ii. Maintain a constant relative observation distance between the UAV and the slope based on laser ranging feedback.

[0070] iii. Expand the spiral radius to cover the entire area from the top to the toe of the slope based on the geometric characteristics of the slope.

[0071] During the flight, synthetic aperture radar and multispectral imaging are activated synchronously to obtain slope displacement time series data and moisture content distribution data.

[0072] In a preferred embodiment of the present invention, the real-time generation process of the slope risk characteristic data of the current operation section includes: solving the displacement vector field of adjacent time-series slope point clouds, and marking the coordinate set of the risk area where the displacement rate exceeds a preset safety threshold.

[0073] A slope moisture content distribution map is constructed and the areas where the moisture content exceeds the preset saturation threshold are outlined. The map is spatially superimposed with the risk area coordinate set to generate a displacement-moisture content composite risk area.

[0074] The slope stability safety reduction factor of the composite risk area is calculated by combining the preset parameters of the slope material, the moisture content data and the area ratio of the composite risk area through geotechnical shear strength reduction simulation.

[0075] It should be noted that the above-mentioned rock and soil shear strength reduction simulation process is: based on the preset parameters of the slope material, an attenuation function of the rock and soil shear strength parameters as the moisture content increases is established. The attenuation function can be pre-calibrated by geotechnical mechanics tests in the early stage of the development of the invention method.

[0076] Based on the spatial distribution of moisture content in the composite risk area, the shear strength calculation of the rock and soil is performed on each grid unit divided in the composite risk area.

[0077] The critical state of overall instability of the slope is determined through virtual mechanical equilibrium analysis.

[0078] The slope stability safety reduction factor is calculated based on the critical state reduction ratio.

[0079] According to each slope stability risk level and its preset safety reduction factor range stored in the WEB cloud, the slope stability risk level of the composite risk area is matched.

[0080] The boundary coordinates of the composite risk area and the slope stability risk level are stored in association to generate slope risk characteristic data.

[0081] The embodiment of the present invention dynamically couples the construction progress map with the drone acquisition mode, drives the matching of the drone's first acquisition mode in the completed section and the second acquisition mode in the current operation section, ensures that the drone's payload resources are accurately adapted to the project requirements, and effectively solves the problem of insufficient data validity caused by mode mismatch or mismatch during the inspection process.

[0082] S5. Generate an inspection report by integrating the ditch bottom abnormality feature data and the slope risk feature data. Generate an alarm instruction and hidden danger location marking information based on the ditch bottom abnormality level or slope stability risk level in the report, and push it to the construction terminal simultaneously.

[0083] In a preferred embodiment of the present invention, after the slope risk characteristic data is generated, the method further includes: collecting meteorological data within a preset time window of the composite risk area, including wind speed and precipitation intensity.

[0084] The moisture content data of the composite risk area and the meteorological data are input into a preset landslide prediction model to calculate the probability of slope instability.

[0085] It should be noted that the construction process of the above-mentioned preset landslide prediction model is as follows: obtaining a regional moisture content experimental data set and a meteorological experimental data set, preprocessing and normalizing the experimental data set, dividing the processed experimental data set into a training set and a test set, using the training set to train the model based on a machine learning algorithm and automatically adjust the weights of the regional moisture content data and the meteorological data, constructing a landslide prediction model by capturing the probability relationship between the regional moisture content data and the meteorological data inducing slope instability, and verifying and evaluating the model through the test set to obtain a trained model, inputting the moisture content data and meteorological data of the composite risk area into the trained landslide prediction model, and outputting the probability of slope instability.

[0086] If the slope instability probability is greater than the preset warning probability threshold, a weather warning label is added to the inspection report.

[0087] With the center coordinates of the composite risk area as the center of the circle, a restricted area for construction machinery is planned with a preset radius, and emergency avoidance instructions are pushed to the construction terminal.

[0088] The embodiment of the present invention effectively establishes a cross-modal data fusion architecture for ditch bottom anomalies and slope risks, generates alarm instructions and hidden danger location marking information based on the ditch bottom state anomaly level or slope stability risk level, and conducts slope instability probability analysis, which is pushed to the construction terminal, promoting the engineering decision-making closed loop of drone water conservancy ditch segmented excavation construction site inspection, and improving the safety and management efficiency of water conservancy project construction.

[0089] The above contents are merely examples and explanations of the concept of the present invention. Those skilled in the art may make various modifications or additions to the described specific embodiments or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, they should all fall within the scope of protection of the present invention.

Claims

1. A construction site drone intelligent inspection method, characterized in that: include: Plan the drone inspection route based on the construction progress map and load the preset intersection position between the completed section and the current operation section; Controlling the drone to fly along the inspection route, and in response to entering a completed section, switching the drone to a first acquisition mode, and generating ditch bottom abnormality feature data of the completed section in real time during point cloud acquisition; When the UAV arrives at the airspace near the preset intersection position, it verifies the boundary structure characteristics of the ditch to determine the validity of the preset intersection position, and confirms the actual intersection position based on the determination result; When the UAV crosses the actual boundary position and enters the current operation section, the UAV is switched to the second acquisition mode, and the slope risk characteristic data of the current operation section is generated in real time through edge waveform variation recognition; The inspection report is generated by integrating the ditch bottom abnormality feature data and the slope risk feature data. Based on the ditch bottom abnormality level or slope stability risk level in the report, an alarm instruction and hidden danger location marking information are generated and pushed to the construction terminal simultaneously; The airspace adjacent to the preset intersection position is defined as a rectangular airspace defined by the preset intersection position as a reference, extending in both forward and reverse directions along the flight trajectory to determine the length, determining the width by the real-time ditch bottom detection width, and determining the height by the current flight maintenance altitude; The verification of ditch boundary structural features includes: In the airspace adjacent to the preset intersection position, the trench bottom surface is divided into continuous strip-shaped units along a direction perpendicular to the trench centerline; Quantify and fuse the elevation gradient change rate, reflection intensity change rate, and image texture change rate between adjacent strip-shaped units to generate a comprehensive boundary feature index at the junction of each adjacent unit; The unit boundary position where the comprehensive boundary characteristic index exceeds the preset verification threshold is determined to be the mutation boundary position from the continuous hardened surface to the exposed soil.

2. The method for intelligent inspection of a construction site by using a drone according to claim 1, characterized in that: Planning the drone inspection route based on the construction progress map includes: Call the coordinate sequence of the ditch centerline of the construction progress map, project the centerline into the geographic coordinate system, bind the path starting point to the coordinate of the starting point of the completed section, and the end point to the coordinate of the latest updated point of the current construction section, and generate the initial reference path for the UAV; Control the drone to perform a pre-flight scan at the starting point of the initial reference path to identify obstacles that conflict with the planned trajectory; Horizontal obstacle avoidance offset correction is performed on the path nodes of the obstacle-affected section to generate the final planned inspection path, wherein the horizontal obstacle avoidance offset correction requires that the relative spatial relationship between the corrected section and the target detection area meets the preset acquisition constraint conditions.

3. The method for intelligent inspection of a construction site by using a drone according to claim 1, characterized in that: The first acquisition modality performs the following operations in order: Determine the flight altitude of the drone based on the trench depth data of the completed section; Control the UAV to maintain a low-disturbance flight attitude so that the emission direction of the laser scanning device on board is perpendicular to the ditch bottom plane; Perform reciprocating zigzag flight along the ditch axis to cover the full width of the ditch bottom, and simultaneously activate the laser scanning device to perform continuous panoramic scanning vertically downward; Panoramic optical image acquisition is triggered at a turning point of the path of the reciprocating zigzag flight, and the central area of ​​the optical image is spatially aligned with the central area of ​​the laser scanning.

4. The method for intelligent inspection of a construction site by using a drone according to claim 3, characterized in that: The real-time generation process of the abnormal characteristic data of the trench bottom of the completed section includes: When scanning the trench bottom of the completed section using a laser scanning device, the elevation deviation area on the trench bottom surface, the abnormal linear area of ​​reflection intensity, and the moisture-sensitive color gamut concentration area based on optical image recognition are detected in real time; When any abnormal area is detected, the full index correlation analysis of the abnormal area is automatically triggered to extract the normalized parameters of the abnormal area's elevation deviation, linear extension scale, and humidity area coverage; According to the ditch material type, a weight distribution rule is configured, and the normalized parameters are linearly weighted and fused to generate a ditch bottom anomaly evaluation index of the abnormal area; According to the abnormal level of each ditch bottom state stored in the WEB cloud and its preset evaluation index range, match the abnormal level of the ditch bottom state in the abnormal area and output the corresponding level label; The level label is bound to the spatial coordinates of the abnormal area to form the ditch bottom abnormality feature data.

5. The method for intelligent inspection of a construction site by using a drone according to claim 1, characterized in that: Determining the validity of the preset intersection position includes: Quantify the coordinate distance between the mutation junction position and the preset junction position; If the coordinate spacing is less than or equal to the preset tolerance spacing threshold, the preset intersection position is determined to be valid and the preset intersection position is confirmed to be the actual intersection position; otherwise, it is determined to be invalid and the sudden change intersection position is confirmed to be the actual intersection position.

6. The method for intelligent inspection of a construction site by using a drone according to claim 1, characterized in that: The second acquisition modality performs the following operations in sequence: Adjust the drone's yaw angle so that its fuselage axis is parallel to the slope, and synchronously control the gimbal pitch angle to match the slope inclination, keeping the synthetic aperture radar beam perpendicular to the slope. Control the UAV to perform variable pitch spiral flight, including: i. Real-time calculation of slope angle and dynamic adjustment of spiral pitch; ii. Maintaining a constant relative observation distance between the UAV and the slope based on laser ranging feedback; iii. Expand the spiral radius to cover the entire area from the top to the toe of the slope based on the geometric characteristics of the slope surface; During the flight, synthetic aperture radar and multispectral imaging are activated synchronously to obtain slope displacement time series data and moisture content distribution data.

7. The method for intelligent inspection of a construction site by using a drone according to claim 6, characterized in that: The real-time generation process of the slope risk characteristic data of the current operation section includes: Calculate the displacement vector field of adjacent time-series slope point clouds and mark the coordinate set of risk areas where the displacement rate exceeds the preset safety threshold; Construct a slope moisture distribution map and outline the areas where the moisture content exceeds the preset saturation threshold, and spatially superimpose it with the risk area coordinate set to generate a displacement-moisture content composite risk area; Based on the preset parameters of the slope material, the moisture content data and the area ratio of the composite risk area, the slope stability safety reduction factor of the composite risk area is calculated through geotechnical shear strength reduction simulation; Match the slope stability risk level of the composite risk area according to each slope stability risk level and its preset safety reduction factor range stored in the WEB cloud; The boundary coordinates of the composite risk area and the slope stability risk level are stored in association to generate slope risk characteristic data.

8. The method for intelligent inspection of a construction site by using a drone according to claim 7, characterized in that: After the slope risk characteristic data is generated, the following steps are also included: Collecting meteorological data within a preset time window of the composite risk area, including wind speed and precipitation intensity; Inputting the moisture content data of the composite risk area and meteorological data into a preset landslide prediction model to calculate the probability of slope instability; If the slope instability probability is greater than the preset warning probability threshold, a weather warning tag is added to the inspection report; With the center coordinates of the composite risk area as the center of the circle, a restricted area for construction machinery is planned with a preset radius, and emergency avoidance instructions are pushed to the construction terminal.

Images