CN114612806A - Method for improving DEM product precision of consumption-level unmanned aerial vehicle - Google Patents
Method for improving DEM product precision of consumption-level unmanned aerial vehicle Download PDFInfo
- Publication number
- CN114612806A CN114612806A CN202210117130.8A CN202210117130A CN114612806A CN 114612806 A CN114612806 A CN 114612806A CN 202210117130 A CN202210117130 A CN 202210117130A CN 114612806 A CN114612806 A CN 114612806A
- Authority
- CN
- China
- Prior art keywords
- point
- ground
- dem
- elevation
- unmanned aerial
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000001914 filtration Methods 0.000 claims abstract description 55
- 230000002159 abnormal effect Effects 0.000 claims abstract description 51
- 239000004744 fabric Substances 0.000 claims abstract description 22
- 238000012545 processing Methods 0.000 claims description 15
- 230000006872 improvement Effects 0.000 claims description 11
- 239000002245 particle Substances 0.000 claims description 9
- 230000009471 action Effects 0.000 claims description 8
- 238000006073 displacement reaction Methods 0.000 claims description 8
- 238000005516 engineering process Methods 0.000 claims description 7
- 230000008569 process Effects 0.000 claims description 5
- 230000008859 change Effects 0.000 claims description 4
- 230000005484 gravity Effects 0.000 claims description 4
- 239000007787 solid Substances 0.000 claims description 4
- 238000012935 Averaging Methods 0.000 claims description 2
- 238000005259 measurement Methods 0.000 abstract description 6
- 238000004088 simulation Methods 0.000 abstract description 4
- 230000000694 effects Effects 0.000 description 27
- 238000011156 evaluation Methods 0.000 description 7
- 238000005070 sampling Methods 0.000 description 7
- 238000000926 separation method Methods 0.000 description 6
- 238000005457 optimization Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 241000196324 Embryophyta Species 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011002 quantification Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 241000084490 Esenbeckia delta Species 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010835 comparative analysis Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 239000011521 glass Substances 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C11/00—Photogrammetry or videogrammetry, e.g. stereogrammetry; Photographic surveying
- G01C11/02—Picture taking arrangements specially adapted for photogrammetry or photographic surveying, e.g. controlling overlapping of pictures
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S19/00—Satellite radio beacon positioning systems; Determining position, velocity or attitude using signals transmitted by such systems
- G01S19/38—Determining a navigation solution using signals transmitted by a satellite radio beacon positioning system
- G01S19/39—Determining a navigation solution using signals transmitted by a satellite radio beacon positioning system the satellite radio beacon positioning system transmitting time-stamped messages, e.g. GPS [Global Positioning System], GLONASS [Global Orbiting Navigation Satellite System] or GALILEO
- G01S19/42—Determining position
- G01S19/43—Determining position using carrier phase measurements, e.g. kinematic positioning; using long or short baseline interferometry
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Computer Graphics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Processing Or Creating Images (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a method for improving the DEM product precision of a consumption-level unmanned aerial vehicle, which comprises the following steps: acquiring an image map, a three-dimensional point cloud and a three-dimensional model of a region to be detected by a consumption-level unmanned aerial vehicle; constructing a gradient-distribution filtering model to filter the three-dimensional point cloud to obtain a ground point cloud, and constructing a ground seed DEM through the ground point cloud; determining the elevations of certain density characteristic points reasonably distributed in the area to be measured through GNSS RTK, calculating the elevation difference between the GNSS RTK measurement of the characteristic points and the unmanned aerial vehicle measurement, and constructing an elevation abnormal value curved surface model of the area to be measured, wherein the elevation difference is hereinafter referred to as an elevation abnormal value; and then, compensating and correcting the ground seed DEM by using the elevation abnormal value curved surface model, and finally improving the product precision of the consumption-level unmanned aerial vehicle DEM. According to the invention, the accuracy of the DEM product of the consumption-level unmanned aerial vehicle is obviously improved through the constructed gradient-cloth simulation filtering model and the constructed elevation abnormal value curved surface model for compensation and correction.
Description
Technical Field
The invention belongs to the technical field of unmanned aerial vehicles, and particularly relates to a method for improving DEM product precision of a consumption-level unmanned aerial vehicle.
Background
As a part of modern Digital map 4D products, Digital Elevation Models (DEMs) are used to promote rapid development of geographic information industry, and how to construct high-quality DEMs by using collected topographic data is paid attention by many researchers. Compared with the traditional remote sensing technology, the UAV photogrammetry can acquire a large amount of point cloud data in a short time and accurately acquire three-dimensional coordinate information of the earth surface. The method provides data support for constructing DEM and Digital Surface Model (DSM) products, and is an important data source for high-resolution Surface modeling.
At present, a low-altitude Unmanned Aerial Vehicle (UAV) aerial survey system is widely applied to the aspects of acquisition of a digital map 4D product, map surveying and drawing, a digital line graph and the like, however, the problems of low precision, weak practicability and the like exist when a digital elevation model product is constructed by using data acquired by a consumer-grade UAV, and therefore how to improve the DEM product precision acquired by the consumer-grade UAV is a hotspot problem and a difficult problem in research of the UAV aerial survey system.
Therefore, a new method for improving the accuracy of the DEM product of the consumer-grade unmanned aerial vehicle is needed to solve the technical problem.
Disclosure of Invention
The invention aims to solve the problems and provide a method for improving the DEM product precision of a consumption-level unmanned aerial vehicle.
The invention achieves the above purpose through the following technical scheme:
a method for improving the precision of a DEM product of a consumption-level unmanned aerial vehicle comprises the following steps:
acquiring an image map, a three-dimensional point cloud and a three-dimensional model of a region to be detected by a consumption-level unmanned aerial vehicle;
constructing a gradient-distribution filtering model to filter the three-dimensional point cloud to obtain a ground point cloud, and constructing a ground seed DEM through the ground point cloud;
selecting certain density characteristic points which are uniformly distributed in the area to be measured, measuring the plane coordinates and the elevation abnormal values of the characteristic points through a GNSS RTK technology, fitting the elevation abnormal values of the characteristic points by utilizing a quadratic polynomial surface model, and constructing a surface model of the elevation abnormal values of the area to be measured;
and compensating and correcting the ground seed DEM by using the elevation abnormal value curved surface model, thereby realizing the improvement of the product precision of the consumption-level unmanned aerial vehicle DEM.
As a further optimization scheme of the invention, the specific steps of acquiring the image map, the three-dimensional point cloud and the three-dimensional model of the area to be measured by the consumer-grade unmanned aerial vehicle are as follows:
laying image control points in the region to be detected;
acquiring an aerial photograph of the area to be measured by a consumption-level unmanned aerial vehicle;
and matching and processing the aerial image and the image control points by using Smart3D Capture unmanned aerial vehicle remote sensing image processing software to obtain an image map, a three-dimensional point cloud and a three-dimensional model of the area to be measured.
As a further optimization scheme of the present invention, the specific steps of filtering the three-dimensional point cloud by the gradient-distribution filtering model to obtain the ground point cloud are as follows:
s1: determining a threshold value of a gradient filtering model;
s2: selecting a ground point and an adjacent point and calculating the distance d between the ground point and the adjacent point;
s3: judging the relation between the distance d and a threshold value, if the distance d is less than or equal to the threshold value, the ground point is a ground seed point, otherwise, the ground seed point is not the ground seed point;
s4: initializing a cloth grid of the cloth filter model, setting cloth grid resolution C, and virtually meshing the ground seed points obtained in the step S3 to obtain particles;
s5: projecting all the ground seed points and mass points to the same horizontal plane, searching each mass point and the nearest ground seed point and recording the elevation value of each mass point and the nearest ground seed point;
s6: calculating the displacement of each mass point under the action of gravity, and comparing the elevation values of the mass point and the nearest ground seed point; if the elevation value of the mass point is less than or equal to the elevation value of the nearest ground seed point, placing the mass point at the position of the ground seed point, and setting the mass point as an immovable point;
s7: calculating the displacement of each mass point under the action of the internal force;
s8: repeating the step S6 and the step S7 until the maximum elevation change value of all the particles is smaller than a preset value or the maximum iteration number I set by a user is reached, and finishing filtering; and if the distance from the seed ground point to the corresponding mass point is less than a threshold value, classifying the point as a ground point to obtain the filtered ground point cloud.
As a further optimization scheme of the invention, the specific steps of constructing the elevation abnormal value surface model are as follows:
continuously measuring the coordinates of the characteristic points for multiple times by selecting certain density characteristic points which are uniformly distributed in the region to be measured and by a GNSS RTK technology, and obtaining the on-site three-dimensional coordinates of the characteristic points by taking the average value;
processing the aerial photography image through Smart3D Capture unmanned aerial vehicle remote sensing image processing software to generate a three-dimensional model of the area to be measured, and acquiring three-dimensional coordinates of the feature points under the three-dimensional model;
obtaining an abnormal value of the characteristic point in the elevation direction through the solid three-dimensional coordinate and the three-dimensional coordinate of the characteristic point under the three-dimensional model;
and constructing a quadratic polynomial fitting plane so as to obtain an elevation abnormal value curved surface model.
As a further optimization scheme of the invention, the specific process of utilizing the elevation abnormal value curved surface model to compensate and correct the ground seed DEM is as follows:
and superposing the ground seed DEM and the elevation abnormal value curved surface model by using a grid calculator tool bar in Arcmap software to obtain a final DEM.
The invention has the beneficial effects that:
according to the method, the gradient-distribution simulation filtering model is constructed to obtain the ground point cloud and construct the ground seed DEM, the elevation abnormal value curved surface model is constructed on the basis, and finally the final DEM is obtained through the DEM difference method, so that the precision of constructing a digital elevation model product by using the consumer-grade UAV collected data is improved.
Drawings
FIG. 1 is a flow chart of a method of the present invention;
FIG. 2 is a schematic diagram of a CSF of the present invention;
FIG. 3 is a GF-CSF acquisition ground point flow diagram of the present invention;
FIG. 4 is a comparison of the filtering effects of the present invention;
FIG. 5 is a graph of the effect of polynomial fitting of the present invention;
FIG. 6 is an image of an experimental area of the present invention, wherein the rectangle is the area of interest;
FIG. 7 is a data processing result diagram of the present invention;
FIG. 8 is a graph of the effect of the fitting of the experimental region according to the present invention;
FIG. 9 is a DEM rendering of the present invention, with a seed DEM on the left and a final DEM on the right;
FIG. 10 is a plan view of a sampling point of the present invention;
fig. 11 is a sampling check chart of the accuracy improvement effect of the present invention.
Detailed Description
The present application will now be described in further detail with reference to the drawings, it being noted that the following detailed description is given by way of illustration only, and should not be construed as limiting the scope of the application, since certain insubstantial modifications and adaptations of the invention will become apparent to those skilled in the art based upon the foregoing description.
As shown in fig. 1-11, a method for improving accuracy of DEM products of consumer-grade unmanned aerial vehicles comprises the following steps:
acquiring an image map, a three-dimensional point cloud and a three-dimensional model of a region to be detected by a consumption-level unmanned aerial vehicle;
constructing a gradient-distribution filtering model to filter the three-dimensional point cloud to obtain a ground point cloud, and constructing a ground seed DEM through the ground point cloud;
selecting certain density characteristic points which are uniformly distributed in the area to be measured, measuring the plane coordinates and the elevation abnormal values of the characteristic points through a GNSS RTK technology, fitting the elevation abnormal values of the characteristic points by utilizing a quadratic polynomial surface model, and constructing a surface model of the elevation abnormal values of the area to be measured;
and compensating and correcting the ground seed DEM by using the elevation abnormal value curved surface model, thereby realizing the improvement of the product precision of the consumption-level unmanned aerial vehicle DEM.
The specific steps of obtaining the image map, the three-dimensional point cloud and the three-dimensional model of the area to be measured by the consumer-grade unmanned aerial vehicle are as follows:
laying image control points in the area to be detected;
acquiring an aerial photograph of the area to be measured by a consumption-level unmanned aerial vehicle;
and matching and processing the aerial image and the image control points by using Smart3D Capture unmanned aerial vehicle remote sensing image processing software to obtain an image map, a three-dimensional point cloud and a three-dimensional model of the area to be measured.
The specific steps of filtering the three-dimensional point cloud by the gradient-cloth filtering model to obtain the ground point cloud are as follows:
s1: determining a threshold value of a gradient filtering model;
s2: selecting a ground point and an adjacent point and calculating the distance d between the ground point and the adjacent point;
s3: judging the relation between the distance d and a threshold value, if the distance d is less than or equal to the threshold value, the ground point is a ground seed point, otherwise, the ground seed point is not the ground seed point;
s4: initializing a cloth grid of the cloth filter model, setting cloth grid resolution C, and virtually meshing the ground seed points obtained in the step S3 to obtain particles;
s5: projecting all the ground seed points and mass points to the same horizontal plane, searching each mass point and the nearest ground seed point and recording the elevation value of each mass point and the nearest ground seed point;
s6: calculating the displacement of each mass point under the action of gravity, and comparing the elevation values of the mass point and the nearest ground seed point; if the elevation value of the mass point is less than or equal to the elevation value of the nearest ground seed point, placing the mass point at the position of the ground seed point, and setting the mass point as an immovable point;
s7: calculating the displacement of each mass point under the action of the internal force;
s8: repeating the step S6 and the step S7 until the maximum elevation change value of all the particles is smaller than a preset value or the maximum iteration number I set by a user is reached, and ending the filtering; and if the distance from the seed ground point to the corresponding mass point is less than a threshold value, classifying the point as a ground point to obtain filtered ground point cloud.
The specific steps of constructing the elevation abnormal value curved surface model are as follows:
selecting characteristic points at equal intervals in the area to be measured through GNSS RTK, continuously measuring the coordinates of the characteristic points for multiple times, and averaging to obtain a field three-dimensional coordinate of the characteristic points;
processing the aerial photography image through Smart3D Capture unmanned aerial vehicle remote sensing image processing software to generate a three-dimensional model of the area to be measured, and acquiring three-dimensional coordinates of the feature points under the three-dimensional model;
obtaining an abnormal value of the characteristic point in the elevation direction through the solid three-dimensional coordinate and the three-dimensional coordinate of the characteristic point under the three-dimensional model;
and constructing a quadratic polynomial fitting plane so as to obtain an elevation abnormal value curved surface model.
The specific process of utilizing the elevation abnormal value curved surface model to compensate and correct the ground seed DEM is as follows:
and superposing the ground seed DEM and the elevation abnormal value curved surface model by using a grid calculator tool bar in Arcmap software to obtain a final DEM.
Gradient filtering algorithm: the gradient filter algorithm (GF) determines whether to reject or accept the selected point by mainly comparing the magnitude of the height difference between the two points; the threshold value of the height difference between two points is defined as a function of the distance between the two points, i.e. delta hmax(d) The filtering kernel function of (1); when filtering is performed, the probability that a point cloud data point with a large elevation value belongs to a ground point is smaller as the distance between two points is reduced. Assuming that A is an original data set, DEM is a ground point set and d is an inter-point distance, the following requirements are metThe points of the filter function are the ground point clouds.
If for a given point piNo adjacent point ρ can be foundjSo that they satisfy the relation (2), then piDivided into ground points.
Cloth simulation filtering algorithm: the cloth simulation filtering algorithm (CSF) is a filtering algorithm based on surface adjustment, and is to invert a point cloud obtained by scanning and then cover the inverted point cloud with a rigid cloth. By analyzing the interaction between the distribution nodes and the points in the corresponding point clouds, the positions of the distribution nodes can be determined to generate an approximate surface shape, and finally, the ground points are extracted from the point cloud entity by comparing the distances between the points in the original point clouds and the generated distribution curved surface. The inversion process is shown in figure 2.
Elevation outlier fitting: the basic idea of the invention for improving the accuracy is to determine a curved surface through a group of known elevation abnormal point positions in a certain area range, and then determine the elevation abnormal values of other measuring points in the area range through the curved surface and the point positions. The determined surface best fits the data so that it reflects the trend of these discrete data changes to minimize the sum of squared errors of the data points. The difference between the image control point and the RTK measurement corresponding point in the Smart3D Capture generation model and the plane coordinates (x, y) thereof form a functional relation:
ΔZ＝f(x,y) (3)
where f (x, y) is a polynomial function, such as a quadratic polynomial, the expression is as follows:
ΔZ＝a0+a1x+a2y+a3xy+a4x2+a5y2 (4)
in the formula a0....a5Are the model parameters to be found.
Therefore, the least square matrix form corresponding to the abnormal elevation fitting plane based on the polynomial at the moment is as follows:
in the formula (x)1,y1),(x2,y2),....,(xn,yn) As plane coordinates of elevation anomaly points, Δ Z1,ΔZ2,...,ΔZnThe elevation abnormal value of the corresponding point is obtained. And (4) solving by applying a normal equation set to obtain model parameters so as to fit the elevation abnormal values of other points in the measurement area.
Constructing a gradient-cloth filtering model: the GF is used for extracting ground points of an experimental area, has good filtering effect on low and short vegetation, but has poor filtering effect on tall trees or the top of a building. The CSF is used for extracting ground points in an experimental area, has poor filtering effect on short vegetation covered by weeds or vegetation beside roads, but has good filtering effect on buildings and medium and high vegetation. In order to obtain accurate DEM of an experimental area, a GF-CSF filtering model is constructed by combining GF and CSF, GF obtains ground seed points, and then CSF is used for filtering the ground seed points to ensure that non-ground point clouds can be deleted as far as possible. The filtering method comprises the following specific steps:
s1: determining a GF threshold;
s2: calculating the distance d between the seed point and the adjacent point;
s3: judging the relation between the distance d and a threshold value to obtain a seed ground point;
s4: initializing a cloth grid: setting the distribution grid resolution C, and carrying out virtual grid on the seed ground points;
s5: projecting all the seed ground point data and the gridded particle data to the same horizontal plane, searching each particle of the cloth and the most adjacent point in the point cloud, and recording the elevation value of the point cloud;
s6: calculating the displacement of each movable mass point in the virtual gridding under the action of gravity, and comparing the elevation value of the point with the elevation value of the corresponding point cloud; if the elevation value of the mass point is equal to or less than the elevation value of the corresponding point cloud, placing the mass point on the position of the point cloud, and setting the mass point as an immovable point;
s7: calculating the displacement of each virtual grid particle under the action of internal force;
s8: repeating S6 and S7 until the maximum elevation change value in all virtual particles is small enough or the maximum iteration number I set by a user is reached, and ending the filtering;
s9: distinguishing between ground and non-ground points. If the distance from a point in the seed ground points to the virtual quality point corresponding to the seed ground points is less than the filtering threshold value, classifying the point as a ground point, otherwise, classifying the point as a non-ground point.
Wherein, the GF-CSF parameter system is [ R, C, I, T ], R is a cloth hardness coefficient, C is a cloth grid resolution, I is a maximum iteration number, and T is a GF-CSF classification threshold.
A flowchart of a procedure for obtaining a ground point cloud based on the GF-CSF method is shown in fig. 3.
Precision evaluation based on GF-CSF filtering model: the setting of each parameter in the GF-CSF parameter system has a direct influence on the filtering effect, and the [ R, C, I, T ] is set to [2,0.6,500,0.4] according to the experience in combination with the topographic characteristics of the experimental area.
Visual qualitative evaluation: whether the ground point data after filtering meets the drawing requirement is observed according to visual qualitative evaluation criteria; as can be seen from fig. 4: generally, buildings and forest trees can be well divided into non-ground points, but short weeds in part of roads cannot be effectively removed, the overall filtering effect of the edge of a test area needs to be improved, the top points of houses close to steep sills are wrongly divided into the ground points, and higher houses are easily mistakenly removed, because the misjudgment probability of dense point cloud data in the matching process is high. Therefore, in the later processing, the dense point cloud data is subjected to rarefying processing and then point cloud matching, so that misjudgment of point cloud matching of partial positions can be effectively reduced.
Evaluation of statistical metrology quantification: the international association for photogrammetry and remote sensing (isps) in 2003 proposed an error assessment standard for filter algorithm results (Sithole G, 2004), in which the filter result errors are grouped into three categories: the missing division error (I) is the percentage of the number of the wrongly divided non-ground points in the ground points to the total number of the ground points, the wrong division error (II) is the percentage of the number of the wrongly divided non-ground points in the non-ground points to the total number of the non-ground points, and the total error (III) is the percentage of the total number of the wrongly divided ground points and the non-ground points to the total point cloud number.
In the formula, a is the number of the ground points wrongly divided into the non-ground points, c is the number of the correctly divided ground points, b is the number of the non-ground points wrongly divided into the ground points, and d is the number of the correctly divided non-ground points. On the point cloud filtering effect, the filtering effect of each filtering algorithm is more visually reflected on the basis of a statistical quantification method. The filtering effect is shown in table 1.
Table 1: error analysis of filtering result of each filtering algorithm
Ground points where missing score errors occur in some of the highly populated ground are misclassified as non-ground points and other ground points occur at the perimeter of the building. Misclassification errors occur mainly in point clouds of low vegetation and are misclassified to ground points. As can be seen from table 1, the missing separation error of the gradient filter in the experiment is 3%, the missing separation error is 33%, and the missing separation error of the cloth filter in the experiment is 16%, and the missing separation error is 10%. The gradient filtering has good filtering effect on surrounding low vegetation when filtering non-ground point cloud, but has obviously poor filtering effect on the point cloud on the upper part of a higher building; cloth filtering is relatively poor to low vegetation filter effect, but has good filter effect to higher building point cloud. Therefore, the filtration of the combination of the two is carried out with the missing division error of 11 percent, the misclassification error of 3 percent and the total error of 7 percent. The point cloud filtering effect of the whole experimental area is good.
The elevation fitting method based on the polynomial surface model comprises the following steps:
the method comprises the steps of selecting feature points at equal intervals in a region to be measured through GNSS RTK, measuring coordinates of the feature points continuously and repeatedly, obtaining a solid three-dimensional coordinate of the feature points by taking an average value, generating an experimental region model through Smart3D Capture high-precision image dense matching technology, obtaining a three-dimensional coordinate of the feature points under the model, obtaining abnormal values of the feature points in the elevation direction, constructing various times of polynomial fitting planes, and constructing an evaluation function equation (9) through various polynomial fitting effects such as Root Mean Square Error (RMSE) and plane overall fitting degree effect (R2).
Then, an optimal polynomial fitting elevation abnormal value curved surface is selected according to the lowest point of the evaluation function curve, and the method specifically comprises the following steps:
acquiring coordinates and elevation abnormal values of points required by a fitting plane;
constructing a polynomial function;
obtaining model parameters according to the feature points;
calculating the RMSE and R under various degree polynomial fitting2A value;
calculating an evaluation function value F, and selecting an optimal degree polynomial according to the F;
and determining a fitting plane to obtain an elevation abnormal value curved surface model.
When the polynomial fitting plane is carried out, the polynomial times are respectively selected to be primary, secondary, tertiary and quaternary to carry out comparative analysis, and the effect of fitting characteristic points of each polynomial is compared. The results are shown in Table 2 and FIG. 5.
Table 2: polynomial fitting effect
As can be seen from Table 2, in terms of root mean square error when each degree polynomial is fitted to a plane, the first, third and fourth RMSE values all exceed 0.02, and the second RMSE value is 0.0145; in the fitting degree of the regression equation as a whole, R is given once, three times and four times2Are all less than 0.5, two times R2The value is 0.7456. As can be seen from fig. 5, the trend of the F curve shows a trend of "reverse-positive distribution" in the curve as a whole, and an inflection point appears at the second order. Therefore, the quadratic polynomial fitting of the elevation abnormal curved surface has high accuracy. Example applications
Study area overview: the experimental area is located in the southern school district of the university of Anhui Ringman of Huai nan City, Anhui province, the length of the experimental area is about 0.324km, the width of the experimental area is about 0.153km, and the area of the measuring area is about 0.050km2. The experimental area has many buildings, smooth inner roads, fluctuant terrain, and the experimental area has distributed vegetation, grassland, water areas and the like. The location of the study area is shown in figure 6. The invention adopts a Xinntom 4Pro four-rotor unmanned aerial vehicle in Xinjiang to obtain image data, and the basic parameters are shown in table 3. The flight height of the unmanned aerial vehicle in the experimental area is 100m, and the flight course and the lateral overlapping degree of the unmanned aerial vehicle are set to be 80%.
Table 3: relevant parameters of unmanned aerial vehicle
Before the unmanned aerial vehicle takes photo by plane, the unmanned aerial vehicle is used for preliminarily acquiring an orthophoto map of an experimental area, determining a terrain condition and an aerial photographing flight line according to the orthophoto map, and preliminarily selecting a control point position. And laying image control points on the spot according to the preselected position and acquiring coordinates of the image control points under the CGCS 2000 coordinate system by utilizing GNSS-RTK. The image control points are distributed by adopting a regional network, the coordinates are obtained by adopting a GNSS RTK method, the GNSS-RTK is used for continuously observing for four times, and the average value is taken as the final coordinates. A total of 58 image control points are acquired in the investigation region and used as control points and examination points.
Data processing: in order to reduce the influence of RTK measuring point errors on the accuracy of a final model as much as possible, aerial photos acquired in an experimental area are subjected to primary processing by using Smart3D Capture professional unmanned aerial vehicle remote sensing image processing software. Measuring coordinates of 58 image control points in the acquired Digital Orthophoto Map (DOM), Digital Surface Model (DSM) and three-dimensional model of the research area, and processing the gross error by using Lauder rule equation (10):
in the formula: x is the number ofiIn order to generate the height difference between the image control point in the model and the RTK measurement corresponding point,
the height difference is the mean value, and n is the number of image control points.
And calculating the absolute value of the error of each sampling point, comparing the absolute value with 3 times of standard deviation, rejecting the sampling points if the absolute value is more than 3 times of standard deviation, and repeating the steps until no rejection exists. And processing the rest image control points by adopting Smart3D Capture high-precision image dense matching technology, and automatically matching the same name points in all the acquired image data. And acquiring accurate terrain and ground object information to generate dense point cloud. The data processing results are shown in fig. 7.
High precision DEM results: and constructing an elevation abnormal value fitting plane taking a quadratic polynomial as a model by combining the ground point coordinates and the abnormal elevation values acquired in the data acquisition and processing stage. The model parameters are calculated as: a is a0＝1.716e+07，a1＝-2.406，a2＝-9.187，a3＝-6.97e-07，a4＝4.908e-06， a51.323 e-06. The results of the model construction are shown in FIG. 8.
Then, constructing a seed DEM (digital elevation model) by using ground point cloud obtained by GF-CSF (glass fiber-CSF); constructing an elevation abnormal value curved surface model by utilizing a curved surface fitted by a quadratic polynomial; and then, compensating and correcting the ground seed DEM by using the elevation abnormal value curved surface model to obtain a final DEM, as shown in FIG. 8. As can be seen from fig. 9, the whole elevation trends of the experimental area before and after the precision is improved are consistent, in order to continuously verify the effectiveness and the accuracy of the improvement of the DEM precision, 20 check points are selected in the experimental area according to the principle of a random sampling method, sampling points are distributed as shown in fig. 10, and coordinates of the check points are acquired by using a data acquisition stage method. Then, the elevations of all the inspection points measured by DEM before and after the precision improvement are compared, and an elevation abnormal value delta Z is calculated, wherein the specific result is shown in a table 4:
table 4: precision-improving effect sampling inspection table
As can be seen from table 4 and fig. 11, in 20 randomly sampled check points, the DEM after the accuracy is improved is closer to the measured elevation, and the elevation abnormal value is significantly reduced compared to that before the improvement. In order to further understand the precision improvement effect, the RMSE value of the elevation abnormal value is calculated by using the data, the RMSE before the improvement is 33.7mm, the RMSE after the improvement is 16.7mm, and the DEM precision is improved. The method effectively solves the problem of insufficient accuracy of the low-accuracy unmanned aerial vehicle aerial survey DEM.
And (4) conclusion:
(1) the missing separation error is 11 percent, the wrong separation error is 3 percent and the total error is 7 percent when filtering by a GF-CSF method. The point cloud filtering effect of the whole experimental area is good. Therefore, the GF-CSF method has good filtering effect on areas with distributed buildings, complicated internal roads, fluctuant terrain, low and medium-high vegetation, grasslands, water areas and the like.
(2) The fitting result of the abnormal value of the elevation is not improved along with the improvement of the order of the polynomial, and the fitting precision of the second-order polynomial is higher than that of the first-order polynomial and the third-order polynomial. According to the method, a quadratic polynomial is used for fitting the DEM, the RMSE is 33.7mm before fitting, the RMSE is 16.7mm after fitting, and the DEM precision is improved.
(3) The method for combining the abnormal elevation value fitting plane after the high-quality ground point acquired by the GF-CSF is used for constructing the DEM has an obvious effect on improving the DEM precision of the low-precision unmanned aerial vehicle, but the improved DEM precision only approaches to the RTK precision by taking the RTK real measurement data as a real value.
The above-mentioned embodiments only express several embodiments of the present invention, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention.
Claims (5)
1. A method for improving the precision of a DEM product of a consumption-level unmanned aerial vehicle is characterized by comprising the following steps:
acquiring an image map, a three-dimensional point cloud and a three-dimensional model of a region to be detected by a consumer-grade unmanned aerial vehicle;
constructing a gradient-distribution filtering model to filter the three-dimensional point cloud to obtain a ground point cloud, and constructing a ground seed DEM through the ground point cloud;
selecting certain density characteristic points which are uniformly distributed in the area to be measured, measuring the plane coordinates and the elevation abnormal values of the characteristic points through a GNSS RTK technology, fitting the elevation abnormal values of the characteristic points by utilizing a quadratic polynomial surface model, and constructing a surface model of the elevation abnormal values of the area to be measured;
and compensating and correcting the ground seed DEM by using the elevation abnormal value curved surface model, thereby realizing the improvement of the product precision of the consumption-level unmanned aerial vehicle DEM.
2. The method for improving the accuracy of DEM products of consumer-grade unmanned aerial vehicles according to claim 1, wherein the method comprises the following steps: the specific steps of acquiring the image map, the three-dimensional point cloud and the three-dimensional model of the area to be detected by the consumer-grade unmanned aerial vehicle are as follows:
laying image control points in the region to be detected;
acquiring an aerial photograph of the area to be measured by a consumption-level unmanned aerial vehicle;
and matching and processing the aerial image and the image control points by using Smart3D Capture unmanned aerial vehicle remote sensing image processing software to obtain an image map, a three-dimensional point cloud and a three-dimensional model of the area to be measured.
3. The method for improving the accuracy of DEM products of consumer-grade unmanned aerial vehicles according to claim 1, wherein the method comprises the following steps: the concrete steps of filtering the three-dimensional point cloud by the gradient-cloth filtering model to obtain the ground point cloud are as follows:
s1: determining a threshold value of a gradient filtering model;
s2: selecting a ground point and an adjacent point and calculating the distance d between the ground point and the adjacent point;
s3: judging the relation between the distance d and a threshold value, if the distance d is less than or equal to the threshold value, the ground point is a ground seed point, otherwise, the ground point is not the ground seed point;
s4: initializing a cloth grid of the cloth filter model, setting cloth grid resolution C, and virtually meshing the ground seed points obtained in the step S3 to obtain particles;
s5: projecting all the ground seed points and mass points to the same horizontal plane, searching each mass point and the nearest ground seed point and recording the elevation value of each mass point;
s6: calculating the displacement of each mass point under the action of gravity, and comparing the elevation values of the mass point and the nearest ground seed point; if the elevation value of the mass point is less than or equal to the elevation value of the nearest ground seed point, placing the mass point at the position of the ground seed point, and setting the mass point as an immovable point;
s7: calculating the displacement of each mass point under the action of the internal force;
s8: repeating the step S6 and the step S7 until the maximum elevation change value of all the particles is smaller than a preset value or the maximum iteration number I set by a user is reached, and finishing filtering; and if the distance from the seed ground point to the corresponding mass point is less than a threshold value, classifying the point as a ground point to obtain filtered ground point cloud.
4. The method for improving the accuracy of DEM products of consumer-grade unmanned aerial vehicles according to claim 2, wherein the method comprises the following steps: the specific steps of constructing the elevation abnormal value curved surface model are as follows:
selecting characteristic points of the area to be measured at a certain density through GNSS RTK, continuously measuring coordinates of the characteristic points for multiple times, and averaging to obtain a field three-dimensional coordinate of the characteristic points;
processing the aerial photography image through Smart3D Capture unmanned aerial vehicle remote sensing image processing software to generate a three-dimensional model of the area to be measured, and acquiring three-dimensional coordinates of the feature points under the three-dimensional model;
obtaining abnormal values of the feature points in the elevation direction through the solid three-dimensional coordinates and the three-dimensional coordinates of the feature points under the three-dimensional model;
and constructing a quadratic polynomial fitting plane so as to obtain an elevation abnormal value curved surface model.
5. The method for improving the accuracy of DEM products of consumer-grade unmanned aerial vehicles according to claim 1, wherein the method comprises the following steps: the specific process of utilizing the elevation abnormal value curved surface model to compensate and correct the ground seed DEM is as follows:
and superposing the ground seed DEM and the elevation abnormal value curved surface model by using a grid calculator tool bar in Arcmap software to obtain a final DEM.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210117130.8A CN114612806A (en) | 2022-02-07 | 2022-02-07 | Method for improving DEM product precision of consumption-level unmanned aerial vehicle |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210117130.8A CN114612806A (en) | 2022-02-07 | 2022-02-07 | Method for improving DEM product precision of consumption-level unmanned aerial vehicle |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114612806A true CN114612806A (en) | 2022-06-10 |
Family
ID=81858489
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210117130.8A Pending CN114612806A (en) | 2022-02-07 | 2022-02-07 | Method for improving DEM product precision of consumption-level unmanned aerial vehicle |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114612806A (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106529469A (en) * | 2016-11-08 | 2017-03-22 | 华北水利水电大学 | Unmanned aerial vehicle airborne LiDAR point cloud filtering method based on adaptive gradient |
| CN106969751A (en) * | 2017-03-13 | 2017-07-21 | 西安科技大学 | A kind of method of the coal mining subsidence amount monitoring calculation based on unmanned aerial vehicle remote sensing |
| CN110569786A (en) * | 2019-09-06 | 2019-12-13 | 中国农业科学院农业资源与农业区划研究所 | fruit tree identification and quantity monitoring method and system based on unmanned aerial vehicle data acquisition |
| CN111598780A (en) * | 2020-05-14 | 2020-08-28 | 山东科技大学 | Terrain adaptive interpolation filtering method suitable for airborne LiDAR point cloud |
| CN113327276A (en) * | 2021-06-30 | 2021-08-31 | 江苏海洋大学 | Method for registering point cloud data of large volume facing mobile measurement |
| KR102336523B1 (en) * | 2021-04-27 | 2021-12-08 | 주식회사 모빌테크 | Mobile mapping system that improves map information using point cloud data |
| CN113971768A (en) * | 2021-10-27 | 2022-01-25 | 国网安徽省电力有限公司马鞍山供电公司 | Unmanned aerial vehicle-based three-dimensional dynamic detection method for power transmission line illegal building |
| CN114067073A (en) * | 2021-11-16 | 2022-02-18 | 安徽理工大学 | TLS point cloud-based mining area building deformation automatic extraction method |
-
2022
- 2022-02-07 CN CN202210117130.8A patent/CN114612806A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106529469A (en) * | 2016-11-08 | 2017-03-22 | 华北水利水电大学 | Unmanned aerial vehicle airborne LiDAR point cloud filtering method based on adaptive gradient |
| CN106969751A (en) * | 2017-03-13 | 2017-07-21 | 西安科技大学 | A kind of method of the coal mining subsidence amount monitoring calculation based on unmanned aerial vehicle remote sensing |
| CN110569786A (en) * | 2019-09-06 | 2019-12-13 | 中国农业科学院农业资源与农业区划研究所 | fruit tree identification and quantity monitoring method and system based on unmanned aerial vehicle data acquisition |
| CN111598780A (en) * | 2020-05-14 | 2020-08-28 | 山东科技大学 | Terrain adaptive interpolation filtering method suitable for airborne LiDAR point cloud |
| KR102336523B1 (en) * | 2021-04-27 | 2021-12-08 | 주식회사 모빌테크 | Mobile mapping system that improves map information using point cloud data |
| CN113327276A (en) * | 2021-06-30 | 2021-08-31 | 江苏海洋大学 | Method for registering point cloud data of large volume facing mobile measurement |
| CN113971768A (en) * | 2021-10-27 | 2022-01-25 | 国网安徽省电力有限公司马鞍山供电公司 | Unmanned aerial vehicle-based three-dimensional dynamic detection method for power transmission line illegal building |
| CN114067073A (en) * | 2021-11-16 | 2022-02-18 | 安徽理工大学 | TLS point cloud-based mining area building deformation automatic extraction method |
Non-Patent Citations (7)
| Title |
|---|
| ZHANG, W, ET AL.: "An Easy-to-Use Airborne LiDAR Data Filtering Method Based on Cloth Simulation", REMOTE SENSING, 15 June 2016 (2016-06-15), pages 1 - 22 * |
| 刘毅, 王磊, 王胜利等.: "改进的布料模拟算法在多波束点云粗差剔除中的应用", 科学技术与工程, 8 November 2021 (2021-11-08), pages 13248 - 13253 * |
| 周拥军 等: "现代土木工程测量", 30 November 2021, 《上海：上海交通大学出版社》, pages: 175 - 179 * |
| 晓晨的博客: "点云地面点滤波(Cloth Simulation Filter, CSF)"布料"滤波算法介绍", CSDN, HTTPS://BLOG.CSDN.NET/WXC_1998/ARTICLE/DETAILS/107206075, 8 July 2020 (2020-07-08), pages 1 - 6 * |
| 林春峰: "GNSS RTK高程拟合控制点选取工具设计与实现", 铁道勘察, 23 December 2021 (2021-12-23), pages 21 - 25 * |
| 段延松 等: "无人机测绘生产", 31 July 2019, 《武汉：武汉大学出版社》, pages: 306 * |
| 贾秉松, 李亮, 王瑞等.: "基于点云数据的山区开采沉陷特征研究", 测绘科学, 30 October 2020 (2020-10-30), pages 88 - 94 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kraus et al. | Advanced DTM generation from LIDAR data | |
| CN108846352B (en) | Vegetation classification and identification method | |
| CN102506824B (en) | Method for generating digital orthophoto map (DOM) by urban low altitude unmanned aerial vehicle | |
| CN109492563A (en) | A kind of tree species classification method based on unmanned plane Hyperspectral imaging and LiDAR point cloud | |
| Fabris et al. | Automated DEM extraction in digital aerial photogrammetry: precisions and validation for mass movement monitoring | |
| KR101547940B1 (en) | An error correction system for data of terrestrial LiDAR on the same plane and the method thereof | |
| CN113034689A (en) | Laser point cloud-based terrain three-dimensional model, terrain map construction method and system, and storage medium | |
| Briese | Three-dimensional modelling of breaklines from airborne laser scanner data | |
| CN110889899B (en) | Digital earth surface model generation method and device | |
| CN109100719A (en) | Combine plotting method with the topographic map of optical image based on satellite-borne SAR image | |
| CN115761303A (en) | Ground object classification method based on airborne laser radar point cloud and remote sensing image data | |
| Baily et al. | Comparative assessment of analytical and digital photogrammetric methods in the construction of DEMs of geomorphological forms | |
| CN111986074A (en) | Real projective image manufacturing method, device, equipment and storage medium | |
| Süleymanoğlu et al. | Comparison of filtering algorithms used for DTM production from airborne lidar data: A case study in Bergama, Turkey | |
| CN108956392B (en) | Unmanned aerial vehicle identification method for tidal flat sediment types | |
| CN111060922B (en) | Tree point cloud extraction method based on airborne laser radar point cloud spatial distribution characteristics | |
| Ismail et al. | DEM derived from photogrammetric generated DSM using morphological filter | |
| Cățeanu et al. | ALS for terrain mapping in forest environments: An analysis of LiDAR filtering algorithms | |
| Li et al. | Registration of Aerial Imagery and Lidar Data in Desert Areas Using the Centroids of Bushes as Control Information. | |
| Partama et al. | A simple and empirical refraction correction method for UAV-based shallow-water photogrammetry | |
| CN114612806A (en) | Method for improving DEM product precision of consumption-level unmanned aerial vehicle | |
| Rahmayudi et al. | Comparison of semi automatic DTM from image matching with DTM from Lidar | |
| Sefercik et al. | Point-based and model-based geolocation analysis of airborne laser scanning data | |
| García et al. | Delineation of individual tree crowns for LiDAR tree and stand parameter estimation in Scottish woodlands | |
| Ye et al. | Gaussian mixture model of ground filtering based on hierarchical curvature constraints for airborne lidar point clouds |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: 232000 Tian Jia An District, Huainan City, Anhui Applicant after: Anhui University of Science and Technology Address before: 232000 tianjia'an District, Hefei City, Anhui Province Applicant before: Anhui University of Science and Technology |