A target area determination method based on marker detection

Abstract

The present invention relates to a target area determination method based on marker detection, comprising: establishing a positional relationship between a marker and a target area in a three-dimensional model space; performing marker detection on n+1 frames of images through a target detection network; matching the detection frame center of the n+1th frame of the image with the marker; obtaining the pixel coordinates Un of the previous n frames of the camera based on a pixel point set U; calculating a basic matrix F according to the pixel point set U and the pixel coordinates Un; triangulating the pixel point set U according to the basic matrix F and camera parameters C to obtain a point cloud P; matching the point cloud Ptag of the marker in the three-dimensional model space with the point cloud P, and obtaining a coordinate matching relationship Rt; mapping the detection area of ​​the three-dimensional model space to the actual space according to the coordinate matching relationship Rt, and projecting the detection area of ​​the actual space to the camera coordinate system through the camera parameters C to obtain a target area R. The present invention can automatically adjust the position of the recognition area with the help of the marker.

Description

Technical Field

[0001] The present invention relates to the technical field of target detection and recognition, and in particular to a target area determination method based on marker detection. Background Art

[0002] Most of the existing target recognition algorithms, such as area intrusion, leaving the post, and personnel counting, require a fixed camera installation position and angle to determine the recognition area (that is, the area photographed by the camera). When the camera deviates, the recognition area needs to be re-marked. This traditional method is not only time-consuming and labor-intensive, but also the accuracy of manually adjusting the recognition area photographed by the camera is not high enough.

[0003] With the continuous updating of information technology and image processing methods, there is an urgent need to find a method to effectively determine the recognition area. Only in this way, when the camera moves, the recognition area it captures can also be updated. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a target area determination method based on marker detection, which can automatically adjust the position of the recognition area within a certain position and angle range with the help of the marker.

[0005] The technical solution adopted by the present invention to solve the technical problem is: to provide a target area determination method based on marker detection, comprising:

[0006] Step (1): establishing a positional relationship between a marker and a target area in a three-dimensional model space, wherein the target area is an area captured by a camera;

[0007] Step (2): perform marker detection on n+1 frames of images taken by the camera at different positions through a target detection network, wherein the n+1th frame of image has a complete marker, and the target detection network is provided with a detection frame for detecting the marker;

[0008] Step (3): Match the center of the detection frame of the n+1th frame image with the marker;

[0009] Step (4): Based on the pixel point set U in the detection frame of the n+1th frame image, obtain the pixel coordinates Un of the camera in different positions of the previous n frames;

[0010] Step (5): Calculate the basic matrix F of the camera posture from the n+1th frame to the previous n frames according to the pixel point set U and the pixel coordinates Un;

[0011] Step (6): triangulate the pixel point set U according to the basic matrix F and the camera parameter C to obtain a point cloud P; wherein the point cloud P corresponds to a marker in the actual space;

[0012] Step (7): Match the point cloud Ptag of the marker in the three-dimensional model space with the point cloud P to obtain the coordinate matching relationship Rt between the three-dimensional model space and the actual space;

[0013] Step (8): Map the detection area of ​​the three-dimensional model space to the actual space according to the coordinate matching relationship Rt to obtain the detection area of ​​the actual space, and project the detection area of ​​the actual space to the camera coordinate system through the camera parameter C to obtain the target area R.

[0014] The target detection network in step (2) is an SSD network.

[0015] The step (4) is specifically as follows: performing optical flow tracking on the pixel point set U within the detection frame of the n+1th frame image to obtain the pixel coordinates Un of the camera in the previous n frames at different positions.

[0016] The step (7) matches the point cloud Ptag of the marker in the three-dimensional model space with the point cloud P through the ICP algorithm to obtain the coordinate matching relationship Rt between the three-dimensional model space and the actual space.

[0017] Beneficial Effects

[0018] Due to the adoption of the above-mentioned technical solution, the present invention has the following advantages and positive effects compared with the prior art: the area determination method provided by the present invention can automatically adjust the position of the recognition area within a certain range of position and angle changes with the help of a marker, and the present invention can be applied to more flexible movable camera scenes such as robot vision. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is a method flow chart of an embodiment of the present invention;

[0020] Figure 2 It is a positional relationship diagram between the three-dimensional model space marker and the target area according to an embodiment of the present invention. DETAILED DESCRIPTION

[0021] The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms fall within the scope limited by the appended claims of the application equally.

[0022] The embodiment of the present invention relates to a target area determination method based on marker detection, see Figure 1 ,include:

[0023] Step (1): Establish the positional relationship between the marker and the target area R in the three-dimensional model space. The target area R is the area captured by the camera. Figure 2 ;

[0024] Step (2): perform marker detection on n+1 frames of images taken by the camera at different positions through a target detection network, wherein the n+1th frame of image has a complete marker, and the target detection network is provided with a detection frame for detecting the marker;

[0025] The target detection network in step (2) is an SSD network, which has a fast running speed and high detection accuracy and can be applied to real-time image detection;

[0026] Step (3): Match the center of the detection frame of the n+1th frame image with the marker;

[0027] Step (4): Based on the pixel point set U in the detection frame of the n+1th frame image, obtain the pixel coordinates Un of the camera in the previous n frames at different positions; wherein the pixel coordinates Un refer to the pixel point set of each detection frame in the previous n frames;

[0028] The step (4) specifically includes: performing optical flow tracking on the pixel point set U in the detection frame of the n+1th frame image to obtain the pixel coordinates Un of the camera in the previous n frames at different positions;

[0029] Step (5): Select a pixel point set of a frame of image from the pixel coordinates Un, and combine the pixel point set U to calculate the basic matrix F of the camera posture from the n+1th frame to the previous n frames;

[0030] Step (6): triangulate the pixel point set U according to the basic matrix F and the camera parameter C (i.e., the camera intrinsic parameter matrix) to obtain a point cloud P. The triangulation is a prior art and will not be described in detail here. The point cloud P corresponds to a marker in the actual space.

[0031] Step (7): Match the point cloud Ptag of the marker in the three-dimensional model space with the point cloud P to obtain the coordinate matching relationship Rt between the three-dimensional model space (i.e., the position of the camera in the three-dimensional model space) and the actual space (i.e., the current position of the camera);

[0032] The step (7) matches the point cloud Ptag of the marker in the three-dimensional model space with the point cloud P by using the ICP algorithm to obtain a coordinate matching relationship Rt between the three-dimensional model space and the actual space;

[0033] Step (8): Map the detection area in the three-dimensional model space to the actual space according to the coordinate matching relationship Rt to obtain the detection area in the actual space, and project the detection area in the actual space to the camera coordinate system through the camera parameter C to obtain the target area R, which is the area in the image taken by the camera.

[0034] The foregoing description of specific exemplary embodiments of the present invention is for the purpose of illustration and demonstration. These descriptions are not intended to limit the present invention to the precise form disclosed, and it is clear that many changes and variations can be made based on the above teachings. The purpose of selecting and describing the exemplary embodiments is to explain the specific principles of the present invention and its practical application, so that those skilled in the art can realize and utilize various different exemplary embodiments of the present invention and various different selections and changes. The scope of the present invention is intended to be limited by the claims and their equivalents.

Claims

1. A target area determination method based on marker detection, characterized in that: include: Step (1): establishing a positional relationship between a marker and a target area in a three-dimensional model space, wherein the target area is an area captured by a camera; Step (2): perform marker detection on n+1 frames of images taken by the camera at different positions through a target detection network, wherein the n+1th frame of image has a complete marker, and the target detection network is provided with a detection frame for detecting the marker; Step (3): Match the center of the detection frame of the n+1th frame image with the marker; Step (4): Based on the pixel point set U in the detection frame of the n+1th frame image, obtain the pixel coordinates Un of the camera in different positions of the previous n frames; Step (5): Calculate the basic matrix F of the camera posture from the n+1th frame to the previous n frames according to the pixel point set U and the pixel coordinates Un; wherein the pixel coordinates Un refer to the pixel point set of each detection frame in the previous n frames; Step (6): triangulate the pixel point set U according to the basic matrix F and the camera parameter C to obtain a point cloud P; Wherein, the point cloud P corresponds to a marker in the actual space; Step (7): Match the point cloud Ptag of the marker in the three-dimensional model space with the point cloud P to obtain the coordinate matching relationship Rt between the three-dimensional model space and the actual space; Step (8): Map the detection area of ​​the three-dimensional model space to the actual space according to the coordinate matching relationship Rt to obtain the detection area of ​​the actual space, and project the detection area of ​​the actual space to the camera coordinate system through the camera parameter C to obtain the target area R.

2. The target area determination method based on marker detection according to claim 1, characterized in that: The target detection network in step (2) is an SSD network.

3. The target area determination method based on marker detection according to claim 1, characterized in that: The step (4) is specifically as follows: performing optical flow tracking on the pixel point set U within the detection frame of the n+1th frame image to obtain the pixel coordinates Un of the camera in the previous n frames at different positions.

4. The target area determination method based on marker detection according to claim 1, characterized in that: The step (7) matches the point cloud Ptag of the marker in the three-dimensional model space with the point cloud P through the ICP algorithm to obtain the coordinate matching relationship Rt between the three-dimensional model space and the actual space.

Images