CN-122015778-A - Automatic screening method of oblique photogrammetry images facing target area

CN122015778ACN 122015778 ACN122015778 ACN 122015778ACN-122015778-A

Abstract

The invention discloses an automatic screening method of an oblique photogrammetry image facing a target area, and belongs to the technical field of oblique photogrammetry and three-dimensional modeling. The technical problem to be solved is to solve the problem that the three solutions of weak texture areas such as water areas, bare lands and the like in the area are easy to fail. The technical scheme is characterized in that a photogrammetry principle and a GIS space analysis technology are fused, a localization elevation datum plane is established through pre-flight calibration to calculate relative altitude, an image geographic range area is generated through coordinate conversion, rotation matrix construction and the like, and then an optimal image subset associated with a target area is automatically screened out through triple space topological relation discrimination, so that invalid image interference is avoided from the source.

Inventors

XU ZENAN
LU YUAN
Li Chanfu
ZHANG SHENG
Ling Jiangyu
Xi Wenhuan
ZHU DONGHAI
Zhou Shuangbaihe
DUAN QIAN

Assignees

广东水电二局集团有限公司
广东省建筑工程集团股份有限公司

Dates

Publication Date: 20260512
Application Date: 20260303

Claims (10)

1. An automatic screening method of oblique photogrammetry images facing to a target area is characterized by comprising the following steps: S1, calculating relative altitude, establishing a localization elevation reference plane, and calculating the relative altitude of each aerial image relative to the localization elevation reference plane; S2, constructing an image geographic range, namely sequentially carrying out coordinate conversion, attitude parameter processing, rotation matrix construction and angular point geodetic coordinate calculation on the basis of a photogrammetry principle and a coordinate conversion technology to obtain a ground coverage geographic range area of each aerial image; S3, intelligent judgment of spatial relationship, preprocessing the vector red line of the target area, and screening out images which are spatially related with the target area and output the images through judgment of multiple spatial topological relationships.
2. The method for automatically screening an oblique photography measurement image of a target area according to claim 1, wherein step S1 specifically comprises: pre-flight calibration, pre-flight shooting is carried out in a region, calibration images are obtained, absolute navigational height values in POS data of each calibration image are extracted and calculated, and an arithmetic average value of the absolute navigational height values is calculated and used as a localization elevation reference plane; Calculating relative altitude, namely acquiring absolute altitude values in POS data of each aerial image of the formal aerial photo, and calculating the relative altitude H of the aerial photo, wherein the specific formula is as follows: Wherein, the For an absolute altitude value of the ship, Is a localization elevation reference plane.
3. The method for automatically screening an oblique photography measurement image facing to a target area according to claim 1, wherein the coordinate transformation specifically comprises: And converting the WGS84 longitude and latitude coordinates recorded by the POS data into plane coordinates under a national geodetic 2000 coordinate system according to WGS84 and CGCS2000 ellipsoid parameters.
4. The method for automatically screening an oblique photography measurement image of a target area according to claim 3, wherein the processing of the attitude parameter comprises: the attitude angle Omega, phi, kappa is extracted from the POS data and the unit of the attitude angle is converted into an arc system.
5. The method for automatically screening an oblique photogrammetry image for a target area according to claim 1, wherein the rotation matrix construction specifically includes: Constructing primitive rotation matrixes around a Z axis, a Y axis and an X axis according to a principle of a colinear equation of photogrammetry and a three-dimensional space rotation transformation theory in a Yaw-Pitch-Roll sequence, and synthesizing a platform attitude rotation matrix through matrix chain multiplication; multiplying the lens installation rotation matrix by the platform attitude rotation matrix to obtain a total rotation matrix Rtotal from an image space coordinate system to a local horizontal coordinate system.
6. The method for automatically screening oblique photography measurement images facing to a target area according to claim 5, wherein the calculating of the geodetic coordinates of the corner point in step S2 specifically comprises: Under an image space coordinate system, defining coordinate vectors of four corner points for each lens of the unmanned aerial vehicle, wherein the coordinate vector format is [ x, y-f ], wherein x and y are normalized coordinates based on the size of a half sensor, and f is the focal length of the lens; rotating the coordinate vector of the angular point to a local horizontal coordinate system by utilizing the total rotation matrix Rtotal to obtain a direction vector V containing three components of east, north and sky; based on the collineation equation idea, intersecting calculation is carried out on the camera projection center, the direction vector V and the ground plane, and the earth coordinates of each angular point are obtained.
7. The automatic screening method of oblique photography measurement images facing to the target area according to claim 6, wherein in the step S2, all the lenses of the unmanned aerial vehicle and four corner points of each lens are traversed, and the coordinates of the corner points and the ground are sequentially connected to form a geographical range area corresponding to each aerial photography image.
8. The automatic target area-oriented oblique photography measurement image screening method according to claim 1, wherein the target area vector red line preprocessing in step S3 specifically comprises: and converting the vector red line of the target area into a point coordinate set, and converting the point coordinate from the WGS84 longitude and latitude coordinate to a plane coordinate under the national earth 2000 coordinate system.
9. The method for automatically screening an oblique photography measurement image for a target area according to claim 8, wherein in step S3, the multiple spatial topological relation discrimination is sequentially performed in the following order: s31, judging the point position, detecting whether at least one of four corner points of the geographic range area of the aerial image falls into the vector red line polygon by adopting a ray method, and judging that space correlation exists if the at least one corner point falls into the vector red line polygon; S32, reversely detecting, if S31 judges that no association exists, detecting whether any vertex of the vector red line polygon falls into the geographic range area of the aerial image, and if so, judging that space association exists; And S33, detecting the intersection of the polygons, if the two steps S31 and S32 are respectively judged to be irrelevant, detecting whether the intersection exists between the geographical range area of the aerial image and the boundary of the vector red line polygon, and if so, judging that the spatial correlation exists.
10. The automatic screening method of oblique photogrammetry images facing to a target area according to claim 9, wherein before the multiple spatial topological relation discrimination in step S3 is executed, the vector red line of the target area is subjected to the outward expansion of a preset distance to generate a buffer polygon, and the buffer polygon is used as a boundary for spatial topological relation discrimination; Or in the calculation of the angular point geodetic coordinates in the step S2, replacing the ground plane with a digital elevation model DEM, and calculating the intersection point of the direction vector V and the real earth surface represented by the DEM.

Description

Automatic screening method of oblique photogrammetry images facing target area Technical Field The invention belongs to the technical field of oblique photogrammetry and three-dimensional modeling, and particularly relates to an automatic screening method of an oblique photogrammetry image facing a target area. Background The oblique photogrammetry technology is used as a core means of modern mapping and three-dimensional modeling, is widely applied to the fields of urban planning, geological disaster monitoring, cultural heritage protection, engineering construction and the like, obtains multi-view high-resolution images through an unmanned plane or an aviation platform carrying a multi-angle lens, realizes accurate orientation of the images and three-dimensional point cloud generation through aerial triangulation, and further constructs a high-precision real-scene three-dimensional model. However, in practical applications, if there are weak texture areas or repeated texture areas such as water areas, bare areas, snow areas, sand areas, etc., the image matching is difficult due to lack of enough feature points, resulting in failure of the three-dimensional solution or distortion of the result. In the prior art, the full-scale empty three-iteration method needs repeated trial and error to remove invalid images, time and labor are consumed, and the manual experience pre-screening method has low efficiency, strong subjectivity and poor consistency and repeatability of results. The related patent documents: The publication of China is No. CN107917699B, publication date is No. 1 month and No. 17 in 2020, the publication discloses a method for improving quality of mountain landform oblique photogrammetry air three, which specifically comprises the steps of determining absolute altitude (comprising a plurality of same or different altitudes) of each route according to ground resolution requirements and mountain fluctuation conditions, obtaining relative altitude and absolute altitude of each route, obtaining oblique image data, GNSS data and IMU data of a mountain through aerial remote sensing flight, sorting the data, adopting a mode of adopting integral air three (carrying out air three aiming at all data) or dividing air three into blocks firstly, merging air three (carrying out air three after merging on daily data respectively, and carrying out air three on merging data) to obtain an air three result and an air three report, wherein the two air three modes comprise SURF characteristic point detection, SURF characteristic point description and RANSAC accurate matching, and the block air three is carried out under the same inclination angle condition, the integral air three is carried out under different inclination angle conditions, the three is compared under different inclination angle conditions, and the three is three-dimensional modeling result is selected to form three-dimensional real land three, and three-dimensional real land space information is formed. The prior art represented by the foregoing documents has at least the following technical problems or drawbacks that have not been solved: (1) The document relies on SURF feature point detection and matching to realize blank three, but the weak texture/repeated texture region lacks sufficient feature points, still causes difficult image matching, cannot generate effective connection points, finally causes blank three resolving interruption or result distortion, and does not solve the core technical problem; (2) The document needs to carry out integral blank three on all acquired inclined image data or combine blank three after processing all data in a blocking way, invalid/interference images covering a weak texture area are not removed from a data source, a large amount of redundant data still exist to participate in calculation, so that the blank three process is time-consuming and labor-consuming, and the aging requirement of a large-scale engineering project is difficult to meet; (3) The document does not relate to the spatial topological relation discrimination of the images and the target areas, does not screen out the image subsets only related to the target areas, still needs to process the image data of the non-target areas in the area, further increases the computational complexity, and does not solve the subjectivity problem of relying on manual experience screening in the traditional method; (4) The relative altitude calculation lacks localization suitability, namely the literature determines relative altitude through the ground resolution requirement and mountain area fluctuation conditions, a localization elevation reference plane aiming at a specific flight task is not established, the terrain features of different areas are difficult to adapt, the accuracy of the relative altitude calculation is limited, and the reliability of the follow-up three-space results can be influenced. In view of this, the present invention has been made. D