Water conservancy project three-dimensional modeling method based on unmanned aerial vehicle remote sensing technology

Abstract

The invention discloses a water conservancy project three-dimensional modeling method based on an unmanned aerial vehicle remote sensing technology, and relates to the technical field of water conservancy project surveying and mapping and three-dimensional modeling. An unmanned aerial vehicle carries a high-resolution optical camera, a multi-line laser radar and a multispectral sensor to synchronously work under the control of a time synchronization mechanism; synchronous data acquisition of high-resolution images, high-precision point clouds and surface reflectance data is realized, the limitation of a single data source is made up, the fusion precision of various acquired data is improved, the precision of generating a three-dimensional terrain model in the later period is improved, and the control instruction output by the reinforcement learning algorithm module is used for realizing the control of the three-dimensional terrain model. Flight parameters of the unmanned aerial vehicle are adjusted in real time, flight stability is improved, surveying and mapping requirements of complex terrains such as water areas and vegetation coverage areas in water conservancy projects are met, a generated water conservancy project three-dimensional model is compared with ground control point data through a precision verification module, precision evaluation is conducted on the three-dimensional model, and a subsequent data acquisition strategy is optimized.

Description

TECHNICAL FIELD

[0001] The application relates to the field of water conservancy engineering surveying and mapping and three-dimensional modeling technology, and particularly relates to a three-dimensional modeling method for water conservancy engineering based on unmanned aerial vehicle remote sensing technology. BACKGROUND

[0002] Water conservancy engineering refers to the engineering for improving water resource utilization, protecting hydrological environment, improving water productivity, and preventing water disasters, etc. At present, three-dimensional modeling of water conservancy engineering mainly relies on traditional surveying and mapping technologies such as total station, RTK, GPS and the like to obtain high-precision single-point coordinates, improve design efficiency and reduce risks. The prior art with the publication number CN119879860A discloses a water conservancy engineering unmanned aerial vehicle oblique photography and three-dimensional modeling method. A first height is set, and an unmanned aerial vehicle flies at the first height to obtain a first image set in a selected aerial photography area. The unmanned aerial vehicle flies sideways outside an obstacle in the first image set at the first height. A second height is set for a lateral height, and an optimal aerial photography path of the unmanned aerial vehicle is selected at the second height. The unmanned aerial vehicle collects multi-radiation topography in the optimal aerial photography path multiple times to obtain a second image set. A topography model is built by using a neural network according to the second image set, thereby solving the problem that, when the existing unmanned aerial vehicle oblique photography technology is used to obtain the topography structure information of such an area, a large amount of image data collected is difficult to accurately match and feature recognition is difficult due to the existence of numerous irregular, complex and changeable elements.

[0003] The above-mentioned technical solution has the following deficiencies: the data processing is relatively complex, the model precision is affected by flight parameters, the traditional single sensor data is difficult to consider both image texture and topographic elevation information, and the fixed flight parameters cannot meet the surveying and mapping requirements of complex terrains such as water areas and vegetation coverage areas in water conservancy engineering. SUMMARY

[0004] Therefore, in order to solve the above-mentioned deficiencies, the application provides a three-dimensional modeling method for water conservancy engineering based on unmanned aerial vehicle remote sensing technology.

[0005] The application provides the following technical solution: a three-dimensional modeling method for water conservancy engineering based on unmanned aerial vehicle remote sensing technology, comprising the following steps: Step 1: controlling an unmanned aerial vehicle system, driving a power module to drive the unmanned aerial vehicle to fly through a flight controller in the unmanned aerial vehicle system, and outputting control instructions from a reinforcement learning algorithm module in the unmanned aerial vehicle system to the flight controller to adjust the flight parameters of the unmanned aerial vehicle in real time during the flight, and the unmanned aerial vehicle carries a remote sensing sensor group to perform a flight task; Step 2: The remote sensing sensor group flies synchronously with the UAV to capture data of the water conservancy project topography. The high-resolution optical camera, multi-line laser radar, and multi-spectral sensor in the remote sensing sensor group work synchronously through a time synchronization mechanism to realize synchronous data collection of high-resolution images, high-precision point clouds, and surface reflectivity data, and to compensate for the limitations of a single data source. Step 3: The remote sensing sensor group transmits the captured data to the data processing platform. The image processing software in the data processing platform preprocesses the remote sensing data by denoising, stitching, and radiation correction. The three-dimensional modeling software in the data processing platform generates a three-dimensional terrain model and a texture model from the preprocessed data.

[0006] Preferably, the reinforcement learning algorithm module constructs a state space, an action space, and a reward function with terrain complexity, sensor data quality, and modeling accuracy as optimization objectives, and generates an optimal flight parameter control strategy through iterative training.

[0007] Preferably, the state space includes current terrain type slope, vegetation coverage, wind speed, and point cloud density. The action space includes altitude adjustment, airspeed setting, and overlap rate setting. The reward function quantitatively calculates the modeling accuracy and data integrity. The reward function is: is the weight coefficient, and the modeling quality = point cloud integrity x precision error reciprocal.

[0008] Preferably, the energy consumption includes airspeed penalty and height adjustment penalty. The airspeed penalty calculation formula is: airspeed penalty = 0.01 x V 2 , where V is the UAV airspeed (m / s). The height adjustment penalty calculation formula is: height adjustment penalty = 0.1 x |ΔH|, where H is the UAV flight height (m).

[0009] Preferably, the time synchronization mechanism in step 2 includes a GPS clock module, a reference clock module, and a synchronous trigger circuit. The GPS clock module calculates the UTC time and the 1PPS signal, with the 1PPS rising edge aligned with the UTC time second boundary, with an accuracy of ±100 nanoseconds. The phase-locked loop circuit locks the frequency and phase of the reference clock to the 1PPS signal of the GPS. The synchronous trigger circuit generates multiple synchronous pulses to simultaneously trigger the high-resolution optical camera, multi-line laser radar, and multi-spectral sensor to start data collection, ensuring that the initial time of the high-resolution optical camera, multi-line laser radar, and multi-spectral sensor is aligned.

[0010] Preferably, the GPS clock module further comprises a calibration module and a dynamic compensation module, the calibration module performs clock calibration every 100 milliseconds, calculates the frequency offset Δf by comparing the 1PPS signal of GPS with the second pulse of the reference clock, and dynamically adjusts through the phase-locked loop, and the dynamic compensation module is used for correcting the frequency drift of the crystal oscillator caused by temperature change in the flight of the unmanned aerial vehicle in real time through the temperature compensation algorithm of the GPS clock module.

[0011] Preferably, the multi-spectral sensor is used for collecting visible light and near-infrared wave band data, obtaining vegetation coverage, water turbidity ecological parameters, and fusing to generate a multi-dimensional water conservancy engineering information model with the three-dimensional terrain model.

[0012] Preferably, the data processing platform further comprises a precision verification module, which compares with ground control point data, multi-source data mutual verification, water conservancy feature inspection, precision evaluation of the three-dimensional model, and feedback of the evaluation results to the reinforcement learning algorithm module, and optimization of subsequent data acquisition strategy.

[0013] Compared with the prior art, the application provides a water conservancy engineering three-dimensional modeling method based on unmanned aerial vehicle remote sensing technology, which has the following beneficial effects: The water conservancy engineering three-dimensional modeling method based on unmanned aerial vehicle remote sensing technology of the application realizes synchronous data acquisition of high-resolution images, high-precision point clouds and surface reflectivity data under the control of a time synchronization mechanism, makes up for the limitations of a single data source, improves the accuracy of the fusion of the collected various data, improves the accuracy of the generation of the three-dimensional terrain model, adjusts the flight parameters of the unmanned aerial vehicle in real time through the control instructions output by the reinforcement learning algorithm module, improves the flight stability, and adapts to the surveying and mapping requirements of complex terrains such as water areas and vegetation coverage areas in water conservancy projects, compares the generated water conservancy engineering three-dimensional model with ground control point data through the precision verification module, evaluates the precision of the three-dimensional model, and optimizes the subsequent data acquisition strategy. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 is a working process schematic diagram of the application; Figure 2 is a schematic diagram of the unmanned aerial vehicle system, remote sensing sensor group and data processing platform framework of the application; Figure 3 is a state space framework schematic diagram of the application; Figure 4 is a schematic diagram of the reinforcement learning algorithm module adjusting the flight mechanism of the unmanned aerial vehicle of the application; Figure 5 is a reward calculation framework schematic diagram of the application; Figure 6The precision verification module precision evaluation schematic diagram of the present application.

[0015] Figure 7 The time synchronization mechanism flow chart of the present application. DETAILED DESCRIPTION

[0016] The embodiments of the present application will be described below in detail with reference to the accompanying drawings. Figures 1-7 The embodiments of the present application will be described below in detail with reference to the accompanying drawings.

[0017] Please refer to Figures 1-7 The present application is a water conservancy project three-dimensional modeling method based on unmanned aerial vehicle remote sensing technology, comprising the following steps: Step 1: Control the unmanned aerial vehicle system, drive the power module to drive the unmanned aerial vehicle to fly through the flight controller in the unmanned aerial vehicle system, and output the control instruction of the reinforcement learning algorithm module in the unmanned aerial vehicle system to the flight controller during the flight process, and adjust the flight parameters of the unmanned aerial vehicle in real time, and the unmanned aerial vehicle carries out the flight task by the remote sensing sensor group; The reinforcement learning algorithm module takes the terrain complexity, sensor data quality and modeling accuracy as the optimization target, constructs the state space, action space and reward function, and generates the optimal flight parameter control strategy through iterative training; The state space includes the current terrain type slope, vegetation coverage, wind speed and point cloud density; According to the terrain data, the slope is calculated, the coverage is calculated according to the vegetation data, the wind speed is monitored in real time through the weather station, and the density is analyzed through the point cloud sensor; The action space includes altitude adjustment, speed setting and overlap rate setting; The initial altitude is 100 meters by default, the initial speed is 20 m / s, and the initial overlap rate is 70%; The reward function is quantitatively calculated according to the modeling accuracy and data integrity, and the reward function is: is the weight coefficient, and the modeling quality = point cloud integrity x precision error reciprocal; The energy consumption includes speed penalty and height adjustment penalty, and the speed penalty refers to: the speed V(m / s) of the unmanned aerial vehicle and the power consumption are in a square relationship, and high-speed flight will greatly increase the energy consumption, and the purpose of the penalty is: to avoid excessive pursuit of efficiency and ignore the endurance capability, and the calculation formula is: speed penalty = 0.01 x V 2 , example: when V = 20(m / s), the penalty value = 0.01 x 20 2= 4 (m / s) ; height adjustment penalty refers to: height change H (m) needs to overcome gravity work W, frequent lifting significantly increases energy consumption, the purpose of the penalty: inhibit unnecessary height fluctuations, improve flight stability, the calculation formula: height adjustment penalty = 0.1 x |AH|, example: climb 5 (m) : penalty value = 0.1 x 5 = 0.5 (m) ; During data acquisition, the flight height of the flat terrain is maintained at a height of 80-100 m, and when water areas or steep slope areas are detected, the reinforcement learning algorithm module automatically reduces the flight height to 10-30 m and reduces the airspeed to 5-15 m / s.

[0018] Step 2: The remote sensing sensor group is driven by the unmanned aerial vehicle system to fly synchronously to collect data on the water conservancy project topography; the remote sensing sensor group includes a high-resolution optical camera, a multi-line laser radar, a multi-spectral sensor, and a gimbal fixed to the bottom of the unmanned aerial vehicle; through a time synchronization mechanism, the high-resolution optical camera, the multi-line laser radar, and the multi-spectral sensor work synchronously to achieve synchronous data acquisition of high-resolution images, high-precision point clouds, and ground reflectivity data, making up for the limitations of a single data source; the high-resolution optical camera acquires positioning data (accuracy ±1 cm) through PPK / RTK at the moment of exposure; through the cooperation of the multi-line laser radar and the high-resolution optical camera, the elevation information of the laser point cloud and the texture information of the optical image can be combined to ensure the density of the point cloud while improving the visualization effect of the model; and through gimbal stabilization control, the attitude angle error of the remote sensing sensor group is ≤0.1°, improving the stability of the remote sensing sensor group during data acquisition; The multi-line laser radar emits laser pulses, which are reflected back to the receiver when encountering an object; by calculating the time difference T between emission and reception and the speed of light C (3 x 10 8 m / s), the distance The multi-line laser radar performs vertical scanning by arranging multiple laser emitters (such as 16 lines or 32 lines) at fixed angular intervals, simultaneously emitting laser beams covering different pitch angles (such as the Velodyne VLP-16, which has a viewing angle of -15° to +15°); horizontal scanning is achieved by rotating the laser through a motor, allowing for 360° horizontal scanning of the multi-line laser radar; The multi-spectral sensor is used to collect visible light and near-infrared band data to obtain vegetation coverage, water turbidity, and ecological parameters, and to generate a multi-dimensional water conservancy information model by integrating with the three-dimensional terrain model; the blue band in the multi-spectral sensor has a wavelength range of 450-515 (nm) and is used to monitor water turbidity, with a penetration depth of ≤5 m; the green band has a wavelength range of 520-600 (nm) and is used to monitor the concentration of chlorophyll; the red band has a wavelength range of 630-690 (nm) and is used to monitor the distribution of suspended solids; and the near-infrared band has a wavelength range of 760-900 (nm) and is used to identify vegetation coverage / water boundaries; The flight parameters of the unmanned aerial vehicle are adjusted through the reinforcement learning algorithm module, so that the image overlap rate of the remote sensing sensor group is increased to 80%-90%, the data acquisition density is enhanced, the flight height, the flight speed and the image overlap rate of the unmanned aerial vehicle are dynamically adjusted in combination with the water area, the steep slope and the vegetation coverage area and other water conservancy engineering terrain features, and the full coverage of the complex area is ensured; The time synchronization mechanism comprises a GPS clock module, a reference clock module and a synchronization trigger circuit. The GPS clock module calculates UTC time and a 1PPS signal. The rising edge of the 1PPS is aligned with the second boundary of the UTC time, and the precision is ±100 nanoseconds. The frequency and phase of the reference clock are locked to the 1PPS signal of the GPS through a phase-locked loop circuit. The synchronization trigger circuit generates a plurality of synchronization pulses to simultaneously trigger the high-resolution optical camera, the multi-line laser radar and the multi-spectral sensor to start data acquisition, so as to ensure that the initial time of the high-resolution optical camera, the multi-line laser radar and the multi-spectral sensor is aligned. The GPS clock module further comprises a calibration module and a dynamic compensation module. The calibration module performs clock calibration every 100 milliseconds. The frequency offset Δf is calculated by comparing the 1PPS signal of the GPS with the second pulse of the reference clock, and the dynamic adjustment is performed through the phase-locked loop. The dynamic compensation module is used to compensate the frequency drift of the crystal oscillator caused by temperature change during the flight of the unmanned aerial vehicle, and the real-time correction is performed through the temperature compensation algorithm of the GPS clock module. The mathematical model of the phase-locked loop is as follows: ftou(t)=fref+Kp\cdote(t)+K\inte(t),dt, wherein fref=10MHz. The reference frequency e(t)=t1PPS-tosc. The time difference between the 1PPS and the crystal oscillator Kp, Ki. The proportional and integral gain coefficients, typical values: Kp=10 -6 , Ki=10 -9 .

[0019] During the data acquisition process, the calibration module continuously calibrates the GPS clock module to ensure that the synchronization trigger circuit simultaneously triggers the high-resolution optical camera, the multi-line laser radar and the multi-spectral sensor to start data acquisition. In this process, when the external environment is-20℃-60℃, the crystal oscillator frequency drift is compensated in real time through the dynamic compensation module, so as to avoid the influence of temperature on the high-resolution optical camera, the multi-line laser radar and the multi-spectral sensor, so that the collected various data cannot be accurately fused, and the accuracy of the three-dimensional terrain model generated in the later stage is reduced. Step 3: The remote sensing sensor group transmits the data collected by the line shooting to the data processing platform, and the data processing platform integrates the remote sensing data and generates a three-dimensional terrain model; the data processing platform includes image processing software and three-dimensional modeling software, the image processing software performs denoising, splicing, and radiation correction preprocessing on the remote sensing data, the three-dimensional modeling software generates a three-dimensional terrain model and a texture model based on the preprocessed data, the data processing platform further includes a precision verification module, which compares with the ground control point data to evaluate the precision of the three-dimensional model, and feeds back the evaluation result to the reinforcement learning algorithm module to optimize the subsequent data collection strategy; denoising includes: removing the interference of the unmanned aerial vehicle circuit (manifested as isolated bright / dark points); stripe noise: inter-line difference of scanning sensors; wave reflection: water surface mirror reflection; radiation correction (eliminate light difference) purpose: convert the gray value (DN) recorded by the sensor into the real ground reflectivity.

[0020] The above description of disclosed embodiments enables one of ordinary skill in the art to make or use the application. Various modifications to these embodiments will be readily apparent to those of ordinary skill in the art, and the generic principles defined herein can be applied to other embodiments without departing from the spirit or scope of the application. Accordingly, the application is not to be limited to these embodiments shown herein but is to be accorded the broadest scope consistent with the principles and novel features disclosed herein.

Claims

1. The present invention discloses a 3D modeling method for water conservancy projects based on unmanned aerial vehicle (UAV) remote sensing technology, which relates to the technical field of surveying and mapping and 3D modeling of water conservancy projects. By using an unmanned aerial vehicle (UAV) equipped with a high-resolution optical camera, a multi-line laser radar, and a multispectral sensor, the UAV operates synchronously under the control of a time synchronization mechanism to achieve synchronous data collection of high-resolution images, high-precision point clouds, and surface reflectivity data, thereby overcoming the limitations of a single data source, improving the accuracy of fusing the collected multiple data, and improving the accuracy of the subsequent generation of a 3D terrain model. By using the control instructions output by the reinforcement learning algorithm module, the flight parameters of the UAV are adjusted in real time to improve flight stability, and the method adapts to the surveying and mapping needs of complex terrains such as water areas and vegetation-covered areas in water conservancy projects. By using an accuracy verification module, the generated 3D model of the water conservancy project is compared with the ground control point data, the accuracy of the 3D model is evaluated, and the subsequent data collection strategy is optimized.

Images