Data collection apparatus and method for calibration of robot vision system
The data collection device and method address the challenge of uniform data acquisition in robot vision systems by minimizing overlap areas and ensuring high-quality data across the 6-degree-of-freedom space, enhancing precision and reliability.
Patent Information
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- NEUROMEKA
- Filing Date
- 2024-12-30
- Publication Date
- 2026-07-09
AI Technical Summary
Existing computer vision-based calibration methods for robot systems face challenges in collecting unbiased image data uniformly across the entire 6-degree-of-freedom variable space, considering both kinematic reachability and camera field of view, leading to potential data bias and inaccurate calibration.
A data collection device and method that uniformly collects image data by sampling arbitrary reference points, generating mask images, and minimizing overlap areas using inverse kinematics and camera field of view constraints to ensure high-quality data acquisition.
The solution ensures consistent performance and high precision in various environments by reducing data bias and system malfunctions, maintaining reliable robot operation through unbiased data collection.
Smart Images
Figure KR2024021420_09072026_PF_FP_ABST
Abstract
Description
Data acquisition device and method for calibration of a robot vision system
[0001] The present invention relates to a data collection device and method for calibration of a robot vision system, and more specifically, to a data collection device and method for calibration of a robot vision system for acquiring image data required for calibration.
[0002] In general, robot calibration is an essential technique for ensuring the accuracy and reliability of robot systems. It defines the precise relationships between the robot's motion, sensors, and external devices, thereby minimizing errors. Robot calibration techniques include physical measurement-based calibration methods, sensor-based calibration methods, and computer vision-based calibration methods. Among these, computer vision-based calibration methods are widely used in robot calibration due to their high precision and accuracy.
[0003] In computer vision-based calibration methods, it is important to accurately calculate and estimate the intrinsic and extrinsic parameters of a camera. Intrinsic and extrinsic parameters enable the camera to correctly interpret 3D information from the real world and support the robot in performing accurate tasks.
[0004] In particular, computer vision-based calibration methods perform calibration based on image data. Therefore, it is crucial to collect high-quality image data in computer vision-based calibration methods. If image data is not collected uniformly and calibration is performed based on data skewed toward a specific range of the field of view or a particular pose of the robot, problems caused by data bias may arise.
[0005] Furthermore, in image data collection, the camera viewpoint is a problem of selecting 6 degrees of freedom variables that include both the position and angle of the camera in 3D space. Each camera viewpoint must satisfy two conditions: that the robot must be kinematically reachable and that the calibration sheet must be included within the camera's field of view. However, in general, the inverse kinematics of a robot has no analytical solution and can only be solved using numerical methodologies. Therefore, it is a difficult problem to collect unbiased data for the entire 6-degree-of-freedom variable space within the allowed range while considering both conditions.
[0006] The objective of the present invention is to provide a data collection device and method for calibrating a robot vision system, which collects data by uniformly considering all areas within the camera's field of view and various reachable robot postures during the calibration process.
[0007] A data collection device for calibration of a robot vision system according to the present invention includes a processor that performs the function of collecting data for calibration of a robot, wherein the function acquires an image of a calibration sheet placed at an arbitrary position at n reference points that serve as a reference for the operating range of the robot; calculates a calibration parameter based on the image; generates and accumulates n mask images at the n reference points by applying the calibration parameter; calculates the position of a new point in time that the robot can reach outside of the reference points; and determines a point in time direction for generating a new mask image in which the area overlapping with the accumulated mask image is minimized at the position of the new point in time.
[0008] The three-dimensional position of the new point in time is extracted through random sampling in the space between the n reference points in time, and the initial point in time direction at the new point in time is set to a direction in which the position of the calibration sheet estimated based on the data at the reference point in time is centered in the camera's field of view, and the reachability of the position of the new point in time and the initial point in time direction can be verified through an inverse kinematics algorithm.
[0009] The above n mask images may include simulation data generated by projecting an image onto the calibration sheet based on the calibration parameters.
[0010] The above n mask images may include image data based on actual shooting obtained by shooting the calibration sheet based on the above calibration parameters.
[0011] The above new mask image can be generated by simulating the capture of the calibration sheet at the above new point in time based on the above calibration parameters.
[0012] In generating the new mask image, the position of the new viewpoint is sampled to verify the kinematic attainability of the posture of looking at the calibration sheet through inverse kinematics, a rotation limit value for capturing the calibration sheet is calculated, and the rotation of the camera viewpoint is simulated within the limit value to generate a new viewpoint posture in which the accumulated mask image and the new mask image overlap minimally.
[0013] The above limit value calculates a value such that the calibration sheet does not go beyond the camera's field of view based on the boundary between the calibration sheet and the camera's field of view, and can be generated based on the minimum pitch value, maximum pitch value, minimum yaw value, and maximum yaw value relative to the direction in which the calibration sheet is viewed so that it is at the center of the camera's field of view.
[0014] The processor can generate a pose of the robot for generating a new mask image in which the area overlapping with the accumulated mask image is minimized.
[0015] The above new mask image can be generated by simulating the shooting of the calibration sheet at the new point in time based on the above calibration parameters, or obtained by actually shooting the above calibration sheet.
[0016] The processor can generate a new mask image in which the area overlapping with the accumulated mask image is minimized, thereby preventing biased data from being generated.
[0017] The above calibration parameters include internal parameters and external parameters of the camera, and the external parameters can calculate an estimate based on the position of the camera.
[0018] In generating the above new mask image, multiple mask images can be generated at multiple different new time points such that the area overlapping with the accumulated mask image is minimized.
[0019] The processor can generate the new mask image and set the result of combining the n time points and the time points for the new mask image as a new reference time point.
[0020] When the new reference point is set, the processor can calculate new calibration parameters based on an image for the new reference point, apply the new calibration parameters to generate and accumulate a mask image at the new reference point, and generate a kinematically reachable camera viewpoint such that a new mask image is generated at a new point other than the new reference point in which the area overlapping with the accumulated mask image is minimized.
[0021] Meanwhile, a method for collecting data for calibration of a robot according to the present invention acquires an image of a calibration sheet placed at an arbitrary position at n reference points that serve as a reference for the operating range of the robot; calculates calibration parameters based on the image; generates and accumulates n mask images at the n reference points by applying the calibration parameters; calculates the position of a new point in time that the robot can reach outside of the reference points; and determines a point direction for generating a new mask image in which the area overlapping with the accumulated mask images is minimized at the position of the new point in time.
[0022] The data acquisition device and method for calibration of a robot vision system according to the present invention include the effect of suppressing data bias to maintain consistent performance and high precision in various working environments and conditions, and reducing system malfunctions or inaccurate results caused by unbalanced data, thereby ensuring efficient and highly reliable robot operation.
[0023] In addition, the data acquisition device and method for calibration of a robot vision system according to the present invention include the effect of providing an algorithm that can sample a viewpoint close to the optimal one with relatively little computation by separating the sampling process of 6-dimensional viewpoint variables of two conditions, kinematic reachability and camera field of view, into position sampling and direction sampling steps, and independently applying kinematic reachability and camera field of view conditions in each step.
[0024] The technical effects of the present invention as described above are not limited to those mentioned above, and other unmentioned technical effects will be clearly understood by those skilled in the art from the description below.
[0025] FIG. 1 is a block diagram schematically illustrating a robot vision system according to the present embodiment, and
[0026] FIG. 2 is a flowchart illustrating a method of operating a data collection device according to the present embodiment, and
[0027] FIG. 3 is a conceptual diagram illustrating a method for teaching the operating range of a robot in a data collection method according to the present embodiment, and
[0028] FIG. 4 is a conceptual diagram illustrating the generation of a mask image in a data collection method according to the present embodiment, and
[0029] FIG. 5 is a conceptual diagram showing the limit values of the pitch and yaw of a camera in the data collection method according to the present embodiment, and
[0030] FIG. 6 is a conceptual diagram illustrating a method for selecting pitch and yaw values that do not overlap with existing mask images in a data collection method according to the present embodiment.
[0031] Embodiments of the present invention will be described in detail below with reference to the attached drawings. However, the embodiments disclosed below are not limited to those disclosed below and may be implemented in various forms; the embodiments provided are merely intended to ensure that the disclosure of the present invention is complete and to fully inform those skilled in the art of the scope of the invention. The shapes of elements in the drawings may be exaggerated for clearer explanation, and elements indicated by the same reference numeral in the drawings represent the same element.
[0032] FIG. 1 is a block diagram schematically illustrating a robot vision system according to the present embodiment.
[0033] As illustrated in FIG. 1, the data collection device (100) according to the present embodiment is linked to a camera (200), a robot controller (300), and a calibration device (400) and performs the function of collecting image data required for the calibration of the robot.
[0034] For example, the data collection device (100) acquires image information captured from the camera (200) and acquires information related to robot control from the robot controller (300). The data collection device (100) may provide image data for calibration to the calibration device. Here, the information related to robot control may include variable information (elevation angle and azimuth angle, etc.) that determines the coordinates or position in the robot coordinate system when the robot is located at a reference point or a specific area.
[0035] Meanwhile, the camera (200) may be mounted on the end of the robot, for example, an end effector. In this case, the camera (200) can acquire image information by capturing a marker existing within the robot's operating range. Additionally, the camera (200) may be separately fixedly positioned at a location spaced apart from the robot. In this case, the camera (200) can acquire image information by attaching a marker to the end of the robot and setting the area where the marker is visible in the camera image as the operating range.
[0036] And the calibration device (400) can acquire information from the camera (200) and the robot controller (300) and perform the function of automatically calibrating not only the intrinsic parameters of the camera (200) but also the offset value between the robot and the camera. Here, the calibration device (400) can perform calibration based on calibration parameters. The calibration parameters may include internal parameters and external parameters.
[0037] For example, the intrinsic parameters of the camera are parameters related to the intrinsic characteristics of the camera (200) and may include focal length, sensor offset, and distortion coefficient. Focal length is a parameter for determining how much the camera (200) magnifies or reduces to capture an image, and sensor offset is a parameter for determining the mechanical alignment between the image sensor of the camera (200) and the lens. And the distortion coefficient may be a parameter for correcting non-linear distortion of the lens.
[0038] And the external parameters of the camera are parameters indicating where the camera (200) and the end of the robot are located relative to each other and in which direction they are facing, and can be composed of 6 degrees of freedom (6-DOF).
[0039] Meanwhile, the data collection device (100) includes a processor (110) that performs the function of collecting high-quality image data. In order to collect high-quality image data, it is important for the robot to assume an optimal posture. Accordingly, the data collection method of the data collection device will be described in detail below with reference to the attached drawings.
[0040] FIG. 2 is a flowchart illustrating the operation method of a data collection device according to the present embodiment, and FIG. 3 is a conceptual diagram illustrating the method of teaching the operating range of a robot in the data collection method according to the present embodiment. FIG. 4 is a conceptual diagram illustrating the generation of a mask image in the data collection method according to the present embodiment, FIG. 5 is a conceptual diagram illustrating the limit values of the pitch and yaw of a camera in the data collection method according to the present embodiment, and FIG. 6 is a conceptual diagram illustrating the method of selecting pitch and yaw values that do not overlap with an existing mask image in the data collection method according to the present embodiment.
[0041] As illustrated in FIGS. 2 to 6, in the operation of the data collection device (100) according to the present embodiment, a calibration sheet is placed at a location where the robot's base is located or on a certain area (robot's work area) from said location (S210). The calibration sheet may be provided in a checkerboard pattern, but the type of calibration sheet is not limited.
[0042] Subsequently, the data collection device (100) learns n points of reference that serve as the basis for the robot's operating area according to a command from the operator. Here, the operator may designate the area that serves as the basis for performing calibration as n points of reference. Accordingly, the data collection device (100) can acquire n seed views from n reference points (S220). However, although FIG. 3 illustrates that there are 4 reference points, this is for the purpose of explaining the present embodiment, and 3 or more reference points may be applied.
[0043] Afterward, the data collection device (100) can acquire image data at n learned points in time. For example, the data collection device (100) acquires four image data by taking shots at each of four reference points through the camera (200), and acquires a robot control signal related to the shooting of image data from the robot controller (300).
[0044] Afterwards, the data collection device (100) calculates the internal parameters of the camera based on four image data and estimates the external parameters (S230).
[0045] For example, in the calculation of intrinsic parameters of a camera, parameters such as focal length, sensor offset, and distortion coefficient can be calculated. At this time, various algorithms disclosed in the prior art may be applied, and through calculation, parameter values for focal length (fx, dy), sensor offset (cx, cy), and distortion coefficient (k1, k2, p1, p2 ....) can be calculated.
[0046] And in the calculation of the camera's extrinsic parameters, the extrinsic parameters can be estimated using the position of the marker relative to the camera.
[0047] For example, in the calculation of external parameters of the camera, an offset estimate between the hand camera coordinate system and the robot end-effector coordinate system can be calculated. Here, to estimate the offset between the coordinate systems, the data acquisition device (100) may use a non-linear optimization algorithm. The data acquisition device (100) converts the position values of the marker relative to the camera (200) in various poses into the robot reference coordinate system ( Camera-robot offset ( using a method that minimizes the error of ) Estimates ). Here, indicates the position of the camera contrast marker at specific sample coordinates. And represents the index of the sample coordinates. Various algorithms, such as BroydenFletcher-GoldfarShanno, can be used for such non-linear optimization algorithms. In particular, for computational efficiency and convergence It can be calculated by setting the six variables of xyz coordinates and zyx Euler angles as a function. This can be expressed as Equation 1 below.
[0048] [Mathematical Formula 1]
[0049]
[0050] Here, It goes into the assumed value, The non-linear algorithm is executed in a way that minimizes the error of. Looking at the bottom equation of Equation 1, the sample coordinates ( ) The sample coordinates that are the previous sample coordinates( For -1), the transformation matrix of the sample coordinate system, Value, and, position of the marker from the camera ( Calculate -1) and take the inverse. Then, sample coordinates ( As the term related to ) is calculated in the forward direction, the sample coordinates( Go to -1) and then sample coordinates( In the path returning to ), only the value of the degree of misalignment between the two points remains. This is expressed by the lower equation above. Also, looking at the upper equation of Mathematical Equation 1, ∥log(ΔRj)∥ represents the norm of the rotation matrix, and r∥ΔPj∥ becomes the norm of the position error, so combined, In relation to this, a norm of the total error is formed, and by taking the sigma (Σ) value, a value is generated by summing the error values in the shooting at all sample coordinates, and it can be seen that the optimization process operates in the direction of minimizing this.
[0051] As another example, an offset estimate between a fixed camera and a base can be calculated. That is, in a work environment where a camera is installed separately, a marker is attached to the robot end, and a reference point is set in the area where the marker is visible in the camera image to calculate the robot end-marker offset ( ) Estimates. And through this, the camera position relative to the robot base ( ) can be estimated. Here, represents the robot's transformation matrix. represents the offset estimate from the robot end to the marker, and is the position of the marker from the camera. This can be expressed as Equation 2 below.
[0052] [Mathematical Formula 2]
[0053]
[0054] -1 becomes the transformation matrix from the marker to the camera, and the sample coordinates ( Previous sample coordinates of ) Camera position and sample coordinates from the robot reference coordinate system at -1) A method to minimize the errors in the rotation matrix and position matrix for the error between the camera position and the robot reference coordinate system in ), including not only the robot base reference camera position but also the robot end-marker offset ( ) can be estimated.
[0055] In this way, when the calculation of the camera's internal parameters and external parameters is completed, the data collection device (100) generates a mask image to generate accumulated image data (S240).
[0056] Referring to FIG. 4, the data collection device (100) generates mask images (v1 to v4) at four previously captured viewpoints using internal and external parameters of the camera.
[0057] Here, the mask images (v1 to v4) may be simulation data generated by projecting n images previously captured without re-capturing the actual calibration sheet. At this time, the data collection device (100) can predict and generate four improved mask images (v1 to v4) by utilizing the internal and external parameters of the camera. Then, the data collection device (100) accumulates the simulated mask images (v1 to v4) to generate accumulated image data (10).
[0058] However, the method of generating simulation mask images (v1~v4) is for the purpose of explaining the present embodiment. In other embodiments, mask images based on previously captured images may also be generated.
[0059] Afterward, the data collection device (100) calculates the location of a new point in time that the robot can reach in addition to the four reference points in time (S250). For example, the data collection device (100) can calculate the location of a plurality of new points in time (from the 5th point in time to the 14th point in time) excluding the four reference points in time that have already been acquired.
[0060] Here, the data collection device (100) calculates the positions of multiple new points other than the four reference points within the operating range of the robot through a random sampling method. Accordingly, the data collection device (100) increases the diversity of data collection and ensures that data collection is not concentrated only in specific locations.
[0061] Additionally, the data collection device (100) sets the initial direction of the new viewpoint based on data from four reference viewpoints (S260). For example, the data collection device (100) can set the initial direction of the new viewpoint in a direction that looks so that the position of the estimated calibration sheet is in the center of the camera's field of view. That is, since the estimated position value of the calibration is derived in the process described above, the camera (200) can adjust its direction so that the calibration sheet is placed exactly in the center.
[0062] Meanwhile, the reachability of the new point in time and the initial direction of the new point in time must be verified through an inverse kinematics algorithm (S270). Inverse kinematics is an algorithm in robotics that calculates the joint angles and movements required to reach a target position and direction. In other words, even if the position and direction of the new point in time are set, it is necessary to verify whether the actual robot can physically reach that position.
[0063] However, robots having a spherical wrist joint or a 3-axis joint structure at their extremities are designed to move flexibly in various directions. That is, such robots possess the characteristic of being able to move a camera attached to the extremity very easily in the desired direction. Accordingly, if the arrival of the initial direction at a new viewpoint position is verified, inverse kinematics verification can be omitted for viewpoint direction samples after a slight roll-pitch-yaw rotation has been applied.
[0064] Meanwhile, when the verification of the position of the new viewpoint and the initial direction of the new viewpoint is completed through the inverse kinematics algorithm, the data collection device (100) can predict a new mask image (v5) of the calibration sheet at the fifth viewpoint by applying the internal and external parameters of the camera (S280). Then, the data collection device (100) calculates pitch and yaw limit values based on the predicted new mask image (v5) so that the calibration sheet does not move out of the camera's field of view.
[0065] Referring to FIG. 5, the data collection device (100) can derive a pitch minimum value, a pitch maximum value, a yaw minimum value, and a yaw maximum value for capturing a new mask image (v5) using the expected new mask image (v5) and the boundary of the camera field of view. At this time, the data collection device (100) can set limit values using the pitch minimum value, pitch maximum value, yaw minimum value, and yaw maximum value relative to the direction of looking so that the calibration sheet is at the center of the camera field of view.
[0066] Afterwards, the data collection device (100) simulates the rotation of the camera viewpoint within the calculated limit value. Then, the data collection device (100) determines the viewpoint direction to generate a new mask image (v5) in which the area overlapping with the existing accumulated mask images (v1~v4) is minimized (S290).
[0067] Referring to FIG. 6, the data collection device (100) simulates various camera viewpoints within a limit value. For example, the data collection device (100) can simulate 10 camera tilting angles. The data collection device (100) determines a viewpoint for generating a new mask image (v5) in which the area overlapping with the accumulated mask images (v1 to v4) is minimized among the new mask images (v5) simulated at the 10 camera tilting angles. For example, the data collection device (100) acquires a viewpoint direction for capturing the new mask image (v5) of FIG. 6a in which the area overlapping with the accumulated mask images (v1 to v4) is minimized among the new mask images (v5) shown in FIG. 6a and FIG. 6b.
[0068] Afterward, when the new viewpoint direction is determined at the 5th viewpoint, the data collection device (100) can perform shooting to acquire a new mask image by operating the camera to the position and direction of the new viewpoint (S300). Then, the data collection device (100) can calculate the position and direction of the new viewpoint from the 6th viewpoint to the 14th viewpoint and perform shooting to acquire actual images from the 6th viewpoint to the 14th viewpoint.
[0069] And the data collection device (100) can reinforce the accumulated image data in which mask images from the first time point to the 14th time point are accumulated. At this time, the data collection device (100) sets the accumulated image data from the first time point to the 14th time point as a new reference time point. Then, the data collection device (100) calculates new calibration parameters based on the accumulated image data for the new reference time point and generates and accumulates mask images at the new reference time point. Then, it can regenerate kinematically reachable camera viewpoints so that a new mask image is generated in which the area overlapping with the mask image accumulated at a time point other than the new reference time point is minimized.
[0070] Accordingly, the data collection device and method for calibration of a robot vision system according to the present invention include the effect of suppressing data bias to maintain consistent performance and high precision in various working environments and conditions, and reducing system malfunctions or inaccurate results caused by unbalanced data, thereby ensuring efficient and highly reliable robot operation.
[0071] In addition, the data acquisition device and method for calibration of a robot vision system according to the present invention separate the process of sampling 6-dimensional viewpoint variables of two conditions, kinematic reachability and camera field of view, into position sampling and direction sampling steps, and by independently applying kinematic reachability and camera field of view conditions in each step, it provides the effect of providing an algorithm that can sample a viewpoint close to the optimal one with relatively little computation.
[0072] An embodiment of the present invention described above and illustrated in the drawings should not be interpreted as limiting the technical scope of the present invention. The scope of protection of the present invention is limited only by the matters described in the claims, and a person skilled in the art may modify or change the technical scope of the present invention in various forms. Accordingly, such modifications and changes will fall within the scope of protection of the present invention insofar as they are obvious to a person skilled in the art.
Claims
1. Includes a processor that performs the function of collecting data for robot calibration, and The above function is, Acquiring images of calibration sheets placed at arbitrary positions at n reference points that serve as the basis for the operating range of the above robot; Calculate calibration parameters based on the above image; Generate and accumulate n mask images at the n reference points by applying the above calibration parameters; Calculate the location of a new point in time that the robot can reach outside of the above reference point in time; A data acquisition device for calibration of a robot vision system, which determines a viewpoint direction to generate a new mask image in which the area overlapping with the accumulated mask image is minimized at the new viewpoint position.
2. In Paragraph 1, The three-dimensional position of the aforementioned new point in time is, It is extracted through random sampling in the space between the above n reference points, and The initial point direction at the above new point is, The position of the calibration sheet, estimated based on the data at the reference point above, is set in a viewing direction such that it is positioned in the center of the camera's field of view, and The location of the new point in time and the direction of the initial point in time are, A data acquisition device for the calibration of a robot vision system, the reachability of which is verified through an inverse kinematics algorithm.
3. In Paragraph 1, The above n mask images are, A data acquisition device for calibration of a robot vision system, comprising simulation data generated by projecting an image onto a calibration sheet based on the above calibration parameters.
4. In Paragraph 1, The above n mask images are, A data acquisition device for calibration of a robot vision system comprising image data based on actual shooting obtained by shooting the calibration sheet based on the calibration parameters.
5. In Paragraph 1, The new mask image above is, A data acquisition device for calibration of a robot vision system, generated by simulating the capture of the calibration sheet at the new point in time based on the above calibration parameters.
6. In Paragraph 1, In the generation of the new mask image above, By sampling the position of the aforementioned new viewpoint, the kinematic attainability of the posture of looking at the calibration sheet is verified through inverse kinematics, and Calculate a rotation limit value for capturing the above calibration sheet, and A data acquisition device for calibration of a robot vision system that simulates the rotation of the camera viewpoint within the above limit value to generate a new viewpoint pose in which the accumulated mask image and the new mask image overlap each other minimally.
7. In Paragraph 6, The above limit value is, Based on the boundary between the calibration sheet and the camera field of view, a value is calculated such that the calibration sheet does not go beyond the camera field of view, and A data acquisition device for calibration of a robot vision system, which is generated based on a minimum pitch value, a maximum pitch value, a minimum yaw value, and a maximum yaw value relative to the direction of looking so that the calibration sheet is positioned at the center of the camera's field of view.
8. In Paragraph 1, The above processor is, A data acquisition device for calibration of a robot vision system, which generates a pose of the robot to generate a new mask image in which the area overlapping with the accumulated mask image is minimized.
9. In Paragraph 1, The new mask image above is, A data acquisition device for calibration of a robot vision system, which is generated by simulating the shooting of the calibration sheet at the new point in time based on the above calibration parameters, or is obtained by actually performing the shooting of the above calibration sheet.
10. In Paragraph 1, The above processor is, A data acquisition device for calibration of a robot vision system that generates a new mask image with a minimized overlapping area with the accumulated mask image to prevent the generation of biased data.
11. In Paragraph 1, The above calibration parameters are, Includes internal parameters and external parameters of the above camera, The above external parameters are, A data acquisition device for calibration of a robot vision system that calculates an estimate based on the position of the above-mentioned camera.
12. In Paragraph 1, In the generation of the new mask image above, A data acquisition device for calibration of a robot vision system that generates multiple mask images in which the overlapping area with the accumulated mask image is minimized at multiple different new viewpoints.
13. In Paragraph 1, The above processor is, A data acquisition device for calibration of a robot vision system, which generates the new mask image and sets the result of combining the n viewpoints and the viewpoints for the new mask image as a new reference viewpoint.
14. In Paragraph 13, The above processor is, When the above new reference point is set, Calculate new calibration parameters based on the image for the above new reference point, and A mask image at the new reference point is generated and accumulated by applying the new calibration parameters, and A data acquisition device for calibration of a robot vision system that generates a kinematically reachable camera viewpoint such that a new mask image is generated at a new viewpoint other than the new reference viewpoint, in which the area overlapping with the accumulated mask image is minimized.
15. In a method for collecting data for robot calibration, Acquiring images of calibration sheets placed at arbitrary positions at n reference points that serve as the basis for the operating range of the above robot; Calculate calibration parameters based on the above image; Generate and accumulate n mask images at the n reference points by applying the above calibration parameters; Calculate the location of a new point in time that the robot can reach outside of the above reference point in time; A data collection method for calibration of a robot vision system, which determines a viewpoint direction to generate a new mask image in which the area overlapping with the accumulated mask image is minimized at the new viewpoint location.