View conversion method

By restoring camera motion parameters and establishing a conversion matrix, the problem that the front view cannot be converted into a top view in the prior art is solved, and more flexible camera installation and wide application are achieved.

CN118900364APending Publication Date: 2024-11-05FORYOU GENERAL ELECTRONICS
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410968495.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-18
Publication Date
2024-11-05

AI Technical Summary

Technical Problem

Existing parking assistive technology cannot convert the front view of an object into a top view when the front view has no top view information, limiting the installation position and application range of the camera.

Method used

By obtaining the spatial coordinates of matching feature points in the front view and the top view, the camera motion parameters are restored, and the transformation matrix is ​​established, thereby realizing the conversion from the front view to the top view.

Benefits of technology

It realizes that view conversion can be performed without the front view information, simplifying the installation conditions of the camera and expanding the application range.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118900364A_ABST
    Figure CN118900364A_ABST
Patent Text Reader

Abstract

The invention provides a view conversion method, which comprises the following steps of: 1, acquiring a front view, and determining a first space coordinate which is a space coordinate of a matched feature point in the front view; 2, a top view is obtained, second space coordinates are determined, and the second space coordinates are space coordinates of matched feature points in the top view; 3, determining a conversion matrix through a fourth preset relational expression according to the first space coordinates and the second space coordinates; and 4, acquiring a front view of the object in real time, and converting the front view into a top view according to the conversion matrix. According to the invention, when the front view does not contain the top view information, the front view of the object can still be converted into the top view, so that the installation condition of the front-view camera is simplified, the installation position of the camera is more convenient, and the application range is wider.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of automatic parking, and in particular to a view conversion method. Background Art

[0002] With the continuous development of automobile technology, parking assistance function, as a tool to improve driver safety and convenience, has gradually become a widespread research object of major automobile companies.

[0003] The parking assistance function is mainly realized by obtaining the top view image around the vehicle body. At present, the conversion of the front view of the object captured by the camera into the top view of the object during parking is mainly realized by perspective transformation technology. However, the perspective transformation technology requires that the captured front view information of the object includes the top view information of the object. This results in that when the captured front view of the object does not have the top view information, the perspective transformation cannot convert the front view of the object into the top view of the object. Summary of the invention

[0004] The present invention provides a view conversion method, which aims to solve the defects in the prior art and achieve the goal of converting the front view of an object into a top view when the front view does not contain top view information, thereby simplifying the installation conditions of the front view camera, making the installation position of the camera more convenient and the application range wider.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is:

[0006] The present invention provides a view conversion method, comprising:

[0007] Step 1: Obtain a front view and determine a first spatial coordinate, where the first spatial coordinate is the spatial coordinate of a matching feature point in the front view;

[0008] Step 2: Obtain a top view and determine a second spatial coordinate, where the second spatial coordinate is the spatial coordinate of a matching feature point in the top view;

[0009] Step 3: Determine a transformation matrix according to the first space coordinates and the second space coordinates through a fourth preset relationship;

[0010] Step 4: Acquire the front view of the object in real time, and convert it into a top view according to the transformation matrix.

[0011] Specifically, the step 1 includes:

[0012] Step 101: Acquire two frames of front view of the target object, and restore the camera motion according to the two frames of front view;

[0013] Step 102: Determine the first spatial coordinates by triangulation.

[0014] Specifically, step 101 includes:

[0015] Step A1, extracting feature points from the two front view frames, performing feature matching, and obtaining matching feature point pairs;

[0016] Step A2: Construct epipolar geometry based on the feature point pair matching and restore the camera motion.

[0017] Specifically, the step A2 includes:

[0018] Step A201, obtaining a first pixel point in a matching feature point pair in a first frame, and obtaining a second pixel point in the matching feature point pair in a second frame;

[0019] Step A202: establishing a first preset relationship according to the imaging principle of a pinhole camera model;

[0020] Step A203, establishing a second preset relationship according to the first coordinates and the second coordinates of the first pixel point and the second pixel point on the normalized plane;

[0021] Step A204: Restore the rotation parameters and translation parameters of the camera using the essential matrix and the eight-point method.

[0022] Specifically, the first preset relationship is:

[0023] s1p1=KP,s2p2=K(RP+t)

[0024] Among them, p1 is the first pixel point in the matching feature point pair, p2 is the second pixel point in the matching feature point pair, s1 and s2 are the depths of the first pixel point and the second pixel point, K is the intrinsic parameter matrix of the camera, R is the rotation parameter of the camera, and t is the translation parameter of the camera.

[0025] Specifically, the second preset relationship is:

[0026]

[0027] Among them, x1 is the first coordinate, x2 is the second coordinate, R is the rotation parameter of the camera, and t is the translation parameter of the camera.

[0028] Specifically, the step 102 includes: determining the first space coordinates according to a third preset relationship.

[0029] Specifically, the third preset relationship is:

[0030] s2x2=s1Rx1+t

[0031]

[0032] Among them, s1 and s2 are the depths of the first pixel and the second pixel, R is the rotation parameter of the camera, and t is the translation parameter of the camera. Represents the antisymmetric matrix of x2.

[0033] Specifically, the fourth preset relationship is:

[0034]

[0035] Among them, α1, α2, …α n is the first spatial coordinate, β1, β2, …β n is the second space coordinate, T is the transposition operation, is the transformation matrix.

[0036] The beneficial effects of the present invention are as follows: the present invention utilizes epipolar geometry to obtain the rotation parameters and translation parameters of the camera from the front view of the object, and then determines the spatial point coordinates of the feature points of the object in the front view and the top view, thereby obtaining a transformation matrix, and converts the spatial point coordinates of the object's front view into the spatial point coordinates of the object's top view through the transformation matrix, completing the conversion of the object's front view to the top view, and realizing that when the front view does not contain top view information (i.e., when the object is photographed in a horizontal posture), the front view of the object can still be converted into a top view, thereby simplifying the installation conditions of the front view camera, making the installation position of the camera more convenient, and having a wider range of applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 is a flow chart of the view conversion method of the present invention;

[0038] Figure 2 It is a schematic diagram of matching feature points of the present invention. DETAILED DESCRIPTION

[0039] The following specifically describes the embodiments of the present invention in conjunction with the accompanying drawings, which are only for reference and illustration and do not constitute a limitation on the scope of patent protection of the present invention.

[0040] In the process described in the specification, claims or drawings of the present invention, the serial numbers of each step (such as step 10, 20, etc.) are included. The serial numbers are only used to distinguish the steps, and the serial numbers themselves do not represent any execution order. It should be noted that the descriptions such as "first" and "second" in this article are only used to distinguish the description objects, etc., and do not represent the order of precedence, nor do they mean that "first", "second", etc. are different types.

[0041] like Figure 1 As shown, this embodiment provides a view conversion method, including:

[0042] Step 1: Obtain a front view and determine a first spatial coordinate, where the first spatial coordinate is the spatial coordinate of a matching feature point in the front view.

[0043] In this embodiment, step 1 includes:

[0044] Step 101: Acquire two frames of front view of the target object, and restore the camera motion according to the two frames of front view.

[0045] Step 102: Determine the first spatial coordinates by triangulation.

[0046] In this embodiment, step 101 includes:

[0047] Step A1: extract feature points from the two front view frames, perform feature matching, and obtain matching feature point pairs.

[0048] Step A2: Construct epipolar geometry based on the feature point pair matching and restore the camera motion.

[0049] In this embodiment, step A2 includes:

[0050] Step A201: Acquire a first pixel point in a matching feature point pair in a first frame, and acquire a second pixel point in the matching feature point pair in a second frame.

[0051] like Figure 2 As shown, the centers of the two cameras are O1 and O2 respectively, p1 in the first frame image I1 is the first pixel point in the matching feature point pair, and correspondingly, p2 in the second frame image I2 is the second pixel point in the matching feature point pair.

[0052] P is the position of the space point corresponding to p1 and p2. In the coordinate system of the first frame, the spatial position of P is P = [X, Y, Z] T .

[0053] Step A202: establishing a first preset relationship according to the imaging principle of the pinhole camera model.

[0054] In this embodiment, the first preset relationship is:

[0055] s1p1=KP,s2p2=K(RP+t)

[0056] Among them, p1 is the first pixel point in the matching feature point pair, p2 is the second pixel point in the matching feature point pair, s1 and s2 are the depths of the first pixel point and the second pixel point, K is the intrinsic parameter matrix of the camera, R is the rotation parameter of the camera, and t is the translation parameter of the camera.

[0057] According to the equality in the sense of scale, the above formula can be written as:

[0058]

[0059] Step A203: establishing a second preset relationship according to the first coordinates and the second coordinates of the first pixel point and the second pixel point on the normalized plane.

[0060] In this embodiment, the second preset relationship is:

[0061]

[0062] Among them, x1 is the first coordinate, x2 is the second coordinate, R is the rotation parameter of the camera, and t is the translation parameter of the camera.

[0063] Step A204: Restore the rotation parameter R and translation parameter t of the camera using the essential matrix and the eight-point method.

[0064] In this proposal, the step 102 includes: determining the first space coordinates according to a third preset relationship.

[0065] In this embodiment, the third preset relationship is:

[0066] s2x2=s1Rx1+t

[0067]

[0068] Among them, s1 and s2 are the depths of the first pixel and the second pixel, R is the rotation parameter of the camera, and t is the translation parameter of the camera. Represents the antisymmetric matrix of x2.

[0069] In this embodiment, the spatial coordinates corresponding to the n feature points in the front view can be expressed as:

[0070] α1=(x1,y1,z1) T ,α2=(x2,y2,z2) T ,α3=(x3,y3,z3) T ,…,α n =(x n ,y n ,z n ) T .

[0071] Step 2: Obtain a top view and determine a second spatial coordinate, where the second spatial coordinate is the spatial coordinate of the matching feature point in the top view.

[0072] In this embodiment, the method for determining the second space coordinates in the top view is the same as the method for determining the first space coordinates in step 1.

[0073] In this embodiment, the spatial coordinates corresponding to the n feature points in the top view can be expressed as:

[0074] β1=(u1,v1,k1) T ,β2=(u2,v2,k2) T ,β3=(u3,v3,k3) T ,…,β n =(u n ,v n ,k n ) T .

[0075] Step 3: Determine a transformation matrix H according to the first space coordinates and the second space coordinates through a fourth preset relationship.

[0076] In this embodiment, the fourth preset relationship is:

[0077]

[0078] Among them, α1, α2, …α n is the first spatial coordinate, β1, β2, …β n is the second space coordinate, T is the transposition operation, is the transformation matrix.

[0079] Step 4: Acquire the front view of the object in real time, and convert it into a top view according to the transformation matrix H.

[0080] The above disclosure is only a preferred embodiment of the present invention, which cannot be used to limit the scope of protection of the present invention. Therefore, equivalent changes made according to the scope of the patent application of the present invention are still within the scope covered by the present invention.

Claims

1. A view conversion method, It is characterized in that include: Step 1: Obtain a front view and determine a first spatial coordinate, where the first spatial coordinate is the spatial coordinate of a matching feature point in the front view; Step 2: Obtain a top view and determine a second spatial coordinate, where the second spatial coordinate is the spatial coordinate of a matching feature point in the top view; Step 3: Determine a transformation matrix according to the first space coordinates and the second space coordinates through a fourth preset relationship; Step 4: Acquire the front view of the object in real time, and convert it into a top view according to the transformation matrix.

2. The view conversion method according to claim 1, It is characterized in that The step 1 comprises: Step 101: Acquire two frames of front view of the target object, and restore the camera motion according to the two frames of front view; Step 102: Determine the first spatial coordinates by triangulation.

3. The view conversion method according to claim 2, It is characterized in that The step 101 comprises: Step A1, extracting feature points from the two front view frames, performing feature matching, and obtaining matching feature point pairs; Step A2: Construct epipolar geometry based on the feature point pair matching and restore the camera motion.

4. The view conversion method according to claim 3, It is characterized in that The step A2 comprises: Step A201, obtaining a first pixel point in a matching feature point pair in a first frame, and obtaining a second pixel point in the matching feature point pair in a second frame; Step A202: establishing a first preset relationship according to the imaging principle of a pinhole camera model; Step A203, establishing a second preset relationship according to the first coordinates and the second coordinates of the first pixel point and the second pixel point on the normalized plane; Step A204: Restore the rotation parameters and translation parameters of the camera using the essential matrix and the eight-point method.

5. The view conversion method according to claim 4, It is characterized in that The first preset relationship is: s 1 p 1 =KP,s 2 p 2 =K(RP+t) Among them, p 1 is the first pixel in the matching feature point pair, p 2 To match the second pixel in the feature point pair, s 1 、s 2 is the depth of the first pixel and the second pixel, K is the intrinsic parameter matrix of the camera, R is the rotation parameter of the camera, and t is the translation parameter of the camera.

6. The view conversion method according to claim 5, It is characterized in that The second preset relationship is: Among them, x 1 is the first coordinate, x 2 is the second coordinate, R is the rotation parameter of the camera, and t is the translation parameter of the camera.

7. The view conversion method according to claim 6, It is characterized in that The step 102 includes: determining the first space coordinates according to a third preset relationship.

8. The view conversion method according to claim 8, It is characterized in that The third preset relationship is: s 2 x 2 =s 1 Rx 1 +t Among them, s 1 、s 2 is the depth of the first pixel and the second pixel, R is the rotation parameter of the camera, t is the translation parameter of the camera, Represents the antisymmetric matrix of x2.

9. The view conversion method according to claim 8, It is characterized in that The fourth preset relationship is: Among them, α 1 , α 2 , …α n is the first space coordinate, β 1 , β 2 ,…β n is the second space coordinate, T is the transposition operation, is the transformation matrix.