Device and method for collecting multi-view images of target objects and automatically marking observation angles
By combining fixed vision sensors and mobile vision sensors, using asymmetric coplanar targets and fixed targets to calculate position transformation, the difficulty of estimating translation distances when the rotation angle is large and the problem of hand-held occlusion of fixed vision sensors is achieved, achieving high Observation angle marking of accuracy.
Patent Information
- Application Number
- CN202010987810.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-09-18
- Publication Date
- 2024-11-19
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
It is difficult to estimate the translation distance when the rotation angle of the mobile vision sensor is large, and the spatial state estimation accuracy of the fixed vision sensor is insufficient in the case of hand-held occlusion, so it is impossible to accurately obtain the observation angle tag.
Using a fixed vision sensor and a mobile vision sensor, the pose transformation matrix between the two is calculated through asymmetric coplanar targets and fixed targets, and the relative pose estimation is realized, and the Kalman filtering fusion perception results are combined to improve the observation angle. Marking accuracy.
It realizes high-precision viewing angle tag acquisition at large and continuous changes angles, solves the problem of insufficient position estimation accuracy of the sensor itself, and improves the accuracy and stability of the tag.
Smart Images

Figure CN112085732B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of sensor detection technology, and more specifically, relates to a device and method for multi-view image acquisition and automatic marking of observation angles of a target object. Background Art
[0002] Obtaining data with accurate labels is a core problem of supervised learning, which is popular in machine learning. Visual data contains rich and diverse information and is an important tool for smart devices to perceive the environment. Using visual data and supervised learning methods to achieve robot target grasping, target pose recognition, and target six-degree-of-freedom tracking requires accurate camera observation angle labels. Therefore, methods for obtaining observation pose labels and multi-angle observation information of target objects have attracted the attention of many scholars.
[0003] Mobile visual sensors can estimate their own position and posture by processing environmental information. In this type of method (for example, see Karlsson N, Bernardo ED, Ostrowski J, etc. The vSLAM Algorithm for Robust Localization and Mapping [C] / / Proceedings of the 2005 IEEE International Conference on Robotics and Automation. IEEE, 2006.; Goncalves, Luis, Di Bernardo, Enrico, Benson, Dave, etc. A Visual Front-end for Simultaneous Localization and Mapping [J]. ieee icra, 2005.), most mobile sensors rely on information such as key features in the image to estimate their own spatial state, which has the advantages of less sensor dependence and a simple and efficient position acquisition process. However, this type of method is difficult to estimate the scale of its own translational motion, requires the environment or target object to have continuously trackable features, and is prone to losing the tracked features when the visual sensor has a large rotation angle.
[0004] The spatial state of the target object can be estimated by a fixed visual sensor. This type of method (for example, refer to Kehl W, Manhardt F, Tombari F, etc. SSD-6D: Making RGB-based 3D detection and 6D pose estimation great again [C] / / 2017 IEEE International Conference on Computer Vision (ICCV). IEEE, 2017.; Asif U, Bennamoun M, Sohel F. Real-time pose estimation of rigid objects using RGB-D imagery [C] / / IEEE Conference on Industrial Electronics & Applications. IEEE, 2013.) estimates and tracks the six-degree-of-freedom spatial state of the target object through visual information, which can be used to obtain the observation angle of the visual sensor. The above method can effectively avoid the problem of visual feature loss caused by excessive rotation angle of the sensor. However, this type of method is difficult to determine the correlation between the housing of the mobile visual sensor and the optical center coordinate system, and the accuracy of spatial state estimation in the presence of handheld occlusion is difficult to meet the needs of label collection. Summary of the invention
[0005] In view of the above defects or improvement needs of the prior art, the present invention provides a device and method for multi-perspective image acquisition and automatic marking of observation angles of target objects, which aims to overcome the shortcomings of mobile visual sensor spatial state estimation being affected by rotation and difficult to accurately obtain translation distance information, as well as the acquisition perspective occlusion and coordinate conversion problems existing in third-party fixed visual sensor spatial state estimation.
[0006] To achieve the above-mentioned object, according to one aspect of the present invention, there is provided a device for multi-view image acquisition and automatic marking of observation angles of a target object, comprising: a fixed visual sensor, an asymmetric coplanar target, a mobile visual sensor, a base coordinate plate and a fixed target;
[0007] The asymmetric coplanar target is fixed above the mobile vision sensor and moves simultaneously with the mobile vision sensor in the three-dimensional Cartesian space relative to the object to be measured; the fixed target is printed on the surface of the base coordinate plate, and the object to be measured is placed on the base coordinate plate;
[0008] Fixed visual sensor, used to obtain fixed-viewing angle images of the object being measured;
[0009] Mobile vision sensor, used to obtain images of the object under test with changing viewing angles;
[0010] An asymmetric coplanar target is used to obtain a pose transformation matrix between a fixed vision sensor and a mobile vision sensor, thereby calculating a relative pose between the fixed vision sensor and the mobile vision sensor;
[0011] Fixed targets are used to estimate the relative pose of the mobile vision sensor with respect to the base coordinate plate.
[0012] Furthermore, the asymmetric coplanar target includes a plurality of target points distributed in an asymmetric coplanar structure; each target point is fixed on a corresponding target point support rod; and each target point support plate is fixed on the same support base.
[0013] Furthermore, the number of targets is selected in the range of 4-6.
[0014] Furthermore, the fixed targets are distributed at the four corners of the base coordinate plate; each fixed target adopts black and white squares distributed in a nine-square grid; different combinations of black and white squares are used to distinguish different positions of the base coordinate plate.
[0015] Furthermore, the fixed visual sensor uses a fixed-focus wide-angle lens with a focal length less than 3.5 mm, the distance from the object to be measured is 10 to 15 times the outer diameter of the object to be measured, and the object to be measured is overlooked at an observation angle of 40° to 50°.
[0016] According to another aspect of the present invention, a method for multi-view image acquisition and automatic marking of observation angles of a target object is provided, comprising:
[0017] S1. Before acquiring multi-view images of the object to be measured, calibrate the coordinate transformation relationship between the base coordinate plate and the fixed visual sensor:
[0018] S101. Acquire an image of an asymmetric coplanar target from a fixed vision sensor, and solve the relative position of the asymmetric coplanar target and the fixed vision sensor based on internal parameters of the fixed vision sensor; wherein the asymmetric coplanar target is located at any position set within the viewing angle of the fixed sensor;
[0019] S102. Collecting an image of a fixed target through a mobile visual sensor, and estimating the relative position and posture of the mobile visual sensor and the base coordinate plate through a homography matrix method;
[0020] S103. Solving the coordinate transformation relationship between the asymmetric coplanar target and the mobile vision sensor according to the least squares method;
[0021] S104. Based on the relative posture of the asymmetric coplanar target and the fixed vision sensor in S101 and the coordinate transformation relationship between the asymmetric coplanar target and the mobile vision sensor in S103, the posture of the mobile vision sensor relative to the fixed vision sensor is obtained; based on the posture of the mobile vision sensor relative to the base coordinate plate in S102, the coordinate transformation relationship between the base coordinate plate and the fixed vision sensor is obtained;
[0022] S2. Place the object to be measured on the base coordinate plate, and place the mobile vision sensor at any position where it can capture the object to be measured and the fixed target, and perform multi-view image acquisition and observation angle marking on the object to be measured:
[0023] S201. Obtain the position and posture of the asymmetric coplanar target through the fixed vision sensor, calculate the position and posture of the fixed vision sensor relative to the mobile vision sensor through the coordinate transformation relationship between the asymmetric coplanar target and the mobile vision sensor obtained in S103, calculate the position and posture of the mobile vision sensor relative to the base coordinate plate through the coordinate transformation relationship between the base coordinate plate and the fixed vision sensor obtained in S104, and use it as the perception result of the fixed vision sensor;
[0024] S202. Collecting an image of a fixed target through a mobile vision sensor, estimating the position and posture of the mobile vision sensor relative to the base coordinate plate, and using it as a perception result of the mobile vision sensor;
[0025] S203. Use Kalman filtering to fuse the perception results of the fixed visual sensor and the mobile visual sensor, use the fusion result as the observation angle marking result, and collect the image of the object under the current observation angle through the mobile visual sensor.
[0026] Furthermore, the relative pose calculation formula of the asymmetric coplanar target and the fixed vision sensor is:
[0027] T=A -1 H
[0028]
[0029] Among them, T is the pose parameter of the asymmetric coplanar target in the fixed vision sensor, A is the internal parameter of the fixed vision sensor obtained by offline calibration, H is the homography matrix, and u i ,v i is the coordinate of target i in the image collected by the fixed vision sensor, x i ,y i is the coordinate value of the target point in the target coordinate system, i=1,2,3,4.
[0030] Furthermore, the coordinate transformation relationship between the asymmetric coplanar target and the mobile vision sensor is calculated using the following formula:
[0031]
[0032] Among them, m and n are different postures of the mobile vision sensor, T m 、T n are the pose parameters of the asymmetric coplanar target in pose m and pose n under the fixed vision sensor, B is the pose parameter of the mobile vision sensor under the base coordinate plate, and X is the coordinate transformation relationship between the asymmetric coplanar target and the mobile vision sensor.
[0033] In general, compared with the prior art, the above technical solution conceived by the present invention can achieve the following beneficial effects.
[0034] The method for estimating the observation angle based on the mobile visual sensor's own collected data relies on stable and traceable image features, which is difficult to deal with the situation where the sensor rotates at a large angle and cannot estimate the translation distance parameters; and the method based on the third-party fixed visual sensor has the problems of collection angle occlusion and coordinate conversion; the present invention aims at the scene where the target perspective and image data need to be collected from multiple angles with large-scale continuous changes. With the help of the pose estimation and information fusion results of the fixed visual sensor and the mobile visual sensor, the high-precision sensor perspective label and visual information are obtained synchronously, which solves the problem that the sensor label acquisition accuracy under pose acquisition cannot be met by the visual sensor's own pose estimation alone. The label has high accuracy and good stability and real-time performance. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] Figure 1 It is a schematic diagram of the structure of the device for collecting multi-view images of target objects and automatically marking observation angles provided by the present invention;
[0036] Figure 2 Schematic diagram of the asymmetric coplanar target structure provided by the present invention;
[0037] Figure 3 It is a flow chart of the method for multi-view image acquisition and automatic marking of observation angles of target objects provided by the present invention;
[0038] Figure 4 is a set of target observation angle estimation results of the present invention;
[0039] Among them, 1 is a fixed visual sensor, 2 is an asymmetric coplanar target, 3 is a mobile visual sensor, 4 is a base coordinate plate, 6 is a fixed target, 21 is a target point, 22 is a target point support rod, and 23 is a support. DETAILED DESCRIPTION
[0040] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.
[0041] like Figure 1 As shown, the first embodiment of the present invention provides a device for multi-view image acquisition and observation angle automatic marking of a target object, comprising: a fixed visual sensor 1, an asymmetric coplanar target 2, a mobile visual sensor 3, a base coordinate plate 4, and a fixed target 6;
[0042] A fixed visual sensor 1, used to obtain a fixed-viewing angle image of a measured object 5;
[0043] The asymmetric coplanar target 2 is fixed above the mobile vision sensor 3; the user holds the mobile vision sensor 3 to make the asymmetric coplanar target 2 and the mobile vision sensor 3 move simultaneously in the three-dimensional Cartesian space relative to the object 5 to be measured, wherein the mobile vision sensor 3 is used to obtain the image of the object 5 with a changed viewing angle; the asymmetric coplanar target 2 is used to obtain the posture transformation matrix between the fixed vision sensor 1 and the mobile vision sensor 3, so as to calculate the relative posture between the fixed vision sensor 1 and the mobile vision sensor 3;
[0044] The present invention designs the target as follows Figure 2 The asymmetric coplanar structure shown in FIG. 21 is a target, 22 is a target support rod, and 23 is a support; preferably, the angles between the target points are 70°, 70°, 70°, and 85°, respectively, and the lengths of the target support rods are 50, 50, 40, and 70 mm, respectively. The asymmetric structure ensures the uniqueness of the pose transformation matrix between the fixed visual sensor and the target, so that the relative pose between the fixed visual sensor 1 and the mobile visual sensor 3 can be obtained by calculating the pose transformation matrix; 4 points are the minimum number of target points required, and the number of target points is selected in the range of 4-6 points.
[0045] In order to measure general small industrial parts, the base coordinate plate 4 is set to a size of 400 mm×400 mm for carrying the object to be measured.
[0046] The fixed target 6 is printed on the surface of the base coordinate plate 4, and the object 5 to be measured is placed on the base coordinate plate 4; the fixed target 6 is used to estimate the relative position and posture of the mobile visual sensor 3 relative to the base coordinate plate 4, and the black and white squares on the target provide the mobile visual sensor 3 with the corner points required for estimating the relative position and posture. Since at least four corner points are required to calculate the position and posture of the object to be measured, the fixed target 6 adopts the form of a nine-square grid (different combinations of black and white squares can be used to distinguish different positions of the base coordinate plate 4), and the total number of black and white squares is not limited. In addition, in order to avoid the corner points being blocked and the position and posture estimation being unable to be completed, the fixed targets 6 are arranged at the four corners of the base coordinate plate 4.
[0047] The fixed visual sensor 1 in the present invention adopts a fixed-focus wide-angle lens with a focal length less than 3.5 mm, the distance from the object to be measured is 10 to 15 times the outer diameter of the object to be measured, and the object to be measured is overlooked at an observation angle of 40° to 50°; the fixed visual sensor 1 and the mobile visual sensor 3 can use cognex in-sight 7000.
[0048] The basic working principle of the above device is as follows: first, the fixed visual sensor 1 collects the image of the asymmetric target 2, and solves the pose of the mobile visual sensor 3 relative to the base coordinate plate 4 through the collected image and the internal parameters of the fixed visual sensor 1; then, the mobile visual sensor 3 collects the image containing the base coordinate plate 4 and the object 5 to be measured, and estimates its own pose relative to the base coordinate plate 4 through the image; finally, the above two pose estimation results are subjected to information fusion calculation to obtain the accurate pose of the mobile visual sensor 3 relative to the base coordinate plate 4. Compared with a single mobile visual sensor, the innovation of this device lies in the introduction of the fixed visual sensor 1 and the asymmetric target 2 for accurate pose estimation, and the pose estimation accuracy of the mobile visual sensor 3 is improved by means of the pose estimation and information fusion results of the fixed visual sensor and the mobile visual sensor.
[0049] A second embodiment of the present invention provides a method for multi-view image acquisition and automatic observation angle marking of multi-target objects, comprising:
[0050] S1. Before acquiring multi-view images of the object to be measured, calibrate the coordinate transformation relationship between the base coordinate plate 4 and the fixed visual sensor 1:
[0051] S101. Acquire an image of the asymmetric coplanar target 2 from the fixed vision sensor 1, and solve the relative position of the asymmetric coplanar target 2 and the fixed vision sensor 1 based on the internal parameters of the fixed vision sensor 1; wherein the asymmetric coplanar target 2 is located at any position set within the viewing angle of the fixed sensor;
[0052] The relative posture calculation formula of the asymmetric coplanar target 2 and the fixed visual sensor 1 is:
[0053] T=A -1 H
[0054]
[0055] Where T is the pose parameter of the asymmetric coplanar target 2 in the fixed visual sensor 1, A is the internal parameter of the fixed visual sensor 1 obtained by offline calibration, H is the homography matrix, u i ,v i is the coordinate of target point i in the image collected by fixed vision sensor 1, x i ,y i is the coordinate value of the target point in the target coordinate system, i=1,2,3,4.
[0056] S102. The image of the fixed target 6 is collected by the mobile visual sensor 3, and the relative position and posture of the mobile visual sensor 3 and the base coordinate plate 4 are estimated by the homography matrix method;
[0057] S103. Solving the coordinate transformation relationship between the asymmetric coplanar target 2 and the mobile visual sensor 3 according to the least squares method;
[0058] The coordinate transformation relationship between the asymmetric coplanar target 2 and the mobile visual sensor 3 is calculated using the following formula:
[0059]
[0060] The above formula can be solved by the least square method, where m and n are different positions of the mobile vision sensor 3, T m , T n are the pose parameters of the asymmetric coplanar target 2 under the fixed vision sensor 1 when it is in pose m and pose n, B is the pose parameter of the mobile vision sensor 3 under the base coordinate plate 4, and X is the coordinate transformation relationship between the asymmetric target 2 and the mobile vision sensor 3.
[0061] S104. Based on the relative posture of the asymmetric coplanar target 2 and the fixed vision sensor 1 in S101 and the coordinate transformation relationship between the asymmetric coplanar target 2 and the mobile vision sensor 3 in S103, the posture of the mobile vision sensor 3 relative to the fixed vision sensor 1 is obtained; based on the posture of the mobile vision sensor 3 relative to the base coordinate plate 4 in S102, the coordinate transformation relationship between the base coordinate plate 4 and the fixed vision sensor 1 is obtained.
[0062] S2. Place the measured object 5 on the base coordinate plate 4, place the mobile sensor 3 at any position where the measured object 5 and the fixed target 6 can be captured, and perform multi-view image acquisition on the measured object:
[0063] S201. Acquire the position and posture of the asymmetric coplanar target 2 through the fixed vision sensor 1, calculate the position and posture of the fixed vision sensor 1 relative to the mobile vision sensor 3 through the coordinate transformation relationship between the asymmetric coplanar target 2 and the mobile vision sensor 3 obtained through S103, calculate the position and posture of the mobile vision sensor 3 relative to the base coordinate plate 4 through the coordinate transformation relationship between the base coordinate plate 4 and the fixed vision sensor 1 obtained through S104, and use it as the perception result of the fixed vision sensor.
[0064] S202. The image of the fixed target 6 is collected by the mobile visual sensor 3, and the position of the mobile visual sensor 3 relative to the base coordinate plate 4 is estimated, which is used as the perception result of the mobile visual sensor;
[0065] S203. Based on Kalman filtering, the perception results of the fixed visual sensor and the perception results of the mobile visual sensor are fused as the estimation result of the observation posture; and the image at the current observation angle is collected by the mobile visual sensor.
[0066] In order to verify the method of the present invention, the pose of the mobile visual sensor is estimated. Figure 4 This is a set of target observation angle estimation results of the present invention. The camera is set at a known posture to capture images of the object to be measured. The posture estimation method proposed in the present invention is compared with the known posture. The translation error is less than 0.1mm and the angle error is less than 0.5°. Therefore, the method of the present invention can accurately collect data on the target image and accurately estimate the observation angle of the sensor relative to the target to be measured.
[0067] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A device for collecting multi-view images of a target object and automatically marking observation angles, characterized in that: include: A fixed vision sensor (1), an asymmetric coplanar target (2), a mobile vision sensor (3), a base coordinate plate (4) and a fixed target (6); The asymmetric coplanar target (2) is fixed above the mobile visual sensor (3) and moves simultaneously with the mobile visual sensor (3) in a three-dimensional Cartesian space relative to the object to be measured (5); the fixed target (6) is printed on the surface of the base coordinate plate (4), and the object to be measured (5) is placed on the base coordinate plate (4); A fixed visual sensor (1) is used to obtain a fixed-viewing angle image of a measured object (5); A mobile visual sensor (3) is used to obtain an image of a measured object (5) with a changing viewing angle; The asymmetric coplanar target (2) is used to obtain a posture transformation matrix between a fixed visual sensor (1) and a mobile visual sensor (3), thereby calculating a relative posture between the fixed visual sensor (1) and the mobile visual sensor (3); the asymmetric coplanar target (2) includes a plurality of target points distributed in an asymmetric coplanar structure; each target point is fixed on a corresponding target point support rod; and each target point support plate is fixed on the same support base; The fixed targets (6) are used to estimate the relative position of the mobile vision sensor (3) relative to the base coordinate plate (4); the fixed targets (6) are distributed at the four corners of the base coordinate plate (4); each fixed target uses black and white squares distributed in a nine-square grid; different combinations of black and white squares are used to distinguish different positions of the base coordinate plate (4).
2. The device for collecting multi-view images of a target object and automatically marking observation angles according to claim 1, characterized in that: The number of targets selected ranges from 4 to 6.
3. A device for collecting multi-view images of a target object and automatically marking observation angles according to claim 1 or 2, characterized in that: The fixed vision sensor uses a fixed-focus wide-angle lens with a focal length of less than 3.5 mm. The distance from the object to be measured is 10 to 15 times the outer diameter of the object to be measured, and the object to be measured is overlooked at an observation angle of 40° to 50°.
4. A method for collecting multi-view images of a target object and automatically marking observation angles, characterized in that: include: S1. Before acquiring multi-view images of the object to be measured, calibrate the coordinate transformation relationship between the base coordinate plate (4) and the fixed visual sensor (1): S101. Acquire an image of the asymmetric coplanar target (2) from the fixed visual sensor (1), and solve the relative position and posture of the asymmetric coplanar target (2) and the fixed visual sensor (1) based on the internal parameters of the fixed visual sensor (1); wherein the asymmetric coplanar target (2) is located at any position set within the visual angle range of the fixed sensor; S102. The mobile visual sensor (3) collects an image of the fixed target (6), and estimates the relative position and posture of the mobile visual sensor (3) and the base coordinate plate (4) by using the homography matrix method; S103. Solving the coordinate transformation relationship between the asymmetric coplanar target (2) and the mobile visual sensor (3) according to the least squares method; S104. Based on the relative position and posture of the asymmetric coplanar target (2) and the fixed visual sensor (1) in S101 and the coordinate transformation relationship between the asymmetric coplanar target (2) and the mobile visual sensor (3) in S103, the position and posture of the mobile visual sensor (3) relative to the fixed visual sensor (1) is obtained; based on the position and posture of the mobile visual sensor (3) relative to the base coordinate plate (4) in S102, the coordinate transformation relationship between the base coordinate plate (4) and the fixed visual sensor (1) is obtained; S2. The object to be measured (5) is placed on the base coordinate plate (4), and the mobile visual sensor (3) is placed at any position where the object to be measured (5) and the fixed target (6) can be captured, and multi-view image acquisition is performed on the object to be measured (5): S201. The position and posture of the asymmetric coplanar target (2) is acquired through the fixed vision sensor (1), and the position and posture of the fixed vision sensor (1) relative to the mobile vision sensor (3) is calculated through the coordinate transformation relationship between the asymmetric coplanar target (2) and the mobile vision sensor (3) obtained in S103. The position and posture of the mobile vision sensor (3) relative to the base coordinate plate (4) is calculated through the coordinate transformation relationship between the base coordinate plate (4) and the fixed vision sensor (1) obtained in S104, and is used as the perception result of the fixed vision sensor; S202. The mobile visual sensor (3) collects an image of the fixed target (6), estimates the position of the mobile visual sensor (3) relative to the base coordinate plate (4), and uses it as the perception result of the mobile visual sensor; S203. Use Kalman filtering to fuse the perception results of the fixed visual sensor and the mobile visual sensor, use the fusion result as the estimation result of the observation posture, and collect the image of the object under observation (5) at the current observation angle through the mobile visual sensor.
5. The method for multi-view image acquisition and automatic observation angle marking of a target object according to claim 4, characterized in that: The relative posture calculation formula of the asymmetric coplanar target (2) and the fixed visual sensor (1) is: T=A -1 H Wherein, T is the pose parameter of the asymmetric coplanar target (2) on the fixed visual sensor (1), A is the internal parameter of the fixed visual sensor (1) obtained by offline calibration, H is the homography matrix, u i ,v i is the coordinate of target point i in the image captured by the fixed vision sensor (1), x i ,y i is the coordinate value of the target point in the target coordinate system, i=1,2,3,4.
6. The method for collecting multi-view images of a target object and automatically marking observation angles according to claim 5, characterized in that: The coordinate transformation relationship between the asymmetric coplanar target (2) and the mobile visual sensor (3) is calculated using the following formula: Where m and n are different positions of the mobile vision sensor (3), T m 、T n are the pose parameters of the asymmetric coplanar target (2) in poses m and n under the fixed visual sensor (1), B is the pose parameter of the mobile visual sensor (3) under the base coordinate plate (4), and X is the coordinate transformation relationship between the asymmetric coplanar target (2) and the mobile visual sensor (3).
