CN-122018530-A - Method for improving unmanned aerial vehicle oblique photography measurement precision

CN122018530ACN 122018530 ACN122018530 ACN 122018530ACN-122018530-A

Abstract

The invention relates to the technical field of photogrammetry, in particular to a method for improving unmanned aerial vehicle oblique photogrammetry accuracy, which comprises the steps of obtaining hardware information, environment data and working data of an unmanned aerial vehicle; the method comprises the steps of constructing a hyper-parameter space, carrying out gridding processing, initializing grid density based on environment data and hardware information to obtain the gridding hyper-parameter space, carrying out iterative hyper-parameter search in the gridding hyper-parameter space, carrying out local adjustment on the grid density in the search process, calculating target coefficients under each iteration number according to the search process so as to determine an optimal hyper-parameter matrix by utilizing the target coefficients, and controlling the unmanned aerial vehicle by utilizing the optimal hyper-parameter matrix so as to obtain image data under the optimal hyper-parameter matrix and uploading the image data. The invention effectively improves the measurement precision of unmanned aerial vehicle oblique photography.

Inventors

LI CHUANFENG
LIU ZHICHUN
ZHANG YIRAN
HAN ZHE
LI FEI
LI CHENHANG
SUN ZEYU
WANG GUOQIANG
LU DAN
SHAO HONGXIANG
GAO XIANG

Assignees

洛阳理工学院

Dates

Publication Date: 20260512
Application Date: 20260414

Claims (10)

1. A method for improving unmanned aerial vehicle oblique photography measurement accuracy, the method comprising the steps of: Acquiring hardware information, environment data and working data of the unmanned aerial vehicle; constructing a hyper-parameter space and performing gridding treatment, and initializing the grid density based on the environment data and the hardware information to obtain the gridding hyper-parameter space; Performing iterative super-parameter search in a gridding super-parameter space, locally adjusting the grid density in the search process, and calculating target coefficients under each iteration number according to the search process, so as to determine an optimal super-parameter matrix by using the target coefficients; And controlling the unmanned aerial vehicle by using the optimal super-parameter matrix, thereby acquiring and uploading the image data under the optimal super-parameter matrix.
2. The method for improving the unmanned aerial vehicle oblique photography measurement precision according to claim 1, wherein the steps of constructing a hyper-parameter space and performing gridding processing, initializing grid density based on environmental data and hardware information to obtain the gridding hyper-parameter space comprise the following specific steps: Constructing a super-parameter space, and determining the space size of the super-parameter space according to the environmental data and the hardware information of the unmanned aerial vehicle; And performing gridding processing on the super-parameter space, and initializing the grid density by utilizing the environment data and the hardware information.
3. The method for improving the oblique photography measurement accuracy of the unmanned aerial vehicle according to claim 2, wherein the steps of constructing the hyper-parameter space and determining the space size of the hyper-parameter space according to the environmental data and the hardware information of the unmanned aerial vehicle comprise the following specific steps: Creating a super-parameter space, wherein the super-parameter space corresponds to a space matrix, and elements in the space matrix comprise flying heights of the unmanned aerial vehicle and inclination angles of 4 inclined cameras; And initializing a space matrix of the hyper-parameter space based on the hardware information of the unmanned aerial vehicle.
4. The method for improving the accuracy of unmanned aerial vehicle oblique photography measurement according to claim 2, wherein the steps of gridding the hyper-parameter space and initializing the grid density by using the environmental data and the hardware information comprise the following specific steps: Respectively presetting a flight altitude step length and an inclination angle step length, and equally dividing corresponding dimensions in a super-parameter space through the flight altitude step length and the inclination angle step length to form an initial grid point set, wherein the initial grid point set comprises a plurality of initial grid points, each grid point corresponds to a matrix and is marked as a super-parameter matrix, and the dimension of the super-parameter matrix is identical to that of the super-parameter space; and calculating grid coefficients according to the environmental data and the hardware information of the unmanned aerial vehicle, and dynamically adjusting the space of each dimension in the super-parameter space based on the grid coefficients to obtain the grid super-parameter space at the current moment.
5. The method for improving unmanned aerial vehicle tilt photogrammetry accuracy according to claim 4, wherein the calculating the grid coefficient according to the environmental data and the hardware information of the unmanned aerial vehicle comprises the following specific steps: The method comprises the steps of obtaining atmospheric pressure data, wind speed data and wind direction data in environment data and electric quantity data of an unmanned aerial vehicle, carrying out standardized processing, presetting a prediction length, respectively predicting the atmospheric pressure data, the wind speed data and the electric quantity data through an ARIMA model to obtain prediction values respectively corresponding to the atmospheric pressure data, the wind speed data and the electric quantity data after the current moment, obtaining a plurality of data points which are in a neighborhood radius range and take the current moment as a center as neighborhood data points of the current moment, and calculating grid coefficients at the current moment according to the change fluctuation conditions of the neighborhood data points respectively corresponding to the atmospheric pressure data, the wind speed data and the electric quantity data at the current moment.
6. The method for improving the oblique photography measurement accuracy of the unmanned aerial vehicle according to claim 4, wherein the step of dynamically adjusting the distance between the dimensions in the hyper-parameter space based on the grid coefficient to obtain the grid hyper-parameter space at the current time comprises the following specific steps: Will be As step length coefficient, multiplying the step length coefficient with the flying height step length and the inclination angle step length respectively to obtain gridding super parameter space at the current time, wherein Indicating the current time The grid coefficient below; An exponential function based on a natural constant is represented.
7. The method for improving the unmanned aerial vehicle oblique photography measurement precision according to claim 1, wherein the iterative super-parameter search is performed in a gridding super-parameter space, and the grid density is locally adjusted in the search process, comprising the following specific steps: Taking each superparameter combination in the gridding superparameter space as a grid point, and uniformly setting the grid points in the gridding superparameter space The method comprises the steps that the longicorn uses grid points as position points when each iteration advances in a gridding hyper-parameter space; setting the advancing step length of each iteration of longicorn as Grid points, wherein For the preset number of longhorns, When the longhorn beetles search in the gridding hyper-parameter space, a matrix formed by each hyper-parameter at the corresponding grid point is obtained and marked as a hyper-parameter matrix under the corresponding iteration times; Inputting a super-parameter matrix under the corresponding iteration times into a control module of the unmanned aerial vehicle in the digital simulation model, performing parameter adjustment on components corresponding to each element in the super-parameter matrix, performing geometric projection simulation based on a Digital Elevation Model (DEM) and camera parameters, calculating theoretical overlapping degree, recording the theoretical overlapping degree under the corresponding iteration times, acquiring a difference absolute value of the theoretical overlapping degree and the reference overlapping degree, and calculating a quality coefficient under the corresponding super-parameter matrix according to the difference absolute value, wherein the difference absolute value and the quality coefficient are in negative correlation; And locally adjusting the grid point density of the gridding super-parameter space according to the change condition of the corresponding quality coefficient in the position iteration process of each longicorn in the gridding super-parameter space.
8. The method for improving the oblique photography measurement accuracy of the unmanned aerial vehicle according to claim 7, wherein the local adjustment of the grid point density of the gridding super parameter space according to the change condition of the corresponding quality coefficient in the position iteration process of each longicorn in the gridding super parameter space comprises the following specific steps: The method comprises the steps of obtaining a sequence formed by corresponding quality coefficients of any longicorn at all iteration times before the current moment, taking the sequence as the quality coefficient sequence of the longicorn at the current moment, calculating a search coefficient of the longicorn at the corresponding iteration times at the current moment according to the variation trend of element numerical values in the quality coefficient sequence, presetting a search radius, and respectively multiplying the flight height step length and the inclination angle step length between grids in the search radius by the search coefficient to obtain a local adjustment result of a grid hyper-parameter space.
9. The method for improving the unmanned aerial vehicle tilt photogrammetry accuracy according to claim 1, wherein the calculating the target coefficient for each iteration number according to the search process, so as to determine the optimal super-parameter matrix by using the target coefficient, comprises the following specific steps: Acquiring the quality coefficients of all the longicorn in the grid hyper-parameter space, and calculating the target coefficient under each iteration number by combining the position distribution of the longicorn in the grid hyper-parameter space; The method comprises the steps of constructing a two-dimensional rectangular coordinate system, taking iteration times as a transverse axis of the two-dimensional rectangular coordinate system, taking a target coefficient as a longitudinal axis of the two-dimensional rectangular coordinate system, obtaining a scatter diagram formed by corresponding target coefficients in the two-dimensional rectangular coordinate system under all iteration times, determining inflection points in the scatter diagram by using an elbow method, recording the iteration times corresponding to the inflection points as target iteration times, and obtaining a super-parameter matrix of the longicorn corresponding to the maximum value in all search coefficients under the target iteration times in a grid super-parameter space to serve as an optimal super-parameter matrix.
10. The method for improving the oblique photography measurement precision of the unmanned aerial vehicle according to claim 1, wherein the unmanned aerial vehicle is controlled by utilizing the optimal super-parameter matrix so as to acquire and upload the image data under the optimal super-parameter matrix, comprises the following specific steps: Inputting an optimal super-parameter matrix into a flight control system of the unmanned aerial vehicle, so as to analyze the optimal super-parameter matrix and extract various control parameters contained in the optimal super-parameter matrix, wherein the optimal super-parameter matrix comprises a flight altitude parameter and respective inclination angle parameters of four inclination cameras; after the flight state of the unmanned aerial vehicle and the gestures of each camera reach preset requirements, a main control unit sends out a synchronous shooting instruction, an AES-256 encryption algorithm is adopted to encrypt and package the image file, and the image file is uploaded to a ground station server or a cloud storage platform through a communication module carried by the unmanned aerial vehicle.

Description

Method for improving unmanned aerial vehicle oblique photography measurement precision Technical Field The invention relates to the technical field of photogrammetry, in particular to a method for improving unmanned aerial vehicle oblique photogrammetry accuracy. Background Because oblique photogrammetry can efficiently acquire multi-view images of ground objects and construct a high-precision three-dimensional geographic model, the method is widely applied to the fields of city modeling, topographic mapping, smart cities and the like. However, in practical application, the traditional oblique photography method is easily subjected to problems of incomplete image coverage, image blurring, difficult matching and the like under a complex environment due to the influence of topographic relief, meteorological condition change and unmanned aerial vehicle hardware performance limitation, so that the accuracy of the three-dimensional reconstruction model is reduced. In the prior art, a shooting mode with fixed flying height and camera inclination angle is generally adopted, the self-adaptive adjustment capability for environmental dynamic change and system hardware constraint is lacked, and the imaging quality and measurement precision in different scenes are difficult to consider. In addition, although a parameter optimization mechanism is introduced in a part of schemes, the schemes depend on global traversal search or static preset parameter libraries, the calculation cost is high, the instantaneity is poor, the parameter adjustment is blind and the efficiency is low, and particularly in a changeable field environment, how to combine unmanned aerial vehicle hardware configuration, real-time environment state and working parameters to construct a dynamic parameter optimization mechanism with environment adaptability becomes a key problem for improving the oblique photogrammetry. Disclosure of Invention The invention provides a method for improving unmanned aerial vehicle oblique photography measurement accuracy, which aims to solve the existing problems. The method for improving the unmanned aerial vehicle oblique photography measurement precision adopts the following technical scheme: One embodiment of the present invention provides a method for improving unmanned aerial vehicle tilt photogrammetry, the method comprising the steps of: Acquiring hardware information, environment data and working data of the unmanned aerial vehicle; constructing a hyper-parameter space and performing gridding treatment, and initializing the grid density based on the environment data and the hardware information to obtain the gridding hyper-parameter space; Performing iterative super-parameter search in a gridding super-parameter space, locally adjusting the grid density in the search process, and calculating target coefficients under each iteration number according to the search process, so as to determine an optimal super-parameter matrix by using the target coefficients; And controlling the unmanned aerial vehicle by using the optimal super-parameter matrix, thereby acquiring and uploading the image data under the optimal super-parameter matrix. Optionally, the building of the hyper-parameter space and the gridding processing are performed, and the grid density is initialized based on the environmental data and the hardware information to obtain the gridding hyper-parameter space, which comprises the following specific methods: Constructing a super-parameter space, and determining the space size of the super-parameter space according to the environmental data and the hardware information of the unmanned aerial vehicle; And performing gridding processing on the super-parameter space, and initializing the grid density by utilizing the environment data and the hardware information. Optionally, the method for constructing the hyper-parameter space and determining the space size of the hyper-parameter space according to the environmental data and the hardware information of the unmanned aerial vehicle includes the following specific steps: Creating a super-parameter space, wherein the super-parameter space corresponds to a space matrix, and elements in the space matrix comprise flying heights of the unmanned aerial vehicle and inclination angles of 4 inclined cameras; And initializing a space matrix of the hyper-parameter space based on the hardware information of the unmanned aerial vehicle. Optionally, the method for performing gridding processing on the hyper-parameter space and initializing the grid density by using the environmental data and the hardware information includes the following specific steps: Respectively presetting a flight altitude step length and an inclination angle step length, and equally dividing corresponding dimensions in a super-parameter space through the flight altitude step length and the inclination angle step length to form an initial grid point set, wherein the initial grid point set comprises a plur