Moxibustion robot 4D posture track self-adaptive generation method and moxibustion robot

Abstract

The invention relates to a moxibustion robot 4D posture track self-adaptive generation method and a moxibustion robot, and the moxibustion robot 4D posture track self-adaptive generation method comprises the steps: controlling a 3D camera to collect a target depth image, and controlling an RGB camera to collect a target RGB image; generating initial three-dimensional point cloud data based on the target depth image; obtaining a preset posture track based on the target RGB image, wherein the preset posture track comprises pixel position information of the target acupuncture point; based on pixel position information of a target acupuncture point, filling and fitting the initial three-dimensional point cloud data by using a proximity search method to obtain target three-dimensional point cloud data; the included angle between the moxibustion head and the mechanical arm is set to be a first preset included angle, and the first preset included angle is 35-55 degrees; and based on the target three-dimensional point cloud data, the first preset included angle and the preset posture track, the track position of the mechanical arm is controlled. The control precision of the moxibustion robot is improved.

Description

Technical Field

[0001] This invention relates to the field of moxibustion robots, specifically to a method for adaptively generating 4D posture trajectories of moxibustion robots and a moxibustion robot. Background Technology

[0002] Currently, most moxibustion robots operate using fixed postures and pre-planned paths. However, if automatic moxibustion is required for different targets, there is a problem of mismatch with the target skin surface.

[0003] In addition, moxibustion is a high-temperature and dangerous physiotherapy method, which requires higher precision in the data controlling the robotic arm. Summary of the Invention

[0004] This invention addresses the problem of improving the control precision of the robotic arm in moxibustion robots.

[0005] In a first aspect, this application provides a method for adaptively generating 4D posture trajectory of a moxibustion robot, the method comprising:

[0006] Control the 3D camera to acquire the target depth image, and the RGB camera to acquire the target RGB image;

[0007] Generate initial 3D point cloud data based on the target depth image;

[0008] A preset posture trajectory is obtained based on the target RGB image, and the preset posture trajectory includes the pixel position information of the target acupoint;

[0009] Based on the pixel position information of the target acupoint, the initial three-dimensional point cloud data is filled and fitted using the nearest neighbor search method to obtain the target three-dimensional point cloud data;

[0010] Set the angle between the moxibustion head and the robotic arm to the first preset angle, which is 35-55 degrees.

[0011] Based on the target 3D point cloud data, the first preset angle, and the preset posture trajectory, the trajectory position of the robotic arm is controlled.

[0012] In some embodiments, the motion trajectory of the robotic arm is controlled based on target 3D point cloud data, a first preset angle, and a preset attitude trajectory, including:

[0013] The Euler angles at the end of the robotic arm are defined based on the following formula:

[0014] R x = [-158.432, 175.109]

[0015] R y = [-15.886, -14.373]

[0016] R z =[137.290, 144.879]

[0017] Where Rx, Ry, and Rz are the Euler angles of the robotic arm's end effector in space.

[0018] In some embodiments, the motion trajectory of the robotic arm is controlled based on target 3D point cloud data, a first preset angle, and a preset attitude trajectory, including:

[0019] The relative positional relationship between the robotic arm's end effector and the user's skin is defined based on the following formula:

[0020] x F =x O -Hsin

[0021] y F =y O

[0022] z F =z O +Hsinα

[0023] Where, x O y O z O Let x be the initial trajectory position. F y F z F Let α be the target trajectory position, α be the first preset included angle, and H be the vertical length of the moxibustion head.

[0024] In some embodiments, controlling the motion trajectory of the robotic arm based on target 3D point cloud data, a first preset angle, and a preset attitude trajectory includes: defining the Euler angles at the end of the robotic arm based on the following formula:

[0025] R x = -166.77

[0026] R y =-15.13

[0027] R z =137.98

[0028] Furthermore, the first preset included angle is 45 degrees.

[0029] In some embodiments, a preset posture trajectory is obtained based on a target RGB image. The preset posture trajectory includes pixel position information of the target acupoint, including:

[0030] Identify key point data information of the human shoulder and hip;

[0031] Based on the distribution relationship between the human shoulder, hip and acupoints, the human acupoints are projected proportionally to obtain the pixel position information of the target acupoints.

[0032] Construct a mapping relationship between the target 3D point cloud data and the target RGB image;

[0033] Based on the mapping relationship between the target 3D point cloud data and the target RGB image, the trajectory position of the robotic arm is confirmed.

[0034] In some embodiments, the step of filling and fitting the initial 3D point cloud data using a nearest neighbor search method based on the pixel position information of the target acupoint to obtain the target 3D point cloud data includes:

[0035] Search for n neighboring pixels based on the pixel location information of the target acupoint;

[0036] Point cloud data {S} of n neighboring pixels is obtained by mapping through n neighboring pixels. i (x i ,y i ,z i |i = 1, 2, 3, ..., n};

[0037] Point cloud data based on n neighboring pixels {S i (x i ,y i ,z i Based on the statistical characteristics of |i=1,2,3,…,n}, calculate the target 3D point cloud data corresponding to the pixel position information of the target acupoint.

[0038] In some embodiments, point cloud data {S} based on n neighboring pixels i (x i ,y i ,z i The statistical characteristics of the target 3D point cloud data, including the pixel position information of the target acupoint, are used to calculate the pixel position information of the target acupoint based on the pixel position information of the target acupoint.

[0039] The target acupoint pixel location information corresponding to the target 3D point cloud data T(x) is calculated based on the following formula. t ,y t ,z t ):

[0040] z avgi =[max(z i )+max(z i )]

[0041] if(z avgi ≥z i )→if((val numi+1)>set_numi)→cal T

[0042]

[0043] Where, val numi The effective number of points in the point cloud data of n neighboring pixels.

[0044] In some embodiments, the method further includes:

[0045] Using the current position information of the robotic arm, search for n neighboring pixels;

[0046] Point cloud data of n neighboring pixels is obtained by mapping through n neighboring pixels;

[0047] Based on the statistical characteristics of the point cloud data of n neighboring pixels, the initial three-dimensional point cloud data is filled to obtain the target three-dimensional point cloud data.

[0048] Secondly, this application provides a moxibustion robot, which includes:

[0049] The device includes a 3D camera, an RGB camera, a robotic arm, a moxibustion head located at the end of the robotic arm, and one or more processors. The one or more processors are respectively connected to the 3D camera, the RGB camera, and the robotic arm. The one or more processors are configured to execute a 4D posture trajectory adaptive generation program for the moxibustion robot to achieve the method described above.

[0050] Thirdly, this application provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores: a 4D posture trajectory adaptive generation program for a moxibustion robot;

[0051] When the 4D posture trajectory adaptive generation program of the moxibustion robot is executed by one or more processors, it implements the steps of the method described above.

[0052] This application obtains the location of the target acupoint based on the target RGB image, and then uses a 3D camera to output three-dimensional point cloud data to control a robotic arm to move to the target acupoint for moxibustion, thus achieving target acupoint matching. By filling the initial three-dimensional point cloud data with the pixel position information of the target acupoint, the accuracy of robotic arm control is improved. Furthermore, the preset angle between the moxibustion head and the robotic arm simplifies the control algorithm of the robotic arm and improves safety. Attached Figure Description

[0053] Figure 1 This is a flowchart of an embodiment of the 4D posture trajectory adaptive generation method for moxibustion robot of this application;

[0054] Figure 2 This is a schematic diagram of the end effector of the robotic arm in one embodiment of the moxibustion robot of this application;

[0055] Figure 3 This is a schematic diagram of the nearest neighbor search method in one embodiment of the 4D posture trajectory adaptive generation method for the moxibustion robot of this application.

[0056] The attached diagram lists the components represented by each number as follows:

[0057] 10 filling points; 20 adjacent faces; 30 adjacent mesh faces; 40 moxibustion heads. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0060] In the description of this invention, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed herein.

[0061] The main purpose of this application is to control the end of the robotic arm of a six-axis moxibustion robot to suspend on the target surface and move according to the location of the acupoints to perform moxibustion on each acupoint.

[0062] Example 1

[0063] The design process of the 4D posture trajectory adaptive generation method for the moxibustion robot in this application includes: acquisition of 3D point cloud data and acupoint recognition, optimization and coordinate transformation of 3D point cloud data, fitting and filling of 3D point cloud data, and posture setting and safety maintenance of the robotic arm. The above four steps are described below in this embodiment.

[0064] Acquisition of 3D point cloud data and acupoint identification

[0065] In this embodiment, 3D point cloud data is acquired by using a 3D camera to capture depth images and then converting the depth images into 3D point cloud data; an RGB camera is used to capture skin images (e.g., the back) of the target user, and the pixel positions of the target acupoints are identified based on the skin images.

[0066] Furthermore, based on the relative positional information of the 3D camera and the RGB camera, after performing hand-eye calibration, tool calibration, dual-target calibration, and optimized calibration, the pixel positions of the 3D point cloud data and the RGB image can be mutually mapped. In this way, acupoint recognition can be performed using the RGB image to obtain the target acupoint for moxibustion. Then, based on the mapping relationship, the 3D point cloud information corresponding to the trajectory position of the robotic arm can be confirmed, thereby controlling the robotic arm to move to the target acupoint for moxibustion. For example, using Mediapipe to identify key point data information of the human shoulder and hip, and according to the distribution relationship between the human shoulder, hip, and acupoints, proportional projection of the acupoints can obtain the pixel information of the target acupoint.

[0067] Optimization and coordinate transformation of 3D point cloud data

[0068] Unlike point cloud data directly acquired by radar, the 3D point cloud data obtained by converting depth images acquired by 3D cameras contains non-target areas and noise interference, thus requiring point cloud segmentation, smoothing, and noise reduction. In this embodiment, point cloud segmentation uses a direct-pass filtering method, setting a value range based on the extreme values ​​and structural range of a certain dimension to generate new point cloud data; point cloud smoothing and noise reduction uses Gaussian filtering according to target requirements, and statistically processes nearby point clouds to determine whether they are noise. Based on the KD_Tree neighborhood search, N neighborhood points are searched, and the average Euclidean distance from the sampling point to the neighborhood points is calculated. Specifically: the 3D point cloud coordinates p = {p i (x i ,y i ,z i Let p be the sample neighborhood search point set p | i = 1, 2, 3, ..., n}. ij ={p i (x ij ,y ij ,z ij The average Euclidean distance d from the sampling point to the neighborhood point is given by |j=1,2,3,…,k}.i It can be represented as:

[0069]

[0070] initial point cloud p i The average distance d between it and its K neighborhood points i mean The standard deviation σ can be expressed as:

[0071]

[0072] In the 3D point cloud acquired by the 3D camera, the distances of all points should form a Gaussian distribution. By setting its mean and variance, and replacing the original 3D point cloud data with the calculated Gaussian-smoothed 3D point cloud data, the effects of point cloud noise reduction and data smoothing can be achieved.

[0073] Fitting and filling 3D point cloud data

[0074] In this application, to provide greater operational control, a fixed attitude trajectory is adopted, requiring only the 3D point cloud information of the target acupoint. Based on this, to ensure data security, accuracy, and output quality, the required 3D point cloud needs to undergo multiple fitting and filling processes based on the location of the target acupoint.

[0075] To address the issue of invalid data in 3D point cloud data, this application employs pixel search for filling. Specifically, the filling method mainly involves two proximity search methods—voxels and pixels. Since it is necessary to supplement invalid points in the 3D point cloud, pixel search is required. Using the pixel position information of the target acupoint, n neighboring pixels are searched, and the corresponding 3D information {S} is mapped from these n neighboring pixels. i (x i ,y i ,z i )|i=1,2,3,…,n}. Analyze the extreme values, significant values, and mean values ​​of neighboring 3D point cloud data to perform target point cloud data T(x t ,y t ,z t Fitting and filling.

[0076] z avgi =[max(z i )+max(z i )]

[0077] if(z avgi ≥z i )→if((val numi +1)>set_numi)→cal T

[0078]

[0079] Where, val numi The effective number of points in the point cloud data of n neighboring pixels.

[0080] In some embodiments, after the initial filling, most of the 3D point cloud data can be filled, increasing the possibility of data output. However, the end effector of moxibustion robots is generally large in size. In order to ensure the safety and accuracy of 4D posture data, it is necessary to perform end effector fitting based on the target skin surface data corresponding to the current position of moxibustion.

[0081] Similarly, for neighbor-to-neighbor sampling, a mesh search method was used to collect m neighboring pixels in order to synchronously fit the area. The three-dimensional mapping relationship is {W}. j (x j ,y j ,z j )|j=1,2,3,…,m}. Similarly, perform a second-order fitting and filling, and output the target 4D data information O(x) based on the current point location information. o ,y o ,z o )

[0082] z avgj =[max(z j )+max(z j )]

[0083] if(z avgj ≥z j )→if((val numj +1)>set_numj)

[0084] if (isnan(T(x) t ,y t ,z t ))→cal O2 :cal O1

[0085] cal O1 :x O =x t

[0086] cal O1 :y O =y t

[0087]

[0088] cal O1 &cal O2 :z O=min(val) numi (z l ))

[0089] In the above formula, the isnan() function is used to check whether its parameter is a non-numeric value.

[0090] Posture setting and safety maintenance of robotic arms

[0091] First, compared to a rotatable moxibustion head design, this application sets the angle between the moxibustion head and the skin surface, ensuring that a preset angle is maintained during moxibustion movement. For example, the angle can be set to 35-55 degrees, and in some embodiments, it can be set to 45 degrees. Furthermore, this application can define the Euler angle of the robotic arm's end effector as follows:

[0092] R x = [-158.432, 175.109]

[0093] R y = [-15.886, -14.373]

[0094] R z =[137.290, 144.879]

[0095] This can prevent problems such as exceeding limits, collisions, and singularities that may occur with six-axis robots during operation.

[0096] Furthermore, to improve the safety of the moxibustion robot, the control data of the robotic arm is limited based on the angle between the moxibustion head and the skin surface, as well as the length of the moxibustion head:

[0097] x F =x O -Hsin

[0098] y F =y O

[0099] z F =z O +Hsinα

[0100] Where, x O y O z O Let x be the initial trajectory position. F y F z F Let α be the target trajectory position, α be the first preset included angle, and H be the vertical length of the moxibustion head.

[0101] Example 1

[0102] Reference Figure 1This application provides a method for adaptively generating 4D posture trajectory of a moxibustion robot, including:

[0103] S100 controls the 3D camera to acquire target depth images and the RGB camera to acquire target RGB images;

[0104] S200: Generate initial 3D point cloud data based on the target depth image;

[0105] S300. Obtain a preset posture trajectory based on the target RGB image, wherein the preset posture trajectory includes pixel position information of the target acupoint;

[0106] S400. Based on the pixel position information of the target acupoint, the initial three-dimensional point cloud data is filled and fitted using the nearest neighbor search method to obtain the target three-dimensional point cloud data.

[0107] S500, Set the angle between the moxibustion head and the robotic arm to a first preset angle, the first preset angle being 35-55 degrees; for example, the first preset angle is 45 degrees.

[0108] S600 controls the trajectory position of the robotic arm based on the target 3D point cloud data, the first preset angle, and the preset attitude trajectory.

[0109] This application obtains the location of the target acupoint based on the target RGB image, and then uses a 3D camera to output three-dimensional point cloud data to control a robotic arm to move to the target acupoint for moxibustion, thus achieving target acupoint matching. By filling the initial three-dimensional point cloud data with the pixel position information of the target acupoint, the accuracy of robotic arm control is improved. Furthermore, the preset angle between the moxibustion head and the robotic arm simplifies the control algorithm of the robotic arm and improves safety.

[0110] In some embodiments, the motion trajectory of the robotic arm is controlled based on target 3D point cloud data, a first preset angle, and a preset attitude trajectory, including:

[0111] The Euler angles at the end of the robotic arm are defined based on the following formula:

[0112] R x = [-158.432, 175.109]

[0113] R y = [-15.886, -14.373]

[0114] R z =[137.290, 144.879]

[0115] Where Rx, Ry, and Rz are the Euler angles of the robotic arm's end effector in space. For example:

[0116] Rx = -166.77

[0117] R y =-15.13

[0118] R z =137.98

[0119] Reference Figure 2 In some embodiments, the movement trajectory of the robotic arm is controlled based on the target 3D point cloud data, a first preset angle, and a preset posture trajectory, including:

[0120] The relative positional relationship between the robotic arm's end effector and the user's skin is defined based on the following formula:

[0121] x F =x O -Hsin

[0122] y F =y O

[0123] z F =z O +Hsinα

[0124] Where, x O y O z O Let x be the initial trajectory position. F y F z F Let α be the target trajectory position, α be the first preset included angle, and H be the vertical length of the moxibustion head.

[0125] Reference Figure 3 The nearest neighbor search method is used to obtain n neighboring pixels. Specifically, based on the pixel position information of the target acupoint, the nearest neighbor search method is used to fill and fit the initial three-dimensional point cloud data to obtain the target three-dimensional point cloud data, including:

[0126] Search for n neighboring pixels based on the pixel location information of the target acupoint;

[0127] Point cloud data {S} of n neighboring pixels is obtained by mapping through n neighboring pixels. i (x i ,y i ,z i |i = 1, 2, 3, ..., n};

[0128] Point cloud data based on n neighboring pixels {S i (x i ,y i ,z iBased on the statistical characteristics of |i=1,2,3,…,n}, calculate the target 3D point cloud data corresponding to the pixel position information of the target acupoint.

[0129] In some embodiments, point cloud data {S} based on n neighboring pixels i (x i ,y i ,z i The statistical characteristics of the target 3D point cloud data, including the pixel position information of the target acupoint, are used to calculate the pixel position information of the target acupoint based on the pixel position information of the target acupoint.

[0130] The target acupoint pixel location information corresponding to the target 3D point cloud data T(x) is calculated based on the following formula. t ,y t ,z t ):

[0131] z avgi =[max(z i )+max(z i )]

[0132] if(z avgi ≥z i )→if((val numi +1)>set_numi)→cal T

[0133]

[0134]

[0135] Where, val numi The effective number of points in the point cloud data of n neighboring pixels.

[0136] Continue to refer to Figure 3 In some embodiments, the method further includes:

[0137] Using the current position information of the robotic arm, search for n neighboring pixels;

[0138] Point cloud data of n neighboring pixels is obtained by mapping through n neighboring pixels;

[0139] Based on the statistical characteristics of the point cloud data of n neighboring pixels, the initial three-dimensional point cloud data is filled to obtain the target three-dimensional point cloud data.

[0140] Similarly, for neighbor-to-neighbor sampling, a mesh search method was used to collect m neighboring pixels in order to synchronously fit the area. The three-dimensional mapping relationship is {W}. j (x j ,y j ,z j)|j=1,2,3,…,m}. Similarly, perform a second-order fitting and filling, and output the target 4D data information O(x) based on the current point location information. o ,y o ,z o )

[0141] z avgj =[max(z j )+max(z j 0]

[0142] if(z avgj ≥z j )→if((val numj +1)>set_numj)

[0143] if (isnan(T(x) t ,y t ,z t ))→cal O2 :cal O1

[0144] cal O1 :x O =x t

[0145] cal O1 :y O =y t

[0146]

[0147] cal O1 &cal O2 :z O =min(val) numi (z l ))

[0148] In the above formula, the isnan() function is used to check whether its parameter is a non-numeric value.

[0149] In this embodiment, searching for n neighboring pixels uses a neighboring mesh surface to address the issue of a large moxibustion head. The neighboring mesh surface is the plane containing the highest point when the current position of the moxibustion head is projected onto the skin surface, and its center point is the center of the moxibustion head projection. During the first filling operation, a neighboring surface search method is used. A neighboring surface refers to a surface within a preset radius of the filling point, i.e., the center point is the target filling point.

[0150] In some embodiments, controlling the motion trajectory of the robotic arm based on target 3D point cloud data, a first preset angle, and a preset attitude trajectory includes: defining the Euler angles at the end of the robotic arm based on the following formula:

[0151] Furthermore, in some embodiments, a preset posture trajectory is obtained based on the target RGB image, the preset posture trajectory including pixel position information of the target acupoint, including:

[0152] Identify key point data information of the human shoulder and hip;

[0153] Based on the distribution relationship between the human shoulder, hip and acupoints, the human acupoints are projected proportionally to obtain the pixel position information of the target acupoints.

[0154] Construct a mapping relationship between the target 3D point cloud data and the target RGB image;

[0155] Based on the mapping relationship between the target 3D point cloud data and the target RGB image, the trajectory position of the robotic arm is confirmed.

[0156] Example 2

[0157] This application provides a method for adaptively generating 4D posture trajectory of a moxibustion robot, the method comprising:

[0158] Control the 3D camera to acquire the target depth image, and the RGB camera to acquire the target RGB image;

[0159] Generate initial 3D point cloud data based on the target depth image;

[0160] A preset posture trajectory is obtained based on the target RGB image, and the preset posture trajectory includes the pixel position information of the target acupoint;

[0161] Based on the pixel position information of the target acupoint, the initial three-dimensional point cloud data is filled and fitted using the nearest neighbor search method to obtain the target three-dimensional point cloud data;

[0162] Set the angle between the moxibustion head and the robotic arm to the first preset angle, which is 35-55 degrees.

[0163] Based on the target 3D point cloud data, the first preset angle, and the preset posture trajectory, the trajectory position of the robotic arm is controlled.

[0164] In some embodiments, the motion trajectory of the robotic arm is controlled based on target 3D point cloud data, a first preset angle, and a preset attitude trajectory, including:

[0165] The Euler angles at the end of the robotic arm are defined based on the following formula:

[0166] R x = [-158.432, 175.109]

[0167] R y = [-15.886, -14.373]

[0168] R z =[137.290, 144.879]

[0169] Where Rx, Ry, and Rz are the Euler angles of the robotic arm's end effector in space.

[0170] In some embodiments, the motion trajectory of the robotic arm is controlled based on target 3D point cloud data, a first preset angle, and a preset attitude trajectory, including:

[0171] The relative positional relationship between the robotic arm's end effector and the user's skin is defined based on the following formula:

[0172] x F =x O -Hsin

[0173] y F =y O

[0174] z F =z O +Hsinα

[0175] Where, x O y O z O Let x be the initial trajectory position. F y F z F Let α be the target trajectory position, α be the first preset included angle, and H be the vertical length of the moxibustion head.

[0176] In some embodiments, controlling the motion trajectory of the robotic arm based on target 3D point cloud data, a first preset angle, and a preset attitude trajectory includes: defining the Euler angles at the end of the robotic arm based on the following formula:

[0177] R x = -166.77

[0178] R y =-15.13

[0179] R z =137.98

[0180] Furthermore, the first preset included angle is 45 degrees.

[0181] In some embodiments, a preset posture trajectory is obtained based on a target RGB image. The preset posture trajectory includes pixel position information of the target acupoint, including:

[0182] Identify key point data information of the human shoulder and hip;

[0183] Based on the distribution relationship between the human shoulder, hip and acupoints, the human acupoints are projected proportionally to obtain the pixel position information of the target acupoints.

[0184] Construct a mapping relationship between the target 3D point cloud data and the target RGB image;

[0185] Based on the mapping relationship between the target 3D point cloud data and the target RGB image, the trajectory position of the robotic arm is confirmed.

[0186] In some embodiments, the step of filling and fitting the initial 3D point cloud data using a nearest neighbor search method based on the pixel position information of the target acupoint to obtain the target 3D point cloud data includes:

[0187] Search for n neighboring pixels based on the pixel location information of the target acupoint;

[0188] Point cloud data {S} of n neighboring pixels is obtained by mapping through n neighboring pixels. i (x i ,y i ,z i |i = 1, 2, 3, ..., n};

[0189] Point cloud data based on n neighboring pixels {S i (x i ,y i ,z i Based on the statistical characteristics of |i=1,2,3,…,n}, calculate the target 3D point cloud data corresponding to the pixel position information of the target acupoint.

[0190] In some embodiments, point cloud data {S} based on n neighboring pixels i (x i ,y i ,z i The statistical characteristics of the target 3D point cloud data, including the pixel position information of the target acupoint, are used to calculate the pixel position information of the target acupoint based on the pixel position information of the target acupoint.

[0191] The target acupoint pixel location information corresponding to the target 3D point cloud data T(x) is calculated based on the following formula. t ,y t ,z t ):

[0192] z avgi =[max(z i )+max(z i )]

[0193] if(z avgi ≥z i )→if((val numi+1)>set_numi)→cal T

[0194]

[0195] Where, val numi The effective number of points in the point cloud data of n neighboring pixels.

[0196] In some embodiments, the method further includes:

[0197] Using the current position information of the robotic arm, search for n neighboring pixels;

[0198] Point cloud data of n neighboring pixels is obtained by mapping through n neighboring pixels;

[0199] Based on the statistical characteristics of the point cloud data of n neighboring pixels, the initial three-dimensional point cloud data is filled to obtain the target three-dimensional point cloud data.

[0200] Secondly, this application provides a moxibustion robot, which includes:

[0201] The device includes a 3D camera, an RGB camera, a robotic arm, a moxibustion head located at the end of the robotic arm, and one or more processors. The one or more processors are respectively connected to the 3D camera, the RGB camera, and the robotic arm. The one or more processors are configured to execute a 4D posture trajectory adaptive generation program for the moxibustion robot to achieve the method described above.

[0202] Thirdly, this application provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores: a 4D posture trajectory adaptive generation program for a moxibustion robot;

[0203] When the 4D posture trajectory adaptive generation program of the moxibustion robot is executed by one or more processors, it implements the steps of the method described above.

[0204] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0205] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0206] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0207] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0208] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0209] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0210] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for adaptively generating 4D posture trajectory of a moxibustion robot, characterized in that, include: Control the 3D camera to acquire the target depth image, and the RGB camera to acquire the target RGB image; Generate initial 3D point cloud data based on the target depth image; A preset posture trajectory is obtained based on the target RGB image, and the preset posture trajectory includes the pixel position information of the target acupoint; Based on the pixel position information of the target acupoint, the initial three-dimensional point cloud data is filled and fitted using the nearest neighbor search method to obtain the target three-dimensional point cloud data; Set the angle between the moxibustion head and the robotic arm to the first preset angle, which is 35-55 degrees. Based on the target 3D point cloud data, the first preset angle, and the preset posture trajectory, the trajectory position of the robotic arm is controlled.

2. The method for adaptive generation of 4D posture trajectory of moxibustion robot as described in claim 1, characterized in that, Based on the target 3D point cloud data, the first preset angle, and the preset attitude trajectory, the motion trajectory of the robotic arm is controlled, including: The Euler angles at the end of the robotic arm are defined based on the following formula: R x ＝[-158.432，175.109] R y ＝[-15.886，-14.373] R z ＝[137.290，144.879] Where Rx, Ry, and Rz are the Euler angles of the robotic arm's end effector in space.

3. The method for adaptive generation of 4D posture trajectory of moxibustion robot as described in claim 2, characterized in that, Based on the target 3D point cloud data, the first preset angle, and the preset attitude trajectory, the motion trajectory of the robotic arm is controlled, including: The relative positional relationship between the robotic arm's end effector and the user's skin is defined based on the following formula: x F =x O -Hsin and F / and O With F =z O +Hsinα Where, x O y O z O Let x be the initial trajectory position. F y F z F Let α be the target trajectory position, α be the first preset included angle, and H be the vertical length of the moxibustion head.

4. The method for adaptive generation of 4D posture trajectory of moxibustion robot as described in claim 1, characterized in that, Based on the target 3D point cloud data, the first preset angle, and the preset attitude trajectory, the motion trajectory of the robotic arm is controlled, including: limiting the Euler angles at the end of the robotic arm based on the following formula: R x ＝-166.77 R y ＝-15.13 R z ＝137.98 Furthermore, the first preset included angle is 45 degrees.

5. The method for adaptive generation of 4D posture trajectory of moxibustion robot as described in claim 1, characterized in that, A preset pose trajectory is obtained based on the target RGB image. The preset pose trajectory includes pixel position information of the target acupoint, including: Identify key point data information of the human shoulder and hip; Based on the distribution relationship between the human shoulder, hip and acupoints, the human acupoints are projected proportionally to obtain the pixel position information of the target acupoints. Construct a mapping relationship between the target 3D point cloud data and the target RGB image; Based on the mapping relationship between the target 3D point cloud data and the target RGB image, the trajectory position of the robotic arm is confirmed.

6. The method for adaptively generating 4D posture trajectory of a moxibustion robot as described in any one of claims 1-5, characterized in that, The target 3D point cloud data is obtained by using the nearest neighbor search method to fill and fit the initial 3D point cloud data based on the pixel position information of the target acupoint, including: Search for n neighboring pixels based on the pixel location information of the target acupoint; Point cloud data {S} of n neighboring pixels is obtained by mapping through n neighboring pixels. i (x i ,y i ,z i |i = 1, 2, 3, ..., n}; Point cloud data based on n neighboring pixels {S i (x i ,y i ,z i Based on the statistical characteristics of |i=1,2,3,…,n}, calculate the target 3D point cloud data corresponding to the pixel position information of the target acupoint.

7. The method for adaptively generating 4D posture trajectory of a moxibustion robot as described in claim 6, characterized in that, Point cloud data based on n neighboring pixels {S i (x i ,y i ,z i The statistical characteristics of the target 3D point cloud data, including the pixel position information of the target acupoint, are used to calculate the pixel position information of the target acupoint based on the pixel position information of the target acupoint. The target acupoint pixel location information corresponding to the target 3D point cloud data T(x) is calculated based on the following formula. t ,y t ,z t ): With avgi =[max(z i )+max(z i )] if(z avgi ≥z i )→if((val numi +1)>set_names)→horse T Where, val numi The effective number of points in the point cloud data of n neighboring pixels.

8. The method for adaptively generating 4D posture trajectory of a moxibustion robot as described in claim 1, characterized in that, Also includes: Using the current position information of the robotic arm, search for n neighboring pixels; Point cloud data of n neighboring pixels is obtained by mapping through n neighboring pixels; Based on the statistical characteristics of the point cloud data of n neighboring pixels, the initial three-dimensional point cloud data is filled to obtain the target three-dimensional point cloud data.

9. A moxibustion robot, characterized in that, include: The method comprises a 3D camera, an RGB camera, a robotic arm, an moxibustion head disposed at the end of the robotic arm, and one or more processors, wherein the one or more processors are respectively connected to the 3D camera, the RGB camera, and the robotic arm, and the one or more processors are configured to execute a 4D posture trajectory adaptive generation program for the moxibustion robot to achieve the method described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores: a 4D posture trajectory adaptive generation program for a moxibustion robot; When the 4D posture trajectory adaptive generation program of the moxibustion robot is executed by one or more processors, it implements the steps of the method as described in any one of claims 1-8.

Images