CN110838146A

CN110838146A - Homonymy point matching method, system, device and medium for coplanar cross-ratio constraint

Info

Publication number: CN110838146A
Application number: CN201910973199.9A
Authority: CN
Inventors: 王晓南; 王西旗
Original assignee: Wuhan Meso Automation Technology Co Ltd
Current assignee: Wuhan Meso Automation Technology Co Ltd
Priority date: 2019-10-14
Filing date: 2019-10-14
Publication date: 2020-02-25

Abstract

The invention relates to a co-planar cross ratio constrained homonymy point matching method, a co-planar cross ratio constrained homonymy point matching system, a co-planar cross ratio constrained homonymy point matching device and a co-planar cross ratio constrained homonymy point matching medium, wherein the co-planar cross ratio constrained homonymy point matching method comprises the steps of obtaining an object space point set, object space point set coordinates, an image space point set and; respectively calculating to obtain an object seed point intersection ratio set of each object point based on a coplanar intersection ratio calculation method, and calculating to obtain an image seed point intersection ratio set of the image point at the central position according to the coordinates of the image point set; obtaining a plurality of potential object seed point intersection ratio sets according to the image space seed point intersection ratio set and all object seed point intersection ratio sets; obtaining a potential homography matrix of each potential object seed point intersection ratio set based on a homography matrix calculation method; respectively obtaining potential homonymy point sets of the object space point sets under each potential homography matrix according to each potential homography matrix; and acquiring a target homonym point set of the object space point set from all potential homonym point sets. The method does not need to be assisted by coded light-reflecting mark points or establish a characteristic descriptor, and realizes high-precision homonymy point matching.

Description

Homonymy point matching method, system, device and medium for coplanar cross-ratio constraint

Technical Field

The invention relates to the technical field of photogrammetry and computer vision, in particular to a method, a system, a device and a medium for matching homonymous points constrained by coplanar cross ratios.

Background

In photogrammetry, the image point of a spatial point in two images is called a homonymous point image point, or simply a homonymous image point, for example, a in fig. 1₁And a₂The image point of any point A on the ground on the left and right images, the point a₁And a₂Namely the image points with the same name; the spatial point and the image point in the image are called the homonymous point, for example, a in FIG. 1₁And A is the same name point, a₂And A is also a homonymous point, wherein a homonymous image point is one of the homonymous points.

The homonymous point matching is used in three-dimensional reconstruction, point cloud splicing, image splicing and camera calibration, and is particularly important for camera calibration in machine vision and digital photogrammetry, and the higher the matching precision of the homonymous point is, the better the camera calibration effect is, and the subsequent other measurement work of the camera is facilitated.

The matching method of the same name point in the existing camera calibration is mainly divided into two types, one type is the matching of the same name point between an image point and an object point (a point in a three-dimensional space), and the matching of the same name point is generally carried out by utilizing a light-reflecting mark point with a code and identifying an independent code of the light-reflecting mark point; the other type is the homonymy point matching between image points, and the homonymy point matching is usually performed by establishing a feature descriptor between the homonymy points, such as a epipolar line geometric constraint, a similarity constraint, an angle distance constraint, a gradient direction constraint and the like; the two types of matching methods have the problems of long matching time, easy mismatching and insufficient matching precision, so that the calibration of the camera is not accurate enough, and the subsequent work of the camera is influenced.

Disclosure of Invention

The technical problem to be solved by the invention is to provide a homonymy point matching method, a homonymy point matching system, a homonymy point matching device and a homonymy point matching medium for coplanar cross-ratio constraint, which are characterized by no need of encoding reflective mark points to assist homonymy point matching or establishing feature descriptors to perform matching, and can overcome the problems that the homonymy point matching in the prior art is long in time consumption and easy to generate mismatching.

The technical scheme for solving the technical problems is as follows:

a homonymy point matching method for coplanar cross-ratio constraint comprises the following steps:

step 1: acquiring an object space point set and object space point set coordinates corresponding to the object space point set, and acquiring an image space point set of the object space point set in an image and image space point set coordinates corresponding to the image space point set;

step 2: on the basis of a coplanar intersection ratio calculation method, respectively calculating to obtain an object seed point intersection ratio set which corresponds to each object point in the object point set one by one according to the object point set coordinates, and calculating to obtain an image seed point intersection ratio set which corresponds to an image point located at the central position of the image point set according to the image point set coordinates;

and step 3: obtaining a plurality of potential object seed point intersection ratio sets according to the image side seed point intersection ratio set and all object seed point intersection ratio sets;

and 4, step 4: respectively calculating potential homography matrixes which correspond to the seed point cross ratio sets of each potential object space one by one on the basis of a homography matrix calculation method; respectively obtaining potential homonymy point sets corresponding to the object space point sets one by one under each potential homonymy matrix according to each potential homonymy matrix;

and 5: and acquiring a target homonymy point set of the object space point set from all potential homonymy point sets to complete homonymy point matching.

The invention has the beneficial effects that: firstly, an object space point set needing to be subjected to concentric point matching and corresponding object space point set coordinates are obtained, then, images of the object space point set, namely an image space point set, and corresponding image space point set coordinates are obtained, calculation and analysis are conveniently carried out according to the object space point set and the image space point set, and therefore the object space point set is conveniently subjected to homonymy point matching; because the four points imaged in the image are still collinear after the collinear four points in the space are subjected to projective transformation and meet the requirement of intersection invariance, based on the coplanar intersection ratio calculation method, an object seed point intersection ratio set corresponding to each other when each object point in an object point set is taken as an object seed point can be obtained, and in the same way, an image point at the central position of the image point set can be obtained as an image seed point, corresponding image seed point intersection ratio sets are obtained, wherein the object seed point intersection ratio set and the image seed point intersection ratio set both comprise a plurality of intersection ratios, the object seed point intersection ratio set most matched with the image seed point intersection ratio set can be obtained by analyzing the image seed point intersection ratio set and all the object seed point intersection ratio sets, namely a potential object seed point intersection ratio set, according to the potential object seed point intersection ratio set, the more accurate projective transformation relation between the image space point set and the object space point set can be conveniently researched in the follow-up process, and the potential homography matrix corresponding to each potential object space seed point intersection ratio set one by one is obtained; based on a homography matrix calculation method, different potential homography matrixes can be obtained by utilizing different potential object seed point cross ratio sets, potential homography point sets of an object point set under different projective transformation relations can be obtained according to the different potential homography matrixes, the potential homography point sets are different due to difference among the potential homography matrixes, and finally, a most reasonable and accurate target homography point set of the object point set is found from all the potential homography point sets;

the homonymy point matching method can obtain a more accurate set of potential object seed point cross ratios based on the coplane cross ratio invariance, can obtain a more accurate set of potential homonymy points by utilizing the projective transformation relation of the homonymy matrix, can obtain a target homonymy point set from the set of potential homonymy points, realizes high-precision matching of homonymy points, does not need to assist the homonymy point matching through a coded light-reflecting mark point or establish a feature descriptor for matching, avoids the problems that the homonymy point matching in the prior art is long in time consumption and easy to generate mismatching, is high in matching precision, and is extremely suitable for the fields of camera calibration, camera shooting measurement and the like.

On the basis of the technical scheme, the invention can be further improved as follows:

further: the specific steps of the step 1 comprise:

step 1.1: obtaining the object space point set by using a calibration plate, and obtaining the object space point set coordinates corresponding to the object space point set according to a calibration plate coordinate system;

step 1.2: and shooting the calibration plate to obtain a calibration plate image, and processing the calibration plate image to obtain the image square point set and the image square point set coordinates corresponding to the image square point set.

Further: in the step 2, the specific step of obtaining an object seed point intersection ratio set corresponding to each object point in the object point set one to one includes:

step 2 a.1: optionally selecting one object point in the object space point set as an object space seed point, obtaining an included angle between a connecting line between the object space seed point and each adjacent object point of the object space seed points and a horizontal direction line, and obtaining N adjacent object points of the corresponding object space seed points from all the adjacent object points of the object space seed points according to the sequence of the included angles from small to large by taking the object space seed point as the center and taking the clockwise direction as the index direction by using a kd-tree index method; wherein N is more than or equal to 5;

step 2 a.2: connecting N adjacent object points pairwise at intervals of one adjacent object point to obtain N adjacent object point connecting lines, respectively calculating to obtain one-to-one corresponding first intersection ratio on each adjacent object point connecting line according to the object space point set coordinates based on a coplanar intersection ratio calculation method, and obtaining an object space seed point intersection ratio set corresponding to the corresponding object space seed point according to all the first intersection ratios on all the adjacent object point connecting lines;

step 2 a.3: traversing each object point in the object point set, and obtaining an object seed point intersection ratio set corresponding to each object point in the object point set one by one according to the methods from the step 2a.1 to the step 2 a.2;

in step 2, the specific step of obtaining the image side seed intersection ratio set corresponding to the image point located at the central position of the image side point set includes:

step 2 b.1: taking the image point located at the central position of the image side point set as an image side seed point, and acquiring N adjacent image points of the image side seed point according to the method in the step 2 a.1;

step 2 b.2: obtaining N adjacent image point connecting lines according to the method in the step 2a.2, respectively calculating to obtain a one-to-one corresponding second intersection ratio on each adjacent image point connecting line according to the image side point set coordinates based on a coplanar intersection ratio calculation method, and obtaining the image side seed point intersection ratio set corresponding to the image side seed points according to all the second intersection ratios on all the adjacent image point connecting lines.

Further: the specific steps of the step 3 comprise:

step 3.1: selecting any one object side seed point cross ratio set from all object side seed point cross ratio sets, and performing difference operation on each second cross ratio in the image side seed point cross ratio set and each first cross ratio in one selected object side seed point cross ratio set in a one-to-one correspondence manner to obtain a cross ratio difference set corresponding to one selected object side seed point cross ratio set;

step 3.2: comparing the minimum cross ratio difference value in a cross ratio difference value set corresponding to the selected object side seed point cross ratio set with a preset cross ratio threshold value, judging whether the minimum cross ratio difference value is smaller than the cross ratio threshold value, if so, selecting one object side seed point cross ratio set as a potential object side seed point cross ratio set, and if not, discarding the selected object side seed point cross ratio set;

step 3.3: and traversing each object seed point cross ratio set in all the object seed point cross ratio sets, and obtaining a plurality of potential object seed point cross ratio sets according to the methods from the step 3.1 to the step 3.2.

Further: in step 4, the specific step of calculating the potential homography matrix corresponding to each potential object seed point intersection ratio set in a one-to-one manner includes:

step 4.1: selecting any one potential object side seed point cross ratio set from all potential object side seed point cross ratio sets, selecting at least four adjacent object points from N adjacent object points corresponding to the selected one potential object side seed point cross ratio set, and selecting at least four adjacent image points from N adjacent image points corresponding to the image side seed point cross ratio set;

step 4.2: according to the selected at least four adjacent object points and the object space point set coordinates, object point homogeneous coordinates corresponding to the at least four adjacent object points one to one are obtained, and according to the selected at least four adjacent image points and the image space point set coordinates, image point homogeneous coordinates corresponding to the at least four adjacent image points one to one are obtained; calculating to obtain a potential homography matrix corresponding to the cross ratio set of the selected potential object space seed points according to the object point homogeneous coordinates corresponding to at least four adjacent object points one to one and the image point homogeneous coordinates corresponding to at least four adjacent image points one to one;

setting the homogeneous coordinate of at least four adjacent object points corresponding to the k-th selected potential object seed point cross-ratio set as

The homogeneous coordinate of at least four adjacent image points corresponding to the image side seed point cross ratio set is (u)_i′,v_i′,1)^TThen, the specific formula of the potential homography matrix corresponding to the k-th selected potential object seed point cross ratio set is calculated as follows:

wherein H^kThe potential homography matrix corresponding to the kth potential object space seed point cross ratio set is provided, i is a positive integer and satisfies that i is more than or equal to 4;

step 4.3: and traversing each potential object seed point cross ratio set in all the potential object seed point cross ratio sets, and obtaining the potential homography matrix corresponding to each potential object seed point cross ratio set in a one-to-one manner according to the methods from the step 4.1 to the step 4.2.

Further: in step 4, the specific step of obtaining a one-to-one correspondence potential homonymy point set of the object point set under each potential homography matrix includes:

step 4.4: selecting any potential homography matrix, performing projective transformation on the image space point set by using the selected potential homography matrix, and finding out transformed object space seed points corresponding to each image point of the image space point set under the corresponding potential homography matrix in the object space point set;

step 4.5: for any one transformation object seed point, acquiring the nearest transformation object point corresponding to the corresponding transformation object seed point from the object point set by using the kd-tree indexing method;

step 4.6: calculating the distance between the nearest transformed object point and the corresponding image point to obtain the nearest distance, comparing the nearest distance with a preset distance threshold value, and judging whether the nearest distance is smaller than the distance threshold value, if so, taking the corresponding image point as a potential homonymous point of the nearest transformed object point under the corresponding potential single matrix, and if not, discarding the corresponding image point;

step 4.7: traversing each image point in the image space point set, obtaining potential homonymy points corresponding to each nearest neighbor transform object point in the object space point set one by one under the corresponding potential homography matrix according to the methods of the steps 4.5 to 4.6, and obtaining the potential homonymy point set of the object space point set under the corresponding potential homography matrix according to all the potential homonymy points of all the nearest neighbor transform object points under the corresponding potential homography matrix;

step 4.8: and traversing each potential homography matrix, and obtaining a potential homonymy point set corresponding to the object point set under each potential homography matrix one by one according to the methods from the step 4.4 to the step 4.7.

Further: the specific steps of the step 5 comprise:

and acquiring the number of potential homonymous points in the potential homonymous point set of the object point set under each potential homonymous matrix, and taking the potential homonymous point set corresponding to the maximum value of the number of the potential homonymous points as the target homonymous point set of the object point set to complete homonymous point matching of the object point set.

According to another aspect of the invention, a co-planar cross-ratio constrained homonymous point matching system is provided, which comprises a point set acquisition module, a calculation module, an analysis module and a matching module;

the point set acquisition module is used for acquiring an object space point set and object space point set coordinates corresponding to the object space point set, and acquiring an image space point set of the object space point set in an image and image space point set coordinates corresponding to the image space point set;

the calculating module is used for respectively calculating to obtain an object side seed point intersection ratio set which corresponds to each object point in the object side point set one by one according to the object side point set coordinates based on a coplanar intersection ratio calculating method, and calculating to obtain an image side seed point intersection ratio set which corresponds to an image point located at the central position of the image side point set according to the image side point set coordinates;

the analysis module is used for obtaining a plurality of potential object seed point intersection ratio sets according to the image side seed point intersection ratio set and all object seed point intersection ratio sets;

the calculation module is further used for respectively calculating potential homography matrixes which are in one-to-one correspondence with the seed point intersection ratio sets of each potential object space based on a homography matrix calculation method;

the analysis module is further configured to obtain potential homonymy point sets corresponding to the object point sets one by one under each potential homography matrix respectively according to each potential homography matrix;

and the matching module is used for acquiring a target homonymy point set of the object space point set from all potential homonymy point sets to complete homonymy point matching.

The invention has the beneficial effects that: based on coplanarity cross ratio invariance, a more accurate set of potential object seed point cross ratios can be obtained, a more accurate set of potential homonymous points can be obtained by utilizing the projective transformation relation of the homography matrix, a target homonymous point set is obtained from the set of potential homonymous points, high-precision matching of homonymous points is realized, matching of homonymous points is not needed to be assisted by coding light-reflecting mark points, matching is also not needed to be carried out by establishing feature descriptors, the problems that time consumption is long and mismatching is easy to occur in homonymous point matching in the prior art are avoided, matching precision is high, and the method and the device are extremely suitable for the fields of camera calibration, camera shooting measurement and the like.

According to another aspect of the present invention, there is provided a co-planar cross-ratio constrained homonym matching device, comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program is operable to implement the steps of a co-planar cross-ratio constrained homonym matching method of the present invention.

The invention has the beneficial effects that: the homonymy point matching method is realized by a computer program stored in a memory and running on a processor, a more accurate set of potential object seed point cross ratios can be obtained based on coplane cross ratio invariance, a more accurate set of potential homonymy points can be obtained by utilizing the projective transformation relation of a homography matrix, a target homonymy point set obtained from the set of potential homonymy points realizes high-precision matching of homonymy points, the homonymy point matching is not required to be assisted by coded light-reflecting mark points, the matching is not required to be carried out by establishing a feature descriptor, the problems that the time consumption of homonymy point matching is long and mismatching is easy to occur in the prior art are solved, the matching precision is high, and the homonymy point matching method is extremely suitable for the fields of camera calibration, camera shooting measurement and the like.

In accordance with another aspect of the present invention, there is provided a computer storage medium comprising: at least one instruction which, when executed, performs a step in a co-planar cross-ratio constrained homonym matching method of the present invention.

The invention has the beneficial effects that: the homonymy point matching method has the advantages that homonymy point matching of the homonymy point matching method is achieved by executing a computer storage medium containing at least one instruction, a more accurate potential object seed point cross ratio set can be obtained based on coplane cross ratio invariance, a more accurate potential homonymy point set can be obtained by utilizing the projective transformation relation of a homography matrix, a target homonymy point set obtained from the potential homonymy point set achieves high-precision matching of homonymy points, matching is not needed to be assisted by coded light reflection mark points and is not needed to be conducted by establishing feature descriptors, the problems that time consumption is long and mismatching is prone to occur in homonymy point matching in the prior art are solved, matching precision is high, and the homonymy point matching method is extremely suitable for the fields of camera calibration, camera shooting measurement and the like.

Drawings

FIG. 1 is a schematic diagram of a model defined by the same name points in the present invention;

FIG. 2 is a flowchart illustrating a co-planar cross-ratio constrained homonymy point matching method according to an embodiment of the present invention;

fig. 3 is a schematic flow chart of acquiring an object space point set, coordinates of the object space point set, an image space point set, and coordinates of the image space point set according to a first embodiment of the present invention;

FIGS. 4-1 and 4-2 are schematic diagrams of an object space point set and an image space point set, respectively, according to an embodiment of the present invention;

FIG. 5 is a schematic diagram of an object seed point intersection ratio set corresponding to each object point according to a first embodiment of the present invention;

FIG. 6 is a schematic diagram of a model for cross-ratio invariance definition according to an embodiment of the present invention;

FIG. 7 is a schematic diagram of a model for calculating an intersection ratio based on an intersection ratio invariance according to an embodiment of the present invention;

FIG. 8 is a schematic diagram of five neighbors of a line primer square seed dot in accordance with an embodiment of the present invention;

FIG. 9 is a schematic diagram of five adjacent object point connecting lines according to a first embodiment of the present invention;

FIG. 10 is a flowchart illustrating a process of obtaining an image side seed intersection ratio set according to a first embodiment of the present invention;

FIG. 11 is a schematic flow chart illustrating a process of obtaining a cross-ratio set of a plurality of potential object seed points according to a first embodiment of the present invention;

12-1 to 12-3 are schematic diagrams of models for determining the minimum cross ratio difference of the object seed cross ratio set according to the first embodiment of the present invention;

FIG. 13 is a diagram illustrating the results of some object seed points in the cross-ratio set of potential object seed points according to an embodiment of the present invention;

fig. 14 is a schematic flow chart illustrating a process of obtaining a one-to-one correspondence potential homonymous point set of an object point set under each potential homography matrix according to a first embodiment of the present invention;

FIG. 15 is a diagram illustrating the result of a set of potential homonymous points corresponding to a set of object space points under one of the potential homography matrices according to an embodiment of the present invention;

FIG. 16 is a diagram illustrating the result of a set of potential homonymous points corresponding to a set of object space points under another potential homography matrix according to an embodiment of the present invention;

fig. 17 is a schematic view of a complete process of matching the same-name points of the object space point set according to the first embodiment of the present invention;

fig. 18 is a schematic structural diagram of a co-planar cross-ratio constrained homonymy point matching system according to a second embodiment of the present invention.

Detailed Description

The principles and features of this invention are described below in conjunction with the following drawings, which are set forth by way of illustration only and are not intended to limit the scope of the invention.

The present invention will be described with reference to the accompanying drawings.

In a first embodiment, as shown in fig. 2, a method for matching homonymous points of coplanar cross-ratio constraints includes the following steps:

s1: acquiring an object space point set and object space point set coordinates corresponding to the object space point set, and acquiring an image space point set of the object space point set in an image and image space point set coordinates corresponding to the image space point set;

s2: on the basis of a coplanar intersection ratio calculation method, respectively calculating to obtain an object seed point intersection ratio set which corresponds to each object point in the object point set one by one according to the object point set coordinates, and calculating to obtain an image seed point intersection ratio set which corresponds to an image point located at the central position of the image point set according to the image point set coordinates;

s3: obtaining a plurality of potential object seed point intersection ratio sets according to the image side seed point intersection ratio set and all object seed point intersection ratio sets;

s4: respectively calculating potential homography matrixes which correspond to the seed point cross ratio sets of each potential object space one by one on the basis of a homography matrix calculation method; respectively obtaining potential homonymy point sets corresponding to the object space point sets one by one under each potential homonymy matrix according to each potential homonymy matrix;

s5: and acquiring a target homonymy point set of the object space point set from all potential homonymy point sets to complete homonymy point matching.

The homonymy point matching method of the embodiment can obtain a more accurate set of potential object seed point cross ratios based on coplane cross ratio invariance, can obtain a more accurate set of potential homonymy points by utilizing a projective transformation relation of a homography matrix, can obtain a target homonymy point set from the set of potential homonymy points, realizes high-precision matching of homonymy points, does not need to assist homonymy point matching through coded light-reflecting mark points, does not need to establish a feature descriptor for matching, avoids the problems that time consumption is long and mismatching is easy to occur in homonymy point matching in the prior art, is high in matching precision, and is extremely suitable for the fields of camera calibration, camera shooting measurement and the like.

Preferably, as shown in fig. 3, the specific step of S1 includes:

s1.1: obtaining the object space point set by using a calibration plate, and obtaining the object space point set coordinates corresponding to the object space point set according to a calibration plate coordinate system;

s1.2: and shooting the calibration plate to obtain a calibration plate image, and processing the calibration plate image to obtain the image square point set and the image square point set coordinates corresponding to the image square point set.

The calibration plate is utilized to conveniently obtain a relatively standard and uniform object point set, so that the method for obtaining coordinates of the object point set is simple, meanwhile, the subsequent relatively standard and uniform image point set and the corresponding coordinates of the image point set are conveniently obtained, the coded light-reflecting mark points are not needed to assist the matching of the same-name points, the subsequent matching of the same-name points is facilitated, the matching precision is improved, and the camera calibration and the photogrammetry are facilitated.

Specifically, in this embodiment, an object space point set including 72 object Points is obtained by numbering each reflective mark point on the calibration board, and is set as a point set Points1, as shown in fig. 4-1, and an object space point set coordinate in a calibration board coordinate system is obtained by measuring the calibration board; shooting the calibration plate by using an industrial camera to obtain a corresponding image square point set comprising 72 image Points, setting the image square point set as a point set Points0, and obtaining the coordinates of each image point by using an image processing method as shown in figure 4-2 to obtain the coordinates of the image square point set; it should be noted that, in the present embodiment, the number of the calibration board, for example, 1, 2, 3 … 72, is different from the code in the conventional calibration method, and is only used for marking each reflective mark point.

Preferably, as shown in fig. 5, in S2, the specific step of obtaining an object seed intersection ratio set corresponding to each object point in the object point set includes:

s2a.1: optionally selecting one object point in the object space point set as an object space seed point, obtaining an included angle between a connecting line between the object space seed point and each adjacent object point of the object space seed points and a horizontal direction line, and obtaining N adjacent object points of the corresponding object space seed points from all the adjacent object points of the object space seed points according to the sequence of the included angles from small to large by taking the object space seed point as the center and taking the clockwise direction as the index direction by using a kd-tree index method; wherein N is more than or equal to 5;

s2a.2: connecting N adjacent object points pairwise at intervals of one adjacent object point to obtain N adjacent object point connecting lines, respectively calculating to obtain one-to-one corresponding first intersection ratio on each adjacent object point connecting line according to the object space point set coordinates based on a coplanar intersection ratio calculation method, and obtaining an object space seed point intersection ratio set corresponding to the corresponding object space seed point according to all the first intersection ratios on all the adjacent object point connecting lines;

s2a.3: traversing each object point in the object point set, and obtaining an object seed intersection ratio set corresponding to each object point in the object point set one by one according to methods from S2a.1 to S2a.2;

as shown in fig. 10, in S2, the specific step of obtaining the image side seed intersection ratio set corresponding to the image point located at the center position of the image side point set includes:

s2b.1: taking an image point located at the central position of the image side point set as an image side seed point, and acquiring N adjacent image points of the image side seed point according to the method of S2a.1;

s2b.2: according to the method of S2a.2, N adjacent image point connecting lines are obtained, on the basis of a coplanar intersection ratio calculation method, according to the image space point set coordinates, second intersection ratios which are in one-to-one correspondence on each adjacent image point connecting line are respectively calculated, and according to all the second intersection ratios on all the adjacent image point connecting lines, the image space seed point intersection ratio set corresponding to the image space seed points is obtained.

By using a kd-tree indexing method, any object point is taken as an object seed point, the center of the object seed point and the clockwise direction are taken as indexing directions, adjacent object points which are close to the object seed point are searched out in a concentrated mode from the object point, namely the adjacent object points, and similarly, the image point at the central position in the concentrated image point is taken as an image seed point, and the adjacent image points of the image seed point are searched out according to the same method; according to the method, N adjacent object points (N is more than or equal to 5) are indexed according to the index direction in the clockwise direction and N adjacent image points are indexed according to the same method in the order that the included angle between a connecting line between each object side seed point and each adjacent object point is from small to large, on one hand, the adjacent object points close to the object side seed points and the adjacent image points close to the image side seed points can be obtained, so that on the basis of the property that the cross ratio is not changed, a corresponding object side seed point cross ratio set and an image side seed point cross ratio set are respectively calculated, and the object side seed point cross ratio set which is most matched with the image side seed points is conveniently analyzed according to all the object side seed point cross ratio sets and the image side seed point cross ratio set, namely, a potential object side seed point cross ratio set; on the other hand, after the most matched object seed point intersection ratio set is obtained, the potential homography matrix corresponding to each potential object seed point intersection ratio set one by one can be conveniently obtained according to a homography matrix calculation method in projective transformation, because the homography matrix describes the relation between a plane and a plane in projective geometry and is usually represented by a 3 x 3 matrix, 8 degrees of freedom are provided, namely 8 unknown quantities are provided, at least 4 homography point pairs are needed for solving the homography matrix, and N is more than or equal to 5, at least 5 homography point pair solving formulas can be provided through N adjacent object points and N adjacent image points, namely at least 5 homography matrix calculation equation sets are established, and the corresponding potential homography matrix is conveniently and directly solved through the at least 5 homography matrix calculation equation sets; the specific operation steps of the kd-tree indexing method are in the prior art, and specific details are not described herein again;

because, according to the definition of cross ratio invariance, the number of collinear points satisfying the cross ratio invariance is at least 4, therefore, in the process of calculating the cross ratio set of the corresponding object seed points based on the property of constant cross ratio, at least five adjacent object points are connected in pairs in a manner of separating one adjacent object point at a time to obtain at least five adjacent object point connecting lines (when N is 5, the connecting lines are similar to a pentagram shape), so as to obtain at least four collinear points (at least including two adjacent object points and two intersection points), thereby conveniently calculating the first cross ratio corresponding to the collinear four points on the connecting line of each adjacent object point, because five adjacent object point connecting lines obtain five first cross ratios according to a cross ratio calculation formula, the at least five first cross ratios form an object seed point cross ratio set of corresponding object seed points, and the corresponding image seed point cross ratio set is calculated based on the property of constant cross ratios in the same way; therefore, according to the same method, an object side seed point intersection ratio set corresponding to each object point in the object side point set as an object side seed point and an image side seed point intersection ratio set corresponding to the image point positioned at the central position of the image side point set as an image side seed point can be obtained, all the object side seed point intersection ratio sets and the image side seed point intersection ratio sets can be conveniently analyzed subsequently, and a plurality of potential object side seed point intersection ratio sets matched with the image side seed point intersection ratio sets can be obtained from the object side seed point intersection ratio sets without establishing a feature descriptor; the number of adjacent object points and adjacent image points of each object seed point is generally the same.

Specifically, regarding the definition of the cross ratio invariance in photogrammetry, the model schematic diagram is shown in fig. 6 and 7, where a straight line L represents a straight line in space, A, B, C and D represent four points on the straight line L, S represents a central point of photography when photography is performed, a straight line L represents an image of the straight line L in an image, A, B, C and D four points form four image points a, b, c and c in the image after projective transformation, and the four image points are still collinear and located on the straight line L, then A, B, C and D four points and the four image points a, b, c and c satisfy the cross ratio invariance, and the corresponding cross ratio calculation formula is:

where CR is an intersection ratio satisfied by the A, B, C and D four points and the a, B, C, and C four image points, | AC | is a distance between a point and C, | BD | is a distance between B point and D point, AD | is a distance between a point and D point, | BC | is a distance between B point and C point, | AC | is a distance between a point and C point, | BD | is a distance between B point and D point, | AD | is a distance between a point and D point, and | BC | is a distance between B point and C point.

Specifically, in this example s2a.1, any object seed point is assumed to be SP₀If N is 5, the corresponding five adjacent object points of the rope are respectively P₀、P₁、P₂、P₃And P₄The model diagram of the specific index is shown in fig. 8; in this example s2a.2, five adjacent object point connecting lines obtained by connecting two by two are shown in fig. 9, and are L_P0P2、L_P1P3、L_P2P4、L_P3P0And L_P4P1Then calculate the link L between the neighboring object points_P0P2The specific formula of the first cross ratio is as follows:

wherein, CR₀To connect a line L at a neighboring object point_P0P2First cross ratio of (1), CP₀Is a line L connecting adjacent object points_P0P2Line L connecting to adjacent object point_P1P3Cross point of (2), CP₁Is a line L connecting adjacent object points_P0P2Line L connecting to adjacent object point_P4P1Point of intersection, | P₀CP₀I is the neighboring object point P₀Intersection point CP₀Distance between, | CP₁P₂I is the intersection point CP₁And the adjacent object point P₂Distance between, | P₀P₂I is the neighboring object point P₀And the adjacent object point P₂Distance between, | CP₁CP₀I is the intersection point CP₁Intersection point CP₀The distance between them;

respectively calculating object seed points SP according to the method₀The five first cross ratios constitute the object seed point SP₀Respectively calculating the object seed point cross ratio set and the image seed point which are in one-to-one correspondence with each object seed point (namely each object point in the point set Points 1) by analogy in turn(image point located at the center position in the point set Points 0) of the corresponding image side seed point intersection set.

Preferably, as shown in fig. 11, the specific step of S3 includes:

s3.1: selecting any one object side seed point cross ratio set from all object side seed point cross ratio sets, and performing difference operation on each second cross ratio in the image side seed point cross ratio set and each first cross ratio in one selected object side seed point cross ratio set in a one-to-one correspondence manner to obtain a cross ratio difference set corresponding to one selected object side seed point cross ratio set;

s3.2: comparing the minimum cross ratio difference value in a cross ratio difference value set corresponding to the selected object side seed point cross ratio set with a preset cross ratio threshold value, judging whether the minimum cross ratio difference value is smaller than the cross ratio threshold value, if so, selecting one object side seed point cross ratio set as a potential object side seed point cross ratio set, and if not, discarding the selected object side seed point cross ratio set;

s3.3: and traversing each object seed point cross ratio set in all the object seed point cross ratio sets, and obtaining a plurality of potential object seed point cross ratio sets according to the methods from S3.1 to S3.2.

In order to find out a plurality of potential object seed point cross ratio sets which are matched with the image seed point cross ratio set, N cross ratio difference values which are in one-to-one correspondence with each object seed point cross ratio set are obtained by performing one-to-one correspondence subtraction operation on each first cross ratio in each object seed point cross ratio set and a second cross ratio in the image seed point cross ratio set, namely, the cross ratio difference value sets which are in one-to-one correspondence with each object seed point cross ratio set, and for the N cross ratio difference values (cross ratio difference value sets) which are in correspondence with any object seed point cross ratio set, the minimum cross ratio difference value is compared with a preset cross ratio threshold value, so that whether one selected object seed point cross ratio set is a matched potential object seed point cross ratio set or not can be judged; by the difference making operation and comparison judgment method, homonymy point matching is not required to be assisted by identifying the coded reflective mark points, and a characteristic descriptor is not required to be established, so that the matching degree of the obtained potential object side seed point cross ratio set and the image side seed point cross ratio set is high, a potential homonymy matrix with high matching degree can be conveniently calculated subsequently, and the homonymy point matching precision is effectively improved; the preset cross ratio threshold value can be set and adjusted according to actual conditions.

Specifically, in this embodiment, the difference operation is performed on five first cross ratios corresponding to one object side seed point cross ratio set and five second cross ratios corresponding to the image side seed point cross ratio set one by one, and the minimum cross ratio difference value in the obtained cross ratio difference value set is compared with the cross ratio threshold, and a partial model schematic diagram is shown in fig. 12-1, fig. 12-2, and fig. 12-3; when the difference operation is performed, namely the first cross ratio and the second cross ratio are sequentially subtracted correspondingly, and an absolute value is obtained, and since the corresponding relation of 5 cross ratio values in the two groups is unknown, the difference operation is performed correspondingly in a way of clockwise rotating around the object seed point.

Specifically, in this embodiment S3.3, two sets of potential object seed point intersection ratio sets meeting the requirement are obtained according to the above method, and the object seed points corresponding to the two sets of potential object seed point intersection ratio sets are "No. 49 object seed point" and "No. 53 object seed point" shown in fig. 13.

Preferably, as shown in fig. 14, in S4, the specific step of calculating the potential homography matrix corresponding to each potential object seed intersection ratio set in a one-to-one manner includes:

s4.1: selecting any one potential object side seed point cross ratio set from all potential object side seed point cross ratio sets, selecting at least four adjacent object points from N adjacent object points corresponding to the selected one potential object side seed point cross ratio set, and selecting at least four adjacent image points from N adjacent image points corresponding to the image side seed point cross ratio set;

s4.2: according to the selected at least four adjacent object points and the object space point set coordinates, object point homogeneous coordinates corresponding to the at least four adjacent object points one to one are obtained, and according to the selected at least four adjacent image points and the image space point set coordinates, image point homogeneous coordinates corresponding to the at least four adjacent image points one to one are obtained; calculating to obtain a potential homography matrix corresponding to the cross ratio set of the selected potential object space seed points according to the object point homogeneous coordinates corresponding to at least four adjacent object points one to one and the image point homogeneous coordinates corresponding to at least four adjacent image points one to one;

wherein H^kSelecting a potential homography matrix corresponding to the k-th selected potential object space seed point cross ratio set, wherein i is a positive integer and satisfies that i is more than or equal to 4;

s4.3: and traversing each potential object seed point cross ratio set in all the potential object seed point cross ratio sets, and obtaining the potential homography matrix corresponding to each potential object seed point cross ratio set one by one according to the methods from S4.1 to S4.2.

Since when S3.2 determines a set of seed point cross ratios of potential object space, the corresponding set of seed point cross ratios of object space has N corresponding neighboring object points, and the set of seed point cross ratios of image space has N corresponding neighboring image points, the N adjacent object points and the N adjacent image points are equivalent to the homonymous points which are preliminarily matched, and at least 4 homonymous point pairs are needed for solving the homonymous matrix, and N is more than or equal to 5, so that the N adjacent object points and the N adjacent image points can provide at least 5 homonymous homonym, therefore, at least four adjacent object points and at least four adjacent image points are selected from the N adjacent object points and the N adjacent image points respectively, and homogeneous coordinates (namely object point homogeneous coordinates) corresponding to the at least four adjacent object points and homogeneous coordinates (namely image point homogeneous coordinates) corresponding to the at least four adjacent image points are obtained respectively according to the object space point set coordinates and the image space point set coordinates in the step 1; the corresponding potential object space seed point cross ratio set can be simultaneously solved through the at least four object point homogeneous coordinates and the at least four image point homogeneous coordinates to solve an equation set of a potential homography matrix, and then the corresponding potential homography matrix can be obtained through solving the equation set, so that the mapping transformation relation between the preliminarily matched homography points is obtained, the transformed object space seed points of each image point in the object space point set can be conveniently searched and found out according to the mapping transformation relation, and finally, the more matched potential homography points can be conveniently found out; by the method for calculating the potential homography matrix, under the condition that the potential homography matrix corresponding to each potential object space seed point cross ratio set one by one can be obtained, the potential object space seed point cross ratio set is obtained based on the cross ratio invariance and the homography matrix calculation method is utilized, the obtained potential homography matrix can reflect the real mapping transformation relation between the image space point set and the object space point set, the subsequent homography point matching of the object space point set is facilitated, a feature descriptor is not required to be established, the method is simple and easy to implement, the time consumption is shorter, and the matching precision is higher.

Specifically, N in this embodiment is 5, so for one group of potential object side seed point cross ratio sets, 4 adjacent object points are optionally selected from corresponding 5 adjacent object points, 4 adjacent image points are optionally selected from 5 adjacent image points, object point set coordinates and image point set coordinates obtained in the previous step are obtained respectively to obtain object point homogeneous coordinates of 4 adjacent object points and image point homogeneous coordinates of 4 adjacent image points, 4 equations for solving the potential homography matrix are combined, and the corresponding potential homography matrix is solved; in the same way, two sets of potential homography matrices are calculated.

Preferably, as shown in fig. 14, in S4, the specific step of obtaining a one-to-one correspondence set of potential homonym points of the object point set under each potential homography matrix includes:

s4.4: selecting any potential homography matrix, performing projective transformation on the image space point set by using the selected potential homography matrix, and finding out transformed object space seed points corresponding to each image point of the image space point set under the corresponding potential homography matrix in the object space point set;

s4.5: for any one transformation object seed point, acquiring the nearest transformation object point corresponding to the corresponding transformation object seed point from the object point set by using the kd-tree indexing method;

s4.6: calculating the distance between the nearest transformed object point and the corresponding image point to obtain the nearest distance, comparing the nearest distance with a preset distance threshold value, and judging whether the nearest distance is smaller than the distance threshold value, if so, taking the corresponding image point as a potential homonymous point of the nearest transformed object point under the corresponding potential single matrix, and if not, discarding the corresponding image point;

s4.7: traversing each image point in the image space point set, obtaining potential homonymy points corresponding to each nearest neighbor transformation object point in the object space point set one by one under the corresponding potential homography matrix according to methods from S4.5 to S4.6, and obtaining the potential homonymy point set of the object space point set under the corresponding potential homography matrix according to all the potential homonymy points of all the nearest neighbor transformation object points under the corresponding potential homography matrix;

s4.8: and traversing each potential homography matrix, and obtaining a potential homonymy point set of the object point set in one-to-one correspondence under each potential homography matrix according to the methods from S4.4 to S4.7.

When a potential homography matrix is calculated by a homography matrix calculation method, each image point in an image side point set can be back-calculated to an object side point set by using the potential homography matrix, a transformation object side seed point under the mapping transformation relation represented by the potential homography matrix is found, for the transformation object side seed point back-calculated by any image point, a plurality of adjacent object points corresponding to the transformation object side seed point can be found by using the kd-tree indexing method of the step 2a.1, wherein the adjacent object point closest to the transformation object side seed point is determined as the nearest transformation object point corresponding to the transformation object side seed point; calculating the distance between the nearest transformation object point and the corresponding image point as the nearest distance, when the nearest distance is smaller than a preset distance threshold value, the corresponding image point and the potential homonymous point of the nearest transformation object point corresponding to the nearest distance, otherwise, the image point is not the potential homonymous point of the nearest transformation object point; by the comparison and judgment method of the nearest distance and the distance threshold, the image point which is matched with each nearest transformation object point under the corresponding potential homography matrix to the highest degree can be matched, and each nearest transformation object point is an object point in an object space point set, so that the potential homonymy point set with the highest degree of matching of the object space point set under the corresponding potential homography matrix can be matched, and the target homonymy point set of the object space point set can be conveniently found out from all the potential homonymy point sets under all the potential homography matrices; the preset distance threshold value can be set and adjusted according to actual conditions.

Preferably, the specific step of S5 includes:

In photogrammetry, because the potential homography matrix represents the potential projective transformation relationship between the object side point set and the image side point set, each potential homography matrix is different, so that the obtained potential homonymy point sets are also different, and the number of the potential homonymy points matched in the step 4 is also different; the image side point set is obtained by shooting the object side point set, so that the number of potential homonymous points is the largest, the potential homonymous matrixes of the potential homonymous points are obtained most reasonably, the potential mapping transformation relation between the corresponding object side point set and the image side point set is the most reasonable, and the potential homonymous point set corresponding to the maximum value of the number of the potential homonymous points is the target homonymous point set of the object side point set; the method for obtaining the target homonymous point set is reasonable and effective.

Specifically, in this embodiment, according to the methods of S4.4 to S4.8 described above, potential homonym point sets of object point sets in one-to-one correspondence with two potential homography matrices are obtained, and as a result, as shown in fig. 15 and fig. 16, where the number of potential homonym points shown in fig. 15 is 15, and the number of potential homonym points shown in fig. 16 is 72, the potential homonym point set shown in fig. 16 is the target homonym point set of the object point set corresponding to the calibration board in this embodiment. Fig. 17 shows a complete flow chart of matching the same-name points of the object space point set in this embodiment.

In a second embodiment, as shown in fig. 18, a co-planar cross-ratio constrained homonymy point matching system includes a point set obtaining module, a calculating module, an analyzing module, and a matching module;

Based on coplanarity cross ratio invariance, a more accurate set of potential object seed point cross ratios can be obtained, a more accurate set of potential homonymous points can be obtained by utilizing the projective transformation relation of the homography matrix, a target homonymous point set is obtained from the set of potential homonymous points, high-precision matching of homonymous points is realized, matching of homonymous points is not needed to be assisted by coding light-reflecting mark points, matching is also not needed to be carried out by establishing feature descriptors, the problems that time consumption is long and mismatching is easy to occur in homonymous point matching in the prior art are avoided, matching precision is high, and the method and the device are extremely suitable for the fields of camera calibration, camera shooting measurement and the like.

Third embodiment, based on the first embodiment and the second embodiment, the present embodiment further discloses a co-planar cross-ratio constrained homonymy point matching device, which includes a processor, a memory, and a computer program stored in the memory and executable on the processor, where the computer program implements specific steps S1 to S5 shown in fig. 2 when running.

The homonymy point matching method is realized by a computer program stored in a memory and running on a processor, a more accurate set of potential object seed point cross ratios can be obtained based on coplane cross ratio invariance, a more accurate set of potential homonymy points can be obtained by utilizing the projective transformation relation of a homography matrix, a target homonymy point set obtained from the set of potential homonymy points realizes high-precision matching of homonymy points, the homonymy point matching is not required to be assisted by coded light-reflecting mark points, the matching is not required to be carried out by establishing a feature descriptor, the problems that the time consumption of homonymy point matching is long and mismatching is easy to occur in the prior art are solved, the matching precision is high, and the homonymy point matching method is extremely suitable for the fields of camera calibration, camera shooting measurement and the like.

The present embodiment also provides a computer storage medium having at least one instruction stored thereon, where the instruction when executed implements the specific steps of S1-S5.

The homonymy point matching method based on the homonymy transformation matrix has the advantages that homonymy point matching is achieved by executing a computer storage medium containing at least one instruction, a computer program stored on a memory is run on a processor, a more accurate set of potential object seed point cross ratios can be obtained based on coplane cross ratio invariance, a more accurate set of potential homonymy points can be obtained by utilizing the projective transformation relation of the homonymy matrix, high-precision matching of homonymy points is achieved without the need of encoding reflecting mark points to assist homonymy point matching and establishing feature descriptors to perform matching, the problems that time consumption is long and mismatching is prone to occur in homonymy point matching in the prior art are solved, matching precision is high, and the homonymy point matching method based on the homonymy transformation matrix is extremely suitable for the fields of camera calibration, camera shooting measurement and the like.

Details of S1 to S5 in this embodiment are not described in detail in the first embodiment and fig. 2 to fig. 17, which are not repeated herein.

The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims

1. A homonymy point matching method for coplanar cross-ratio constraint is characterized by comprising the following steps:

2. The coplanar cross-ratio constrained homonymous point matching method according to claim 1, wherein the specific steps of step 1 include:

3. The coplanar cross-ratio constrained homonymous point matching method of claim 1, wherein in the step 2, the specific step of obtaining an object seed cross-ratio set corresponding to each object point in the object point set in a one-to-one manner includes:

4. The coplanar cross-ratio constrained homonymous point matching method according to claim 3, wherein the specific steps of step 3 include:

5. The coplanar cross-ratio constrained homonymous point matching method of claim 4, wherein in the step 4, the specific step of calculating the potential homonymous matrix corresponding to each potential object seed cross-ratio set in a one-to-one manner includes:

Homogeneous coordinates of at least four adjacent image points corresponding to the image-side seed point cross ratio set are (u'_i,v′_i,1)^TThen, the specific formula of the potential homography matrix corresponding to the k-th selected potential object seed point cross ratio set is calculated as follows:

wherein H_kSelecting a potential homography matrix corresponding to the k-th selected potential object space seed point cross ratio set, wherein i is a positive integer and satisfies that i is more than or equal to 4;

6. The coplanar cross-ratio constrained homonymous point matching method of claim 4, wherein in the step 4, the specific step of obtaining a one-to-one correspondence potential homonymous point set of the object space point set under each potential homography matrix comprises:

7. The coplanar cross-ratio constrained homonymous point matching method of claim 6, wherein the specific steps of step 5 include:

8. A co-planar cross ratio constrained homonymous point matching system is characterized by comprising a point set acquisition module, a calculation module, an analysis module and a matching module;

9. Apparatus for co-planar geometric constraint co-site matching, comprising a processor, a memory and a computer program stored in the memory and executable on the processor, the computer program when executed implementing the method steps as claimed in any one of claims 1 to 7.

10. A computer storage medium, the computer storage medium comprising: at least one instruction which, when executed, implements the method steps of any one of claims 1 to 7.