WO2021023315A1 - 一种基于人眼凝视点的手眼协调抓取方法 - Google Patents

一种基于人眼凝视点的手眼协调抓取方法 Download PDF

Info

Publication number
WO2021023315A1
WO2021023315A1 PCT/CN2020/119316 CN2020119316W WO2021023315A1 WO 2021023315 A1 WO2021023315 A1 WO 2021023315A1 CN 2020119316 W CN2020119316 W CN 2020119316W WO 2021023315 A1 WO2021023315 A1 WO 2021023315A1
Authority
WO
WIPO (PCT)
Prior art keywords
eye
grasping
foreground
joint
camera
Prior art date
Application number
PCT/CN2020/119316
Other languages
English (en)
French (fr)
Inventor
熊蔡华
李全林
万中华
Original Assignee
华中科技大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华中科技大学 filed Critical 华中科技大学
Publication of WO2021023315A1 publication Critical patent/WO2021023315A1/zh

Links

Images

Classifications

    • BPERFORMING OPERATIONS; TRANSPORTING
    • B25HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
    • B25JMANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
    • B25J9/00Programme-controlled manipulators
    • B25J9/16Programme controls
    • B25J9/1656Programme controls characterised by programming, planning systems for manipulators
    • B25J9/1664Programme controls characterised by programming, planning systems for manipulators characterised by motion, path, trajectory planning
    • BPERFORMING OPERATIONS; TRANSPORTING
    • B25HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
    • B25JMANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
    • B25J9/00Programme-controlled manipulators
    • B25J9/16Programme controls
    • B25J9/1694Programme controls characterised by use of sensors other than normal servo-feedback from position, speed or acceleration sensors, perception control, multi-sensor controlled systems, sensor fusion
    • B25J9/1697Vision controlled systems

Definitions

  • the invention belongs to the technical field of hand-eye coordination, and more specifically, relates to a hand-eye coordination grasping method based on the gaze point of the human eye.
  • the gaze point of the human eye can be emitted by the person effortlessly, and can convey certain information to the outside world. For example, by observing the eyes of another person, you can know the person’s approximate gaze direction and position, combined with scene information, You can also learn about objects that the person is interested in.
  • the gaze point of the human eye can be emitted by the person effortlessly, and can convey certain information to the outside world. For example, by observing the eyes of another person, you can know the person’s approximate gaze direction and position, combined with scene information, You can also learn about objects that the person is interested in.
  • disabled people with inconvenient hands and feet they have the need to grab things, but they cannot use traditional human-computer interaction. At this time, they can use eye trackers to extract gaze points and analyze their intentions to achieve a simple and intuitive human Machine interaction.
  • Patent CN 109079777A discloses a robotic arm hand-eye coordination operation system, which estimates the position and posture of the target workpiece through machine vision, and controls the robotic arm to perform operations. The system does not use human vision control operations and requires manual settings in advance.
  • Patent CN 109048918A discloses a visual guidance method for wheelchair robotic arm robots. It uses machine vision to find objects of interest manually input by the user and the robotic arm performs the grasping task.
  • Patent US8888287B2 proposes a human-computer interaction system based on a three-dimensional gaze tracker, which uses a depth camera and an RGB camera to extract the gaze direction and gaze point for human-computer interaction.
  • the system can extract the three-dimensional gaze point of the human eye. It is used for human-computer interaction, but it is not used to control the robotic arm to grab the object of interest to the user in the actual scene.
  • the present invention provides a hand-eye coordinated grasping method based on the gaze point of the human eye, which determines the position of the object of interest for the user based on the gaze point of the human eye, and guides the machine based on the position
  • the arm realizes the grasping of objects, and has the advantages of accurate grasping and convenient operation.
  • the present invention proposes a hand-eye coordinated grasping method based on the gaze point of the human eye, which includes the following steps:
  • S1 uses an eye tracker to determine the gaze point of the human eye, and determines the position of the user's object of interest based on the gaze point of the human eye;
  • S2 Discretization traversal obtains a set of feasible grasping postures of the robot arm used to perform grasping actions at various positions in space and stores them in the server;
  • step S3 accesses the server according to the position of the user's object of interest determined in step S1 to query and obtain a set of feasible grasping postures at that position;
  • S4 obtains multiple sets of solutions for each joint angle of the manipulator based on the inverse solution of each grasping pose in the set of feasible grasping poses, and determines a set of optimal solutions from the multiple sets of solutions of joint angles, that is, obtains the corresponding joints of the manipulator
  • the optimal joint angle, the grasping posture corresponding to this group of optimal solutions is the best grasping posture
  • Each joint of the S5 robotic arm moves to the corresponding optimal joint angle, and then grasps the user's interested object in the optimal grasping posture, thereby completing the hand-eye coordinated grasping based on the human eye gaze point.
  • step S1 includes the following sub-steps:
  • S11 uses the left-eye camera and the right-eye camera on the eye tracker to respectively identify the center of the left and right pupils of the user to extract the information of the human eye;
  • S12 maps the left and right pupil centers obtained by the recognition to the foreground left camera to obtain a two-dimensional gaze point
  • S13 extracts the object anchor frame in the foreground left camera, and then determines the object of interest to the user according to the positional relationship between the two-dimensional gaze point and the object anchor frame;
  • S14 Perform three-dimensional reconstruction on the object of interest to the user to obtain the position of the object of interest in the foreground left camera;
  • S15 converts the position of the object of interest in the foreground left camera to the base coordinate system of the robot arm to determine the position of the object of interest to the user.
  • step S4 the IK fast inverse solution module is preferably used to inversely solve each joint angle of the robotic arm.
  • the target recognition and tracking algorithm is preferably used to extract the anchor frame of the object in the foreground left camera, specifically:
  • the initialized tracking algorithm performs tracking of the object. If the object is lost, the tracking algorithm is re-initialized using the result of the target recognition algorithm to obtain the anchor frame of the object by tracking.
  • a three-dimensional reconstruction algorithm is used to perform three-dimensional reconstruction of the object of interest to the user, specifically:
  • S141 calibration obtains the internal and external parameters and reprojection matrix of the foreground left camera and the foreground right camera;
  • S142 uses the internal and external parameters of the foreground left camera and the foreground right camera to correct and align the images in the foreground left camera and the foreground right camera;
  • S143 obtains the binocular disparity value of the image pixels of the foreground left camera and the foreground right camera through a feature matching algorithm
  • the reprojection matrix obtained by the binocular disparity value and the dual target is reconstructed to obtain the three-dimensional coordinates of each pixel in the image corresponding to the foreground left camera coordinate system, thereby completing the three-dimensional reconstruction of the object.
  • step S15 specifically includes the following sub-steps:
  • the multiple infrared receivers on the eye tracker are combined with the infrared generator located in the base coordinate system of the robotic arm to obtain the transformation matrix of the coordinate system of the eye tracker relative to the base coordinate system of the robotic arm;
  • S152 obtains the transformation matrix between the foreground left camera and the coordinate system of the eye tracker through calibration
  • S153 converts the pose of the object of interest in the foreground left camera to the base coordinate system of the manipulator according to the two transformation matrices.
  • a set of optimal solutions is determined from multiple sets of joint angle solutions based on maximizing the sum of the relative angle margins of all joints as the optimization criterion, preferably to minimize the absolute value of the relative angle deviation
  • the mean value of is an optimization criterion to determine a set of optimal solutions among multiple sets of solutions of joint angles.
  • each joint of the manipulator arm moves to the corresponding joint angle and then moves the pre-acquired execution error to make the manipulator arm reach the target position, and then grabs the user's interested object in the best grasping posture .
  • the execution error is obtained by adopting the following steps:
  • the pre-grabbing position is obtained according to the position of the user's object of interest and the best grasping pose deviating from the preset offset distance in the reverse direction of the grasp, and the server is accessed to query the set of feasible grasping poses for the pre-grabbing position ;
  • the present invention uses an eye tracker to determine the gaze point of the human eye, and based on the gaze point of the human eye, determines the position of the user's object of interest, and then guides the mechanical arm to realize the grasp of the user's object of interest based on the determined position of the user's object of interest It has the advantages of accurate grasping and convenient operation. It is very suitable for disabled persons with inconvenient hands and feet, and can effectively grasp objects of interest.
  • the present invention uses the kinematics inverse solution module IK fast to realize the inverse solution of the grasping posture, and can accurately obtain the target rotation angle of each joint of the mechanical arm.
  • the present invention generates a set of reachable space data by pre-discretizing the position and posture of the traversal working space, thereby improving the efficiency of the kinematics inverse solution, thereby realizing the real-time inverse solution.
  • the present invention moves each joint of the manipulator to the corresponding joint angle and then moves the execution error (that is, error compensation) obtained in advance, so that the manipulator can accurately reach the target position and realize accurate grasping of the object.
  • the method of the present invention extracts the centers of the pupils of both eyes to determine the gaze point of the user, which can effectively ensure the accuracy of gaze point recognition, thereby ensuring the accuracy of determining the pose of the object of interest; the present invention uses two foreground cameras and two The eye camera realizes the recognition and pose estimation of the user's objects of interest, and can be used in actual work scenarios.
  • the user's binocular information is used to control the robotic arm for the user to grab the object of interest, which brings convenience to the disabled.
  • the existing target recognition algorithm is generally used to extract the objects in the camera. Due to the complex background of the actual work and the unstable picture obtained due to camera shake, the recognition rate is low.
  • the present invention uses the target recognition algorithm to extract the tracking algorithm.
  • the result of the target recognition algorithm is used to reinitialize the tracking algorithm, which greatly improves the recognition success of common objects in the foreground camera. rate.
  • the present invention uses the three-dimensional reconstruction algorithm to perform three-dimensional reconstruction of the object of interest to the user, which can accurately locate the coordinates of the target object in the foreground camera coordinate system, and can obtain the pose of the target object in combination with the target object posture recognition.
  • FIG. 1 is a flowchart of a hand-eye coordinated grasping method based on the gaze point of the human eye according to an embodiment of the present invention
  • Figure 2 is a practical application diagram of a grasping device for implementing a hand-eye coordinated grasping method provided by an embodiment of the present invention
  • Figure 3 is a physical diagram of a grasping device for implementing a hand-eye coordinated grasping method provided by an embodiment of the present invention
  • Figure 4 is a schematic structural diagram of an eye tracker provided by an embodiment of the present invention.
  • FIG. 5 is a schematic diagram of object extraction according to an embodiment of the present invention.
  • Fig. 6 is a schematic diagram of the inverse solution realized by the conversion relationship between the joint angle and the homogeneous transformation matrix.
  • 1-Infrared receiver 2-Inner ring on head, 3-Outer ring subject, 4-Foreground left camera, 5-Foreground right camera, 6-Left-eye camera, 7-Right-eye camera, 8-Outer ring upper gusset , 9-moving rotating part, 10-bracket gusset, 11-manipulator, 12-frame, 13-manipulator, 14-forearm tracker, 15-first link, 16-second link, 17-lighthouse , 18-Robot base, 19-small ball, 20-bottle.
  • an embodiment of the present invention provides a hand-eye coordinated grasping method based on the gaze point of the human eye, which includes the following steps:
  • S1 uses an eye tracker to determine the gaze point of the human eye, and determines the position of the user's object of interest based on the gaze point of the human eye;
  • S2 Discretization traversal obtains a set of feasible grasping postures of the robot arm used to perform grasping actions at various positions in space and stores them in the server;
  • step S3 accesses the server according to the position of the user's object of interest determined in step S1 to query and obtain a set of feasible grasping postures at that position;
  • S4 obtains multiple sets of solutions for each joint angle of the manipulator based on the inverse solution of each grasping pose in the set of feasible grasping poses, and determines the optimal solution from the multiple sets of solutions of joint angles, that is, obtains the optimal joint corresponding to each joint of the manipulator Angle, after determining a set of optimal joint angles, the grasping pose corresponding to the set of joint angles is the best grasping pose;
  • Each joint of the S5 robotic arm moves to the corresponding optimal joint angle, and then grabs the user's interested object in the best grasping posture, thereby completing the hand-eye coordinated grasping based on the human eye gaze point.
  • the device used to implement the above method of the present invention includes a robotic arm 11 and an eye tracker.
  • the robotic arm is fixed on the frame 12 and serves as the right hand (or left hand) of the disabled and the object to be grasped (Bottle, ball, etc.) placed in front of the eye tracker.
  • the eye tracker includes an outer ring body 3 and an inner ring 2 located inside the outer ring body 3.
  • the outer ring body 3 and the inner ring 2 are connected by two moving rotating parts 9.
  • the upper part of the main body 3 is provided with an upper gusset 8 of the outer ring, and both sides of the lower part are provided with gusset gussets 10 of the outer ring.
  • the eye movement When in use, the eye movement is worn on the head of the human body.
  • the front of the human face; the foreground left camera 4 and the foreground right camera 5 are set on the front side of the outer ring main body 3 to collect the image information of the scene in front of the user.
  • the left eye camera 6 and the right eye camera 7 are set on the front end of the two brackets , Used to collect the information of the left and right images of the human eye.
  • the infrared is emitted by an infrared generator located behind the eye tracker.
  • the distance between the infrared generator and the eye tracker is preferably more than one meter.
  • the eye tracker can transmit the data of the foreground left camera 4, foreground right camera 5, left eye camera 6 and right eye camera 7 to the data processing system (computer) through the data cable, and the relevant data of the infrared receiver is sent through wireless Bluetooth To the data processing system (computer) for head pose estimation.
  • the present invention adopts the lighthouse positioning technology, and installs the positioning mark (infrared receiver) on the eye tracker to realize a high-precision head posture estimation.
  • the lighthouse positioning technology consists of a lighthouse (infrared generator) and an infrared receiver.
  • the lighthouse is set in the base coordinate system of the robotic arm.
  • the infrared receiver is installed on the eye tracker, specifically a set of infrared receiving diodes. When receiving the infrared laser, it can generate response pulses and input them into the data processing system.
  • the lighthouse has a global exposure at the beginning of each rotation cycle, and then the motor takes the linear laser transmitter across the entire space, and each cycle is performed by two orthogonal motors.
  • the infrared receiver on the eye tracker generates a longer-time pulse signal during global exposure, and a shorter-time pulse signal during subsequent laser sweeps. By processing these pulse signals, the time difference between the global exposure and the laser scanning to a certain infrared receiver can be obtained. Assuming that the scanning time difference between the horizontal and vertical rotation of the i-th receiver is ⁇ t i1 and ⁇ t i2 respectively , because the motor speed r It is known that the sweep angle of the corresponding horizontal motor and vertical motor can be further calculated:
  • s is the amplification factor
  • x i (u i, v i, 1) T
  • R and t represent rotation matrix of 3 ⁇ 3 and 3 ⁇ 1 is translation vector
  • P i (X i, Y i, Z i, 1) T
  • each point provides two linear constraints, [R
  • the foreground left camera 4 and the foreground right camera 5 are binocular cameras, which are fixed on the front of the outer ring body through the mounting hole.
  • the two foreground cameras can collect the image information of the scene in front of the user, and identify the target object through the target recognition algorithm, and then use the binocular 3D
  • the reconstruction algorithm obtains the three-dimensional coordinates of the target object.
  • the left-eye camera 6 and the right-eye camera 7 use two infrared cameras, which are fixed on the front end of the bracket gusset 10, located under the outer ring main body 3. By adjusting the two bracket gussets 10, the left-eye camera 6 and the right-eye camera 7 can be made Aim at the user’s left and right eyes.
  • infrared light source there is an infrared light source on the outside of the camera.
  • the infrared light source is close to the optical axis of the camera, a dark pupil effect will occur, that is, the color of the pupil area in the infrared camera image becomes darker, and the color of the iris and other areas become lighter. , Is conducive to the extraction of pupils.
  • step S1 includes the following sub-steps:
  • S11 uses the left-eye camera 6 and the right-eye camera 7 on the eye tracker to respectively identify the center of the left and right pupils of the user, so as to extract the information of the human eye;
  • S12 maps the left and right pupil centers obtained by recognition to the foreground left camera 4 to obtain a two-dimensional gaze point
  • S13 extracts the object anchor frame in the foreground left camera, and then determines the object of interest to the user according to the positional relationship between the two-dimensional gaze point and the object anchor frame;
  • S14 Perform three-dimensional reconstruction on the object of interest to obtain the position of the object of interest in the foreground left camera 4.
  • the existing conventional posture estimation method can also be used to estimate the posture of the object of interest to obtain the object of interest in the foreground left. Posture in camera 4;
  • S15 converts the position of the object of interest in the foreground left camera to the base coordinate system of the robot arm to determine the position of the object of interest to the user.
  • step S11 includes the following sub-steps:
  • S111 collects the images taken by the left-eye camera 6 and the right-eye camera 7 respectively and performs smoothing processing to obtain a smoothed grayscale image, wherein the smoothing processing is grayscale and filtering operations;
  • S112 Send the smoothed gray image to the edge detector (preferably into the canny edge detector) to obtain the edge points, and perform filtering processing to filter out the noise (that is, filter out the edge points that are obviously not the boundary of the pupil).
  • the edge detector preferably into the canny edge detector
  • filtering processing to filter out the noise (that is, filter out the edge points that are obviously not the boundary of the pupil).
  • Those skilled in the art can design filtering rules for the edge points corresponding to the lower pupil boundary.
  • Those skilled in the art can design filtering rules according to their needs. I will not go into details here.
  • the remaining edge points corresponding to the pupil boundary are constructed as a pupil edge point set, which is performed by the canny edge detector. Edge detection, high detection efficiency and high precision;
  • S113 performs ellipse fitting on the edge points corresponding to the remaining pupil boundary after filtering to obtain the coordinates (x l , y l ) and (x r , y r ) of the left and right pupil centers.
  • the random sampling consensus algorithm is used for the pupil boundary
  • the corresponding edge points are fitted to the ellipse to obtain the coordinates (x l , y l ) and (x r , y r ) of the center of the left and right pupils. From the parametric equation of the ellipse, there are 5 free variables, so random sampling is consistent
  • the ellipse fitting is performed by the linear algorithm, at least 5 edge points are required to fit the parameter equation of the ellipse.
  • random sampling consistency adopts an iterative method. The specific steps are as follows:
  • a ⁇ F are the coefficients to be calculated
  • S1132 calculates the support function values from all inner points to the ellipse in the pupil edge point concentration.
  • the inner points are defined as:
  • is the normalization coefficient
  • is the preset value, which can be selected according to actual needs, for example, 0.5;
  • the support function is defined as: Substituting the inner point corresponding to the ellipse into the following formula to calculate the support function value from the inner point to the ellipse:
  • a and b are the major axis and minor axis of the ellipse, Is the gray gradient of the point (x, y); according to the definition of the support function, the ratio of the short axis to the long axis The larger is, or the gray gradient direction at the inner point of the image is closer to the normal direction at the inner point of the ellipse family (ie The greater), the higher the support function value;
  • S1133 repeats steps S1131 to S1132 for a preset number of times (for example, 20 times) to fit multiple ellipses, select the ellipse corresponding to the maximum support function value, and the center of the ellipse is the pupil center.
  • a commonly used algorithm in the prior art can be used to map the identified pupil center to the foreground left camera to obtain a two-dimensional gaze point.
  • the present invention preferably uses a Gaussian process regression algorithm to calculate the coordinates of the pupil center (x l , y l ) and (x r , y r ) are mapped to the foreground camera to obtain the gaze point (x s , y s ).
  • a Gaussian process is a set of any finite random variable that has a joint Gaussian distribution. The basic principle is to construct a training set before prediction.
  • the training set includes a series of left and right pupil center coordinates and the corresponding gaze point coordinates on the foreground left camera.
  • the data in the training set is collected in advance, and the four dimensions are formed by the left and right pupil centers during prediction.
  • Vector ie test point x *
  • K(X,X) K(x * ,X)
  • K(x * ,X) K(x * ,X)
  • Its mathematical model is:
  • f is the set of gaze point coordinates on the foreground left camera in the training set
  • step S13 includes the following sub-steps:
  • S131 uses the target recognition algorithm to identify the anchor frame of the object in the foreground left camera.
  • the cylinder on the left represents the water cup
  • the circle on the right represents the ball.
  • the dotted line that includes the water cup and the ball is the correspondence between the water cup and the ball.
  • the anchor frame of the object is used to initialize the tracking target in the tracking algorithm.
  • the target recognition algorithm and the tracking algorithm are synchronized, and the target recognition algorithm and the tracking algorithm can use existing conventional methods;
  • S132 initializes the tracking algorithm to track the object. If the object is lost, the tracking algorithm is re-initialized using the real-time recognition result of the target recognition algorithm, and then the initialized tracking algorithm is used to continue tracking the object to obtain the anchor frame of the object. ,
  • the above method can effectively improve the success rate of object recognition in the foreground left camera;
  • S133 determines the object of interest to the user according to the positional relationship between the two-dimensional gaze point mapped to the left camera in the foreground and the object anchor frame, where, as shown in Figure 5, if the gaze point falls within the anchor frame of the water cup (as shown in Figure 5)
  • the black dot located in the anchor frame of the water glass thinks that the user is interested in the water glass, and if the gaze point does not fall into the anchor frame of the water glass (as the white dot located outside the anchor frame of the water glass in Figure 5), it is considered that the user is not interested in the water glass;
  • the gaze point falls into the anchor frame of the ball (as shown in the black dot located in the ball anchor frame in Figure 5)
  • the user is considered to be interested in the ball, if the gaze point does not fall within the anchor frame of the ball (as shown in Figure 5, it is located in the ball anchor frame).
  • the white dot outside the anchor box thinks that the user is not interested in the ball.
  • the three-dimensional reconstruction of the object of interest to the user specifically includes:
  • S141 obtains the internal and external parameters of the foreground left camera and the foreground right camera through dual target determination, specifically including the internal parameter matrix and external parameter matrix of the foreground left and right cameras, and derives the reprojection matrix Q through the internal and external parameter matrix;
  • S142 uses the internal and external parameters of the binocular camera to correct and align the images in the foreground left camera and the foreground right camera, which is a prior art and will not be repeated here;
  • S143 obtains the binocular disparity value d of the image pixels of the foreground left camera and the foreground right camera through the feature matching algorithm.
  • the feature matching adopts the normalized cross-correlation method, and the correlation measurement method is:
  • p(x,y) represents the coordinates of any point in the foreground left camera
  • W p represents the rectangular area centered on p
  • L(x,y) represents the gray at the point (x,y) in the foreground left camera image
  • R(x+d,y) represents the gray value at the point (x+d,y) in the foreground right camera image
  • the d that maximizes the correlation in the above formula is the binocular disparity value
  • S144 uses the binocular disparity value and the re-projection matrix Q obtained by the dual target to reconstruct the three-dimensional coordinates of each pixel in the foreground left camera image in the foreground left camera coordinate system to complete the three-dimensional reconstruction of the object.
  • the principle is:
  • [X Y Z W] are the homogeneous coordinates in the foreground left camera coordinate system
  • (x, y) are the two-dimensional coordinates in the foreground left camera image coordinate system.
  • step S15 preferably adopts the following steps to convert the position of the object of interest in the foreground left camera to the base coordinate system of the robotic arm:
  • S151 uses multiple infrared receivers on the eye tracker in combination with the infrared generator (ie lighthouse) located in the base coordinate system of the robotic arm to obtain the transformation matrix of the coordinate system of the eye tracker relative to the base coordinate system of the robotic arm, which specifically includes:
  • ⁇ i is the horizontal sweep angle of the motor that drives the infrared generator to rotate horizontally
  • ⁇ i is the vertical sweep angle of the motor that drives the infrared generator to rotate vertically
  • S1512 uses the following formula to perform direct linear transformation to obtain the transformation matrix [R
  • S152 obtains the transformation matrix between the foreground left camera and the coordinate system of the eye tracker through calibration
  • S153 converts the pose of the object of interest in the foreground left camera to the base coordinate system of the manipulator according to the two transformation matrices.
  • step S2 the discretization traversal to obtain the feasible grasping posture of the robot arm used to perform the grasping action at each position in space refers to the discretization to obtain the reachable positions of the robot arm in its working space, and then the discretization to obtain The set of feasible grasping postures corresponding to each position point.
  • the discretization method of position points is:
  • the maximum distance between the end of the robotic arm and the shoulder joint of the robotic arm is r, the volume of the sphere with this radius
  • the discretization method of grasping posture is:
  • p i is the i-th position point
  • P i represents the attitude set point solvable
  • the upper limit of p i M Solution of points may pose determined
  • the maximum value is 576.
  • there are many solvable postures in front of the manipulator base and the maximum number of solvable postures reaches 94/576, while there are fewer solvable postures at the edge of the workspace.
  • step S4 the kinematics inverse solution module is specifically used to obtain the joint angles of the robotic arm based on the inverse solution of the grasping posture.
  • the IK fast inverse solution module is used.
  • the problems solved by the inverse solution module are as follows:
  • Is the transformation matrix between the robot arm end effector (manipulator) coordinate system O 8 to the robot arm base coordinate system O 0 (that is, represents the pose of the end effector), Is the constant transformation matrix between the coordinate system O 8 of the end effector of the robot arm and the coordinate system O 7 of the second link end of the robot arm, where, i 1,2,...,7 are the linkage transformations corresponding to the seven joints of the robotic arm, as shown in Figure 6, the coordinate system O i is the coordinate system where the joint angle ⁇ i is located, and the coordinate system O 8 is not marked in Figure 6 Out, because the coordinate system O 8 is the required posture, it is a multi-solution.
  • d(se) and d(ew) are the lengths of the first link of the robot arm and the second link of the robot arm, respectively. mm and 255.5mm; the vectors n, o, a and p are the posture vector (ie grasping posture) and position vector (ie target position) of the robot arm end effector coordinate system O 8 in the robot arm base coordinate system O 0 That is, the location of the object of interest to the user).
  • c is the abbreviation of cos
  • ⁇ i is the i-th joint angle
  • s is the abbreviation of sin
  • a i-1 is the distance from Z i-1 to Z i along the X i-1 axis
  • ⁇ i-1 is i-1 about the axis of rotation X from the Z i-1 to the angle Z i is, d i along the Z i axis from the X i-1 the moving distance of X i to
  • X i is the X-axis coordinate system O i a
  • Z i It is the Z axis of the coordinate system O i , see Figure 6.
  • the mechanical arm used in the present invention contains 7 degrees of freedom and one redundant degree of freedom.
  • its kinematics inverse solution may have an infinite number of groups (that is, more than one can be solved by formula (2)).
  • Group joint angles each group of joint angles contains 7 joint angles, namely ⁇ 1 ⁇ 7 ). Therefore, a certain rule is set according to the actual situation to find the optimal set of solutions.
  • the inverse solution of robot kinematics generally follows the principle of the shortest stroke, that is, to minimize the amount of movement of each joint.
  • the general mechanical arm is a tandem mechanism. A small rotation of the shoulder joint will have a great impact on the position of the end wrist joint. , So it should be weighted, that is, move the end joints more and move the front joints less. Based on this, the present invention uses maximizing the sum of the relative angle margins of all joints as the optimization criterion to find the optimal set of joint angles:
  • the optimization objective of formula (1) is not convenient for optimization, and can be equivalent to minimizing the mean value of the absolute value of the relative angle deviation, namely:
  • aad is expressed as:
  • the standard deviation can be added to the optimization function, so the optimization objective function becomes:
  • calculating the execution error of each joint angle after step S4 includes:
  • the tracker (tracker) fixed on the forearm of the robot arm (the second link) is used to measure the position of the end of the robot arm: the tracker on the forearm of the robot arm can get the tracker coordinate system relative to the robot base coordinate system Transformation matrix
  • the tracker is set on the forearm, so the transformation matrix from the wrist joint (the end of the second link) coordinate system O 5 to the tracker coordinate system It can be obtained by one calibration, so the transformation from the wrist joint coordinate system O 5 to the manipulator base coordinate system (ie coordinate system O 0 ) can be obtained
  • the kinematics inverse solver IK fast to solve the following formula, which can be inverted to obtain the least square solution (overdetermined solution) of the first 5 joints as the actual rotation angle
  • step S4 calculate the difference between the target joint angle and the real joint angle to obtain the execution error of each joint of the manipulator, that is, use the difference between the target rotation angle and the corresponding actual rotation angle to get the execution error of the first five joints, and then use each execution error to obtain step S4 Compensation for each optimal joint angle of each optimal joint angle, that is, each optimal joint angle plus the corresponding execution error to obtain the compensated joint angle.
  • step S5 each joint of the manipulator is moved to the corresponding compensated joint angle. Ensure that the actual position reached by the robotic arm is the target position (that is, the position of the user's object of interest) to achieve high-precision grasping.
  • the invention can accurately locate the object that the user observes (interest), and control the end of the robotic arm to reach the position of the object in a anthropomorphic manner.
  • the anthropomorphic manipulator can grab the object of interest and perform specified actions, such as drinking water and carrying a small ball, It can help patients with severe physical disabilities grasp objects of interest.

Landscapes

  • Engineering & Computer Science (AREA)
  • Robotics (AREA)
  • Mechanical Engineering (AREA)
  • User Interface Of Digital Computer (AREA)
  • Manipulator (AREA)

Abstract

一种基于人眼凝视点的手眼协调抓取方法,其包括如下步骤:S1利用眼动仪确定人眼凝视点,并基于人眼凝视点确定用户感兴趣物体(20)的位置;S2离散化遍历得到机械臂(11)在空间各位置的可行抓取姿态集合并存储在服务器中;S3根据用户感兴趣物体的位置访问服务器,以查询获得该位置的可行抓取姿态集合;S4基于可行抓取姿态集合中各抓取姿态反解得到机械臂各关节角的多组解,在关节角的多组解中确定出最优解,即获得机械臂各关节对应的最优关节角;S5机械臂各关节运动至对应的最优关节角,然后以最佳抓取姿态抓取用户感兴趣物体。基于人眼凝视点的手眼协调抓取方法具有抓取准确、操作方便的优点。

Description

一种基于人眼凝视点的手眼协调抓取方法 【技术领域】
本发明属于手眼协调技术领域,更具体地,涉及一种基于人眼凝视点的手眼协调抓取方法。
【背景技术】
一方面,人眼凝视点可以由人毫不费力的发出来,并且可以向外界传达出一定的信息,例如,通过观察他人的眼睛,能够知道该人大致的凝视方向和位置,结合场景信息,还能得知该人感兴趣的物体。另一方面,对于手脚不便的残疾人而言,其有抓取东西的需求,但无法使用传统的人机交互方式,此时可以通过眼动仪提取凝视点分析其意图来实现简单直观的人机交互。综上,设计一个操作简便并且能捕捉人眼凝视点用于控制机械臂手抓取用户感兴趣物体的系统是一个亟待解决的问题。
目前具有较多的眼动仪装置和机械臂手装置,但机械臂大多是基于机器视觉的,很少有利用到人眼视觉,更没有人将眼动仪装置提取的三维凝视点用于控制机械臂手抓取感兴趣物体(即所谓的手眼协调)。专利CN 109079777A公开了一种机械臂手眼协调作业系统,通过机器视觉的方法估计目标工件的位置和姿态,控制机械臂进行作业,该系统并未用到人眼视觉控制作业,需预先手动设置作业对象。专利CN 109048918A公开了一种轮椅机械臂机器人的视觉引导方法,通过机器视觉寻找用户手动输入的感兴趣物体并由机械臂执行抓取任务,同样的,该方法并未考虑人眼视觉,需由残疾人通过传统交互方式预先确定操作对象。专利US8888287B2提出了一种基于三维凝视点追踪器的人机交互系统,其采用了一个深度相机和一个RGB相机提取凝视方向和凝视点用于人机交互,该系统可以提取人眼三维凝视点用于人机交互,但并未用于在实际场景中控制机械臂手抓取用户感兴趣的物体。
【发明内容】
针对现有技术的以上缺陷或改进需求,本发明提供了一种基于人眼凝视点的手眼协调抓取方法,其基于人眼凝视点确定出用户感兴趣物体的位置,并基于该位置指导机械臂实现物体的抓取,具有抓取准确、操作方便的优点。
为实现上述目的,本发明提出了一种基于人眼凝视点的手眼协调抓取方法,其包括如下步骤:
S1利用眼动仪确定人眼凝视点,并基于人眼凝视点确定用户感兴趣物体的位置;
S2离散化遍历得到用于执行抓取动作的机械臂在空间各位置的可行抓取姿态集合并存储在服务器中;
S3根据步骤S1确定的用户感兴趣物体的位置访问服务器,以查询获得该位置的可行抓取姿态集合;
S4基于可行抓取姿态集合中各抓取姿态反解得到机械臂各关节角的多组解,并在关节角的多组解中确定出一组最优解,即获得机械臂各关节对应的最优关节角,该组最优解对应的抓取位姿即为最佳抓取位姿;
S5机械臂各关节运动至对应的最优关节角,然后以最佳抓取位姿抓取用户感兴趣物体,以此完成基于人眼凝视点的手眼协调抓取。
作为进一步优选的,步骤S1包括如下子步骤:
S11利用眼动仪上的左眼相机和右眼相机分别识别用户的左右瞳孔的中心以提取人眼的信息;
S12将识别获得的左右瞳孔中心映射至前景左相机中以得到二维凝视点;
S13提取前景左相机中的物体锚框,然后根据二维凝视点与物体锚框的位置关系确定用户感兴趣的物体;
S14对用户感兴趣的物体进行三维重建以得到感兴趣物体在前景左相机中的位置;
S15将感兴趣物体在前景左相机中的位置转换至机械臂基坐标系下以此确定用户感兴趣物体的位置。
作为进一步优选的,步骤S4中,优选采用IK fast反解模块反解得到机械臂 的各关节角。
作为进一步优选的,步骤S13中,优选利用目标识别与追踪算法提取前景左相机中的物体锚框,具体为:
首先,利用目标识别算法识别得到前景左相机中物体的锚框,利用物体锚框初始化追踪算法中的追踪目标,该目标识别算法和追踪算法同步进行;
然后,初始化后的追踪算法进行物体的追踪,若跟丢物体,则利用目标识别算法的结果重新初始化追踪算法,以此追踪获得物体的锚框。
作为进一步优选的,步骤S14中,利用三维重建算法对用户感兴趣的物体进行三维重建,具体为:
S141标定得到前景左相机和前景右相机的内外参数及重投影矩阵;
S142利用前景左相机和前景右相机内外参数校正并对齐前景左相机和前景右相机中的图像;
S143通过特征匹配算法得到前景左相机和前景右相机图像像素的双目视差值;
S144利用双目视差值与双目标定得到的重投影矩阵重建得到图像中各个像素点对应在前景左相机坐标系下的三维坐标,以此完成物体的三维重建。
作为进一步优选的,步骤S15具体包括如下子步骤:
S151利用眼动仪上的多个红外接收器结合位于机械臂基坐标系内的红外发生器得到眼动仪坐标系相对于机械臂基坐标系的变换矩阵;
S152通过标定获得前景左相机与眼动仪坐标系之间的变换矩阵;
S153根据上述两个变换矩阵将感兴趣物体在前景左相机中的位姿转换至机械臂基坐标系下。
作为进一步优选的,步骤S4中,以最大化所有关节的相对角度裕度的和为优化准则在关节角的多组解中确定出一组最优解,优选以最小化相对角度偏差的绝对值的均值为优化准则在关节角的多组解中确定出一组最优解。
作为进一步优选的,步骤S5中,机械臂各关节运动至对应的关节角后再运动预先获取的执行误差后以使得机械臂到达目标位置,然后以最佳抓取位姿抓 取用户感兴趣物体。
作为进一步优选的,所述执行误差采用如下步骤获取:
首先,根据用户感兴趣物体的位置及最佳抓取位姿沿抓取反方向偏离预设的偏置距离得到预抓取位置,访问服务器以查询获得该预抓取位置的可行抓取姿态集合;
然后,基于可行抓取姿态集合中各抓取姿态反解得到机械臂各关节角的多组解,在关节角的多组解中确定出最优解,即目标关节角;
接着,使机械臂各关节运动至对应的目标关节角,然后测量机械臂末端的位置并反解得到机械臂各关节的真实关节角;
最后,计算目标关节角与真实关节角之差得到机械臂各关节的执行误差。
总体而言,通过本发明所构思的以上技术方案与现有技术相比,主要具备以下的技术优点:
1.本发明通过眼动仪确定人眼凝视点,并基于人眼凝视点确定出用户感兴趣物体的位置,然后基于确定出的用户感兴趣物体的位置指导机械臂实现用户感兴趣物体的抓取,具有抓取准确、操作方便的优点,非常适用于手脚不便的残疾人,可有效抓取其感兴趣的物体。
2.本发明通过运动学反解模块IK fast实现抓取姿态的反解,可准确得到机械臂各个关节的目标转角。
3.本发明通过预先离散化遍历工作空间的位置与姿态,生成可达空间数据集合,提高了运动学反解的效率,从而可以实现实时反解。
4.本发明通过使机械臂各关节运动至对应的关节角后再运动预先获取的执行误差(即进行误差补偿),以此使得机械臂可准确到达目标位置,实现物体的精确抓取。
5.本发明的方法提取了双眼瞳孔中心以确定用户的凝视点,可有效保证凝视点识别的准确性,进而保证感兴趣物体位姿确定的准确性;本发明借助两个前景相机和两个眼睛相机实现用户感兴趣物体的识别与位姿估计,可用于实际工作场景中,利用用户双眼信息控制机械臂手为用户抓取感兴趣物体,为残疾人 士带来便利。
6.现有的一般单独使用目标识别算法提取相机中的物体,由于实际工作背景复杂且由于相机抖动其获取的画面不稳定,导致识别率低下,本发明利用目标识别算法融合追踪算法的方式提取前景左相机中的物体,其中目标识别算法识别得到物体锚框用来初始化追踪算法,追踪算法跟丢时利用目标识别算法的结果来重新初始化追踪算法,从而大大提高前景相机中常见物体的识别成功率。
7.本发明利用三维重建算法对用户感兴趣的物体进行三维重建,可以精确定位目标物体在前景相机坐标系下的坐标,结合目标物体姿态识别可以得到目标物体的位姿。
【附图说明】
图1是本发明实施例提供的基于人眼凝视点的手眼协调抓取方法的流程图;
图2是本发明实施例提供的用于执行手眼协调抓取方法的抓取装置的实际应用图;
图3是本发明实施例提供的用于执行手眼协调抓取方法的抓取装置的实物图;
图4是本发明实施例提供的眼动仪的结构示意图;
图5是本发明实施例提供的感兴趣物体提取的示意图;
图6是由关节角与齐次变换矩阵的转换关系实现反解的示意图。
在所有附图中,相同的附图标记用来表示相同的元件或结构,其中:
1-红外线接收器,2-头戴内圈,3-外圈主体,4-前景左相机,5-前景右相机,6-左眼相机,7-右眼相机,8-外圈上扣板,9-移动旋转部分,10-支架扣板,11-机械臂,12-机架,13-机械手,14-小臂tracker,15-第一连杆,16-第二连杆,17-灯塔,18-机械臂基座,19-小球,20-瓶子。
【具体实施方式】
为了使本发明的目的、技术方案及优点更加清楚明白,以下结合附图及实施例,对本发明进行进一步详细说明。应当理解,此处所描述的具体实施例仅仅用以解释本发明,并不用于限定本发明。此外,下面所描述的本发明各个实 施方式中所涉及到的技术特征只要彼此之间未构成冲突就可以相互组合。
如图1所示,本发明实施例提供了一种基于人眼凝视点的手眼协调抓取方法,其包括如下步骤:
S1利用眼动仪确定人眼凝视点,并基于人眼凝视点确定用户感兴趣物体的位置;
S2离散化遍历得到用于执行抓取动作的机械臂在空间各位置的可行抓取姿态集合并存储在服务器中;
S3根据步骤S1确定的用户感兴趣物体的位置访问服务器,以查询获得该位置的可行抓取姿态集合;
S4基于可行抓取姿态集合中各抓取姿态反解得到机械臂各关节角的多组解,在关节角的多组解中确定出最优解,即获得机械臂各关节对应的最优关节角,确定一组最优的关节角后该组关节角对应的抓取位姿即为最佳抓取位姿;
S5机械臂各关节运动至对应的最优关节角,然后以最佳抓取姿态抓取用户感兴趣物体,以此完成基于人眼凝视点的手眼协调抓取。
如图2-3所示,本发明用于执行上述方法的装置包括机械臂11及眼动仪,机械臂固定在机架12上,充当残疾人右手(也可以左手),待抓取的物体(瓶子、球等)置于眼动仪前方。如图4所示,眼动仪包括外圈主体3和位于外圈主体3内部的头戴内圈2,外圈主体3和头戴内圈2通过两个移动旋转部分9连接,该外圈主体3上方设置有外圈上扣板8、下方两侧设置有支架扣板10,使用时,眼动议戴在人体头部,外圈上扣板8处于人体头部的后方,支架扣板处于人体面部的前方;前景左相机4和前景右相机5设于外圈主体3的前侧,用于采集用户前方场景图像信息,左眼相机6和右眼相机7设于两支架扣板的前端,用于采集人眼左右图像的信息,在外圈上扣板上均匀分布有多个红外线接收器1,用于接收外部的红外线,该红外线例如由设于眼动仪后方的红外发生器发射,红外发生器与眼动仪之间的距离优选超过一米。使用时,眼动仪可通过数据线将前景左相机4、前景右相机5、左眼相机6和右眼相机7数据传入数据处理系统(电脑),红外线接收器的相关数据通过无线蓝牙发送至数据处理系统 (电脑),以用于头部姿态估计。
进一步的,本发明采用灯塔定位技术,在眼动仪上安装定位标记(红外接收器),实现较高精度的头部姿态估计。灯塔定位技术由灯塔(红外发生器)和红外接收器构成,灯塔设置在机械臂基坐标系内,其内部有两个旋转方向正交的电机,两个电机上分别安装有线性红外激光发射器,每个旋转周期都能扫过视野内空间中的所有点。红外接收器安装在眼动仪上,具体为一组红外接收二极管,当接收到红外激光时,能产生响应脉冲,输入到数据处理系统中。灯塔在每个旋转周期的起始时刻有一个全局曝光,随后电机带着线性激光发射器扫过整个空间,每个周期中两个正交电机各进行一次。眼动仪上的红外接收器在全局曝光时产生一个较长时间的脉冲信号,在随后的激光扫射时产生一个较短时间的脉冲信号。处理这些脉冲信号,能够得到全局曝光与激光扫射到某个红外接收器的时间差,假设第i个接收器水平方向和竖直方向旋转激光的扫射时间差分别为Δt i1、Δt i2,由于电机转速r已知,可进一步算出对应水平电机和竖直电机的扫射角度:
Figure PCTCN2020119316-appb-000001
进而,每个红外接收器在灯塔虚拟平面上2D点的坐标可表示为:
Figure PCTCN2020119316-appb-000002
同时,眼动仪坐标系下各红外接收器的3D点坐标P i=(X i,Y i,Z i),可由外圈主体设计参数得到;
3D到2D点对的运动是一个PnP问题,可以利用直接线性变换求解:
sx i=[R|t]P i
其中,s为放大因子,x i=(u i,v i,1) T,为特征点的归一化平面齐次坐标表示,R和t分别表示3×3的旋转矩阵和3×1的平移向量,P i=(X i,Y i,Z i,1) T,为特征点对应空间点的齐次坐标,每组点对提供了两个线性约束,[R|t]共有12维,最少可通过6对匹配点可以求解该变换。
前景左相机4与前景右相机5为双目相机,通过安装孔固定在外圈主体前部,两个前景相机可以采集用户前方场景图像信息,并通过目标识别算法识别目标物体,接着通过双目三维重建算法得到目标物体的三维坐标。左眼相机6和右眼相机7采用的是两个红外相机,固定在支架扣板10前端,位于外圈主体3下方,通过调节两支架扣板10可使得左眼相机6和右眼相机7分别对准用户左右眼,两个眼睛相机外侧分别有一个红外光源,当红外光源靠近相机光轴时会产生暗瞳效应,即红外相机图像中瞳孔区域颜色变深,虹膜及其他区域颜色变浅,有利于瞳孔的提取。
具体的,步骤S1包括如下子步骤:
S11利用眼动仪上的左眼相机6和右眼相机7分别识别用户的左右瞳孔的中心,以此提取人眼的信息;
S12将识别获得的左右瞳孔中心映射至前景左相机4中以得到二维凝视点;
S13提取前景左相机中的物体锚框,然后根据二维凝视点与物体锚框的位置关系确定用户感兴趣的物体;
S14对用户感兴趣的物体进行三维重建以得到感兴趣物体在前景左相机4中的位置,也可采用现有常规姿态估计方法对用户感兴趣的物体进行姿态估计以获得感兴趣物体在前景左相机4中的姿态;
S15将感兴趣物体在前景左相机中的位置转换至机械臂基坐标系下以此确定用户感兴趣物体的位置。
进一步的,步骤S11包括如下子步骤:
S111分别采集左眼相机6和右眼相机7拍摄的图像并进行平滑处理以获得平滑后的灰度图,其中平滑处理为进行灰度化与滤波操作;
S112将平滑后的灰度图送入边缘检测器(优选送入canny边缘检测器)中得到边缘点,并进行过滤处理以滤除杂点(即过滤掉明显不属于瞳孔边界的边缘点)留下瞳孔边界对应的边缘点,具体如何过滤,本领域技术人员可以根据需要设计过滤规则,在此不赘述,留下的瞳孔边界对应的边缘点构建为瞳孔边缘点集,采用canny边缘检测器进行边缘检测,检测效率高,精度高;
S113对过滤剩下的瞳孔边界对应的边缘点进行椭圆拟合以得到左右瞳孔中心的坐标(x l,y l)和(x r,y r),优选的利用随机抽样一致性算法对瞳孔边界对应的边缘点进行椭圆拟合以得到左右瞳孔中心的坐标(x l,y l)和(x r,y r),由椭圆的参数方程知其自由变量有5个,因此在使用随机抽样一致性算法进行椭圆拟合时,至少需要5个边缘点才能拟合出椭圆的参数方程,为了得到最佳的拟合椭圆,随机抽样一致性采用了迭代的方式,其具体步骤如下:
S1131从瞳孔边缘点集中随机选取5个点以拟合出椭圆的平面参数方程,椭圆的平面参数方程为:
Q(x,y)=Ax 2+Bxy+Cy 2+Dx+Ey+F
式中,A~F为待求系数;
S1132计算瞳孔边缘点集中所有内点到椭圆的支撑函数值,具体的,内点定义为:
inliers={(x,y)|error(Q,x,y)<ε}
式中,
Figure PCTCN2020119316-appb-000003
其为损失函数,α为归一化系数,ε为预先设定值,可根据实际需要进行选择,例如取0.5;
支撑函数定义为,即将椭圆对应的内点代入下式中计算内点到椭圆的支撑函数值:
Figure PCTCN2020119316-appb-000004
式中,a,b分别为椭圆的长轴和短轴,
Figure PCTCN2020119316-appb-000005
为点(x,y)的灰度梯度;根据支撑函数的定义可知,短轴与长轴的比值
Figure PCTCN2020119316-appb-000006
越大,或者图像在内点处的灰度梯度方向与该椭圆族上内点处法线方向越接近(即
Figure PCTCN2020119316-appb-000007
越大),支撑函数值越高;
S1133重复步骤S1131~S1132预设次数(例如20次)以拟合出多个椭圆,选取支撑函数值最大时对应的椭圆,该椭圆的中心即为瞳孔中心。
更进一步的,可采用现有技术中常用的算法例如多项式拟合方法等将识别 的瞳孔中心映射至前景左相机中以得到二维凝视点,本发明优选利用高斯过程回归算法将瞳孔中心的坐标(x l,y l)和(x r,y r)映射到前景相机中得到凝视点(x s,y s),具体的,高斯过程是任意有限个随机变量均具有联合高斯分布的集合,其基本原理是在预测之前构建训练集,该训练集中包括一系列左右瞳孔中心坐标以及与其对应的在前景左相机上的凝视点坐标,训练集中的数据预先采集,预测时通过左右瞳孔中心构成四维向量(即测试点x *),计算K(X,X)、K(x *,X),然后代入期望值计算公式中获得期望值,其即为对应凝视点(x s,y s),具体的,其数学模型为:
Figure PCTCN2020119316-appb-000008
其中,f为训练集中前景左相机上凝视点坐标的集合,X为输入向量x的集合,此处输入向量为左右瞳孔中心坐标组成的4维向量x=(x l,y l,x r,y r),K(X,X)为训练集的对称正定协方差矩阵,K(x *,X)为实际测得的x *与训练集X之间的n×1协方差矩阵,k(x *,x *)为测试点自身的协方差,预测值的期望为:
Figure PCTCN2020119316-appb-000009
其中,
Figure PCTCN2020119316-appb-000010
为期望值,即为通过高斯过程回归得到的前景相机中凝视点的预测值(x s,y s)。
优选的,步骤S13包括如下子步骤:
S131利用目标识别算法识别得到前景左相机中物体的锚框,如图5所示,左边的圆柱体代表水杯,右边的圆代表球,将水杯和球包含在内的虚线即为水杯和球对应的锚框,利用物体锚框初始化追踪算法中的追踪目标,该目标识别算法和追踪算法同步进行,目标识别算法和追踪算法均可采用现有常规的方法;
S132初始化后的追踪算法进行物体的追踪,若跟丢物体,则利用目标识别算法实时识别的结果重新初始化追踪算法,然后利用初始化后的追踪算法继续进行物体追踪,以此追踪获得物体的锚框,通过上述方法可有效提高前景左相机中物体识别的成功率;
S133然后根据映射至前景左相机中的二维凝视点与物体锚框的位置关系确 定用户感兴趣的物体,其中,如图5所示,若凝视点落入水杯锚框内(如图5中位于水杯锚框内的黑点)认为用户对该水杯感兴趣,若凝视点未落入水杯锚框内(如图5中位于水杯锚框之外的白点)认为用户对该水杯不感兴趣;若凝视点落入球的锚框内(如图5中位于球锚框内的黑点)认为用户对该球感兴趣,若凝视点未落入球的锚框内(如图5中位于球锚框之外的白点)认为用户对该球不感兴趣。
优选的,步骤S14对用户感兴趣的物体进行三维重建具体为:
S141通过双目标定得到前景左相机和前景右相机的内外参数,具体包括前景左右相机的内参矩阵和外参矩阵,通过内外参数矩阵推导出重投影矩阵Q;
S142利用双目相机内外参数校正对齐前景左相机和前景右相机中的图像,其为现有技术,在此不赘述;
S143通过特征匹配算法得到前景左相机和前景右相机图像像素的双目视差值d,优选的,特征匹配采用归一化互相关方法,其相关性度量方式为:
Figure PCTCN2020119316-appb-000011
其中,p(x,y)表示前景左相机中任意点的坐标,W p表示以p为中心的矩形区域,L(x,y)表示前景左相机图像中点(x,y)处的灰度值,
Figure PCTCN2020119316-appb-000012
表示前景左相机图像中W p的灰度均值,R(x+d,y)表示前景右相机图像中点(x+d,y)处的灰度值,
Figure PCTCN2020119316-appb-000013
表示前景右相机图像中点(x+d,y)对应的矩形区域中的灰度均值,上式中使相关性最大的d就是所求的双目视差值;
S144利用双目视差值与双目标定得到的重投影矩阵Q重建得到前景左相机图像中各个像素点在前景左相机坐标系下的三维坐标,以此完成物体的三维重建,其原理为:
[X Y Z W] T=Q*[x y d 1] T
其中,[X Y Z W]为前景左相机坐标系下的齐次坐标,(x,y)为前景左相机图像坐标系下的二维坐标。
更具体而言,步骤S15优选采用下述步骤将感兴趣物体在前景左相机中的 位置转换至机械臂基坐标系下:
S151利用眼动仪上的多个红外接收器结合位于机械臂基坐标系内的红外发生器(即灯塔)得到眼动仪坐标系相对于机械臂基坐标系的变换矩阵,具体包括:
S1511测得每个红外接收器在红外发生器虚拟平面上的2D坐标(u i,v i):
Figure PCTCN2020119316-appb-000014
其中,α i为带动红外发生器水平旋转的电机在水平方向的扫射角度,β i为带动红外发生器竖直旋转的电机在竖直方向的扫射角度;
S1512利用下式进行直接线性变换求解获得眼动仪坐标系相对于机械臂基坐标系的变换矩阵[R|t]:
sx i=[R|t]P i
上式即是求解3D到2D点对运动的方法(PnP问题),其中,s为放大因子,x i=(u i,v i,1) T,R和t分别表示3×3的旋转矩阵和3×1的平移向量,P i=(X i,Y i,Z i,1) T,(X i,Y i,Z i)表示红外接收器在眼动仪坐标系下的3D坐标,每组点对提供了两个线性约束,[R|t]共有12维,因此最少可通过6对匹配点求解变换矩阵[R|t],故红外接收器设置6个以上;
S152通过标定获得前景左相机与眼动仪坐标系之间的变换矩阵;
S153根据上述两个变换矩阵将感兴趣物体在前景左相机中的位姿转换至机械臂基坐标系下。
步骤S2中,离散化遍历得到用于执行抓取动作的机械臂在空间各位置的可行抓取姿态指的是离散化获得机械臂在其工作空间内可达的各位置点,然后离散化获得各位置点对应的可行抓取姿态集合。
在一个优选实施例中,位置点的离散化方式为:
机械臂末端离机械臂肩关节的最大距离为r,以此为半径的球体体积
Figure PCTCN2020119316-appb-000015
三个轴的离散化采样间隔为d(其根据抓取物体的大小进行设定),于是整个机械臂工作空间内的采样点数量(即位置点数量)为N=V/d 3
抓取姿态的离散化方式为:
将姿态角以间隔为0.255弧度等分,则一点的姿态采样数(即可行抓取姿态)为576个,利用IKfast反解工具包判断各位置点对应的姿态可解性,建立机械部可达空间数据库:
Figure PCTCN2020119316-appb-000016
其中,p i为第i个位置点,i的上限是N=V/d 3
Figure PCTCN2020119316-appb-000017
表示p i点的可解姿态集合,其上限M由p i点的可解姿态数决定,最大值为576。一般的,机械臂基座前方的可解姿态较多,其最高可解姿态数达到94/576个,而工作空间边缘的可解姿态较少。
步骤S4中,具体采用运动学反解模块基于抓取姿态反解得到机械臂的各关节角,优选采用IK fast反解模块,反解模块求解的问题如下:
Figure PCTCN2020119316-appb-000018
其中,
Figure PCTCN2020119316-appb-000019
为机械臂末端执行器(机械手)坐标系O 8到机械臂基坐标系O 0之间的变换矩阵(即代表末端执行器的位姿),
Figure PCTCN2020119316-appb-000020
为机械臂末端执行器坐标系O 8到机械臂第二连杆末端坐标系O 7之间的常量变换矩阵,其中,
Figure PCTCN2020119316-appb-000021
i=1,2,…,7为机械臂七个关节对应的连杆变换,如图6所示,坐标系O i为关节角θ i所在的坐标系,坐标系O 8未在图6中标出,因为坐标系O 8是要求的姿态,是个多解,图6中d(se)、d(ew)分别是机械臂第一连杆和机械臂第二连杆的长度,长度分别为329.2mm和255.5mm;向量n,o,a和p分别为机械臂末端执行器坐标系O 8在机械臂基坐标系O 0下的姿态矢量(即抓取姿态)和位置矢量(即目标位置,也即用户感兴趣物体的位置)。通过上述公式(一)可以解出各变换矩阵即
Figure PCTCN2020119316-appb-000022
然后利用
Figure PCTCN2020119316-appb-000023
采用下式(二)反解得到机械臂各关节角:
Figure PCTCN2020119316-appb-000024
其中,c为cos的简写,θ i为第i关节角,s为sin的简写,a i-1为沿X i-1轴从Z i-1移动到Z i的距离,α i-1为绕X i-1轴从Z i-1旋转到Z i的角度,d i为沿Z i轴从X i-1移动到X i的距离,X i为坐标系O i的X轴,Z i为坐标系O i的Z轴,可参见图6。
具体的,本发明采用的机械臂包含7个自由度,有一个冗余自由度,在工作空间内,它的运动学反解可能会有无穷多组(即通过公式(二)可求解出多组关节角,每组关节角中均包含7个关节角,即θ 1~θ 7),因此,根据实际情况设定一定的规则从中找到最优的一组解。机器人运动学反解一般遵循最短行程原理,即使得每个关节的活动量最小,同时,一般机械手臂是串联机构,肩关节转动一个很小的角度都会对末端手腕关节的位置造成很大的影响,所以应该加权处理,即多移动末端关节,少移动前段关节。基于此,本发明以最大化所有关节的相对角度裕度的和为优化准则以找到最优的一组关节角:
Figure PCTCN2020119316-appb-000025
其中,θ={θ i},i=1…7,即每个关节角离极限位置越远越好,这样关节角的灵活度更高,更少受关节范围的限制。
式(1)的优化目标不便于优化,可以等效为最小化相对角度偏差的绝对值的均值,即:
Figure PCTCN2020119316-appb-000026
其中,μ=mean(aad)为相对角度偏差绝对值的均值,相对角度偏差的绝对值aad表示为:
Figure PCTCN2020119316-appb-000027
同时为了防止个别关节角出现较大偏差,可以将标准差加入到优化函数当 中,故优化目标函数变为:
Figure PCTCN2020119316-appb-000028
其中,σ=var(aad)为相对角度偏差绝对值的方差。
实际操作中,采用上述方法实现物体的抓取时,仍存在一定的误差,为了提高抓取的精度,本发明优选进行补偿处理。具体的,在步骤S4之后计算出各关节角的执行误差,包括:
首先,根据用户感兴趣物体的位置与确定的最佳抓取姿态沿抓取反方向偏离预设的偏置距离得到预抓取位置,访问服务器以查询获得该预抓取位置的可行抓取姿态集合;具体的,用户感兴趣物体的位置为c,得到预抓取位置为p=c+z R·b,其中,z R为最佳抓取姿态下机械手的主轴向量(垂直于掌心的方向),b为预设的偏置距离,优选设为10cm;
然后,基于可行抓取姿态集合中的各抓取姿态反解得到机械臂各关节角的多组解,在关节角的多组解中确定出最优解,即目标关节角;
接着,使机械臂各关节运动至对应的目标关节角,然后测量机械臂末端的位置并反解得到机械臂各关节的真实关节角,具体是获得机械臂1-5关节的真实转角,具体的利用固定在机械臂小臂(为第二连杆)上的tracker(追踪器)测量得到机械臂末端的位置:通过机械臂小臂上的tracker可以得到该tracker坐标系相对于机器人基坐标系的变换矩阵
Figure PCTCN2020119316-appb-000029
又该tracker设置在小臂上,因此腕关节(第二连杆末端)坐标系O 5到该tracker坐标系的变换矩阵
Figure PCTCN2020119316-appb-000030
可以通过一次标定获得,于是可以求得腕关节坐标系O 5到机械臂基坐标系(即坐标系O 0)的变换
Figure PCTCN2020119316-appb-000031
再利用运动学反解器IK fast求解下式可反解得到前5个关节的最小二乘解(超定解)作为实际转角
Figure PCTCN2020119316-appb-000032
Figure PCTCN2020119316-appb-000033
最后,计算目标关节角与真实关节角之差得到机械臂各关节的执行误差,即用目标转角与对应的实际转角做差得到前五个关节的执行误差,然后利用各执行误差对步骤S4获得的各最优关节角进行补偿,即将各最优关节角加上对应 的执行误差获得补偿后的关节角,最后在步骤S5中使机械臂各关节运动至对应的补偿后的关节角,以此保证机械臂实际到达的位置为目标位置(即用户感兴趣物体的位置),以实现高精度抓取。
本发明可以精确定位用户观察(感兴趣)的物体,并控制机械臂末端以拟人的方式到达物体位置,拟人机械手可以抓取感兴趣物体并执行指定的动作,比如说喝水与搬运小球,可帮助严重肢体残疾患者抓取感兴趣物体。
本领域的技术人员容易理解,以上所述仅为本发明的较佳实施例而已,并不用以限制本发明,凡在本发明的精神和原则之内所作的任何修改、等同替换和改进等,均应包含在本发明的保护范围之内。

Claims (9)

  1. 一种基于人眼凝视点的手眼协调抓取方法,其特征在于,包括如下步骤:
    S1利用眼动仪确定人眼凝视点,并基于人眼凝视点确定用户感兴趣物体的位置;
    S2离散化遍历得到用于执行抓取动作的机械臂在空间各位置的可行抓取姿态集合并存储在服务器中;
    S3根据步骤S1确定的用户感兴趣物体的位置访问服务器,以查询获得该位置的可行抓取姿态集合;
    S4基于可行抓取姿态集合中各抓取姿态反解得到机械臂各关节角的多组解,并在关节角的多组解中确定出一组最优解,即获得机械臂各关节对应的最优关节角,该组最优解对应的抓取位姿即为最佳抓取位姿;
    S5机械臂各关节运动至对应的最优关节角,然后以最佳抓取位姿抓取用户感兴趣物体,以此完成基于人眼凝视点的手眼协调抓取。
  2. 如权利要求1所述的基于人眼凝视点的手眼协调抓取方法,其特征在于,步骤S1包括如下子步骤:
    S11利用眼动仪上的左眼相机和右眼相机分别识别用户的左右瞳孔的中心以提取人眼的信息;
    S12将识别获得的左右瞳孔中心映射至前景左相机中以得到二维凝视点;
    S13提取前景左相机中的物体锚框,然后根据二维凝视点与物体锚框的位置关系确定用户感兴趣的物体;
    S14对用户感兴趣的物体进行三维重建以得到感兴趣物体在前景左相机中的位置;
    S15将感兴趣物体在前景左相机中的位置转换至机械臂基坐标系下以此确定用户感兴趣物体的位置。
  3. 如权利要求2所述的基于人眼凝视点的手眼协调抓取方法,其特征在于,步骤S13中,优选利用目标识别与追踪算法提取前景左相机中的物体锚框,具 体为:
    首先,利用目标识别算法识别得到前景左相机中物体的锚框,利用物体锚框初始化追踪算法中的追踪目标,该目标识别算法和追踪算法同步进行;
    然后,初始化后的追踪算法进行物体的追踪,若跟丢物体,则利用目标识别算法的结果重新初始化追踪算法,以此追踪获得物体的锚框。
  4. 如权利要求2所述的基于人眼凝视点的手眼协调抓取方法,其特征在于,步骤S14中,利用三维重建算法对用户感兴趣的物体进行三维重建,具体为:
    S141标定得到前景左相机和前景右相机的内外参数及重投影矩阵;
    S142利用前景左相机和前景右相机内外参数校正并对齐前景左相机和前景右相机中的图像;
    S143通过特征匹配算法得到前景左相机和前景右相机图像像素的双目视差值;
    S144利用双目视差值与双目标定得到的重投影矩阵重建得到图像中各个像素点对应在前景左相机坐标系下的三维坐标,以此完成物体的三维重建。
  5. 如权利要求2所述的基于人眼凝视点的手眼协调抓取方法,其特征在于,步骤S15具体包括如下子步骤:
    S151利用眼动仪上的多个红外接收器结合位于机械臂基坐标系内的红外发生器得到眼动仪坐标系相对于机械臂基坐标系的变换矩阵;
    S152通过标定获得前景左相机与眼动仪坐标系之间的变换矩阵;
    S153根据上述两个变换矩阵将感兴趣物体在前景左相机中的位姿转换至机械臂基坐标系下。
  6. 如权利要求1所述的基于人眼凝视点的手眼协调抓取方法,其特征在于,步骤S4中,优选采用IK fast反解模块反解得到机械臂关节角的多组解。
  7. 如权利要求1-6任一项所述的基于人眼凝视点的手眼协调抓取方法,其特征在于,步骤S4中,以最大化所有关节的相对角度裕度的和为优化准则在关节角的多组解中确定出一组最优解,优选以最小化相对角度偏差的绝对值的均值为优化准则在关节角的多组解中确定出一组最优解。
  8. 如权利要求1-7任一项所述的基于人眼凝视点的手眼协调抓取方法,其特征在于,步骤S5中,机械臂各关节运动至对应的关节角后再运动预先获取的执行误差后以使得机械臂到达目标位置,然后以最佳抓取位姿抓取用户感兴趣物体。
  9. 如权利要求7所述的基于人眼凝视点的手眼协调抓取方法,其特征在于,所述执行误差采用如下步骤获取:
    首先,根据用户感兴趣物体的位置及最佳抓取位姿沿抓取反方向偏离预设的偏置距离得到预抓取位置,访问服务器以查询获得该预抓取位置的可行抓取姿态集合;
    然后,基于可行抓取姿态集合中各抓取姿态反解得到机械臂各关节角的多组解,在关节角的多组解中确定出最优解,即目标关节角;
    接着,使机械臂各关节运动至对应的目标关节角,然后测量机械臂末端的位置并反解得到机械臂各关节的真实关节角;
    最后,计算目标关节角与真实关节角之差得到机械臂各关节的执行误差。
PCT/CN2020/119316 2019-08-06 2020-09-30 一种基于人眼凝视点的手眼协调抓取方法 WO2021023315A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910721469.7 2019-08-06
CN201910721469.7A CN110605714B (zh) 2019-08-06 2019-08-06 一种基于人眼凝视点的手眼协调抓取方法

Publications (1)

Publication Number Publication Date
WO2021023315A1 true WO2021023315A1 (zh) 2021-02-11

Family

ID=68890465

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/119316 WO2021023315A1 (zh) 2019-08-06 2020-09-30 一种基于人眼凝视点的手眼协调抓取方法

Country Status (2)

Country Link
CN (1) CN110605714B (zh)
WO (1) WO2021023315A1 (zh)

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112907586A (zh) * 2021-03-30 2021-06-04 贵州大学 基于视觉的机械臂控制方法、装置、系统和计算机设备
CN113146633A (zh) * 2021-04-23 2021-07-23 无锡信捷电气股份有限公司 一种基于自动贴盒系统的高精度手眼标定方法
CN113319863A (zh) * 2021-05-11 2021-08-31 华中科技大学 机器人磨抛加工整体叶盘的工件装夹位姿优化方法和系统
CN113610921A (zh) * 2021-08-06 2021-11-05 沈阳风驰软件股份有限公司 混合工件抓取方法、设备及计算机可读存储介质
CN113673479A (zh) * 2021-09-03 2021-11-19 济南大学 基于视觉关注点识别物体的方法
CN113822946A (zh) * 2021-10-09 2021-12-21 上海第二工业大学 一种基于计算机视觉的机械臂抓取方法
CN114067309A (zh) * 2021-10-25 2022-02-18 大连理工大学 一种基于多视角三维重建的甜椒识别与采摘顺序确定方法
CN114147714A (zh) * 2021-12-02 2022-03-08 浙江机电职业技术学院 自主机器人机械臂控制参数计算方法和系统
CN114734444A (zh) * 2022-04-27 2022-07-12 博众精工科技股份有限公司 一种目标定位方法、装置、电子设备及存储介质
CN114998573A (zh) * 2022-04-22 2022-09-02 北京航空航天大学 一种基于rgb-d特征深度融合的抓取位姿检测方法
CN115990889A (zh) * 2023-03-23 2023-04-21 武汉益模科技股份有限公司 一种基于多感受野的复合定位方法、装置及计算机系统
CN116968037A (zh) * 2023-09-21 2023-10-31 杭州芯控智能科技有限公司 一种多机械臂协同任务调度方法
CN117251058A (zh) * 2023-11-14 2023-12-19 中国海洋大学 一种多信息体感交互系统的控制方法

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110605714B (zh) * 2019-08-06 2021-08-03 华中科技大学 一种基于人眼凝视点的手眼协调抓取方法
CN113043334B (zh) * 2021-02-23 2022-12-06 上海埃奇机器人技术有限公司 一种基于机器人的光伏电池串定位方法
CN113128417B (zh) * 2021-04-23 2023-04-07 南开大学 一种基于头部姿态的双区域眼动跟踪方法
CN115256367A (zh) * 2021-04-29 2022-11-01 中国科学院沈阳自动化研究所 一种基于双目立体成像的机械臂手眼标定方法

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120165979A1 (en) * 2010-12-28 2012-06-28 Samsung Electronics Co., Ltd. Robot and control method thereof
US20120173019A1 (en) * 2010-12-29 2012-07-05 Samsung Electronics Co., Ltd. Robot and control method thereof
CN103271784A (zh) * 2013-06-06 2013-09-04 山东科技大学 基于双目视觉的人机交互式机械手控制系统和控制方法
CN109877827A (zh) * 2018-12-19 2019-06-14 东北大学 一种连杆机械手的非定点物料视觉识别与抓取装置及方法
CN110032278A (zh) * 2019-03-29 2019-07-19 华中科技大学 一种人眼感兴趣物体的位姿识别方法、装置及系统
CN110605714A (zh) * 2019-08-06 2019-12-24 华中科技大学 一种基于人眼凝视点的手眼协调抓取方法

Family Cites Families (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9607347B1 (en) * 2015-09-04 2017-03-28 Qiang Li Systems and methods of 3D scanning and robotic application of cosmetics to human
CN105537824B (zh) * 2016-01-27 2017-08-25 华南理工大学 一种基于机械臂手眼协调自主焊接控制方法
CN106553195B (zh) * 2016-11-25 2018-11-27 中国科学技术大学 工业机器人抓取过程中的物体6自由度定位方法及系统
CN107471218B (zh) * 2017-09-07 2020-09-11 南京理工大学 一种基于多目视觉的双臂机器人手眼协调方法
CN108972494B (zh) * 2018-06-22 2020-09-22 华南理工大学 一种仿人机械手抓取控制系统及其数据处理方法
CN109079777B (zh) * 2018-08-01 2021-09-10 北京科技大学 一种机械臂手眼协调作业系统
CN109333536A (zh) * 2018-10-26 2019-02-15 北京因时机器人科技有限公司 一种机器人及其物体抓取方法和装置
CN109664317B (zh) * 2019-01-24 2022-06-14 深圳勇艺达机器人有限公司 机器人的物体抓取系统及方法
CN109986560B (zh) * 2019-03-19 2023-02-14 埃夫特智能装备股份有限公司 一种面向多目标种类的机械臂自适应抓取方法

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120165979A1 (en) * 2010-12-28 2012-06-28 Samsung Electronics Co., Ltd. Robot and control method thereof
US20120173019A1 (en) * 2010-12-29 2012-07-05 Samsung Electronics Co., Ltd. Robot and control method thereof
CN103271784A (zh) * 2013-06-06 2013-09-04 山东科技大学 基于双目视觉的人机交互式机械手控制系统和控制方法
CN109877827A (zh) * 2018-12-19 2019-06-14 东北大学 一种连杆机械手的非定点物料视觉识别与抓取装置及方法
CN110032278A (zh) * 2019-03-29 2019-07-19 华中科技大学 一种人眼感兴趣物体的位姿识别方法、装置及系统
CN110605714A (zh) * 2019-08-06 2019-12-24 华中科技大学 一种基于人眼凝视点的手眼协调抓取方法

Cited By (23)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112907586A (zh) * 2021-03-30 2021-06-04 贵州大学 基于视觉的机械臂控制方法、装置、系统和计算机设备
CN112907586B (zh) * 2021-03-30 2024-02-02 贵州大学 基于视觉的机械臂控制方法、装置、系统和计算机设备
CN113146633A (zh) * 2021-04-23 2021-07-23 无锡信捷电气股份有限公司 一种基于自动贴盒系统的高精度手眼标定方法
CN113146633B (zh) * 2021-04-23 2023-12-19 无锡信捷电气股份有限公司 一种基于自动贴盒系统的高精度手眼标定方法
CN113319863B (zh) * 2021-05-11 2023-06-16 华中科技大学 机器人磨抛加工整体叶盘的工件装夹位姿优化方法和系统
CN113319863A (zh) * 2021-05-11 2021-08-31 华中科技大学 机器人磨抛加工整体叶盘的工件装夹位姿优化方法和系统
CN113610921A (zh) * 2021-08-06 2021-11-05 沈阳风驰软件股份有限公司 混合工件抓取方法、设备及计算机可读存储介质
CN113610921B (zh) * 2021-08-06 2023-12-15 沈阳风驰软件股份有限公司 混合工件抓取方法、设备及计算机可读存储介质
CN113673479A (zh) * 2021-09-03 2021-11-19 济南大学 基于视觉关注点识别物体的方法
CN113822946A (zh) * 2021-10-09 2021-12-21 上海第二工业大学 一种基于计算机视觉的机械臂抓取方法
CN113822946B (zh) * 2021-10-09 2023-10-20 上海第二工业大学 一种基于计算机视觉的机械臂抓取方法
CN114067309B (zh) * 2021-10-25 2024-04-26 大连理工大学 一种基于多视角三维重建的甜椒识别与采摘顺序确定方法
CN114067309A (zh) * 2021-10-25 2022-02-18 大连理工大学 一种基于多视角三维重建的甜椒识别与采摘顺序确定方法
CN114147714A (zh) * 2021-12-02 2022-03-08 浙江机电职业技术学院 自主机器人机械臂控制参数计算方法和系统
CN114998573A (zh) * 2022-04-22 2022-09-02 北京航空航天大学 一种基于rgb-d特征深度融合的抓取位姿检测方法
CN114998573B (zh) * 2022-04-22 2024-05-14 北京航空航天大学 一种基于rgb-d特征深度融合的抓取位姿检测方法
CN114734444B (zh) * 2022-04-27 2023-06-27 博众精工科技股份有限公司 一种目标定位方法、装置、电子设备及存储介质
CN114734444A (zh) * 2022-04-27 2022-07-12 博众精工科技股份有限公司 一种目标定位方法、装置、电子设备及存储介质
CN115990889A (zh) * 2023-03-23 2023-04-21 武汉益模科技股份有限公司 一种基于多感受野的复合定位方法、装置及计算机系统
CN116968037A (zh) * 2023-09-21 2023-10-31 杭州芯控智能科技有限公司 一种多机械臂协同任务调度方法
CN116968037B (zh) * 2023-09-21 2024-01-23 杭州芯控智能科技有限公司 一种多机械臂协同任务调度方法
CN117251058A (zh) * 2023-11-14 2023-12-19 中国海洋大学 一种多信息体感交互系统的控制方法
CN117251058B (zh) * 2023-11-14 2024-01-30 中国海洋大学 一种多信息体感交互系统的控制方法

Also Published As

Publication number Publication date
CN110605714B (zh) 2021-08-03
CN110605714A (zh) 2019-12-24

Similar Documents

Publication Publication Date Title
WO2021023315A1 (zh) 一种基于人眼凝视点的手眼协调抓取方法
CN109255813B (zh) 一种面向人机协作的手持物体位姿实时检测方法
US11030455B2 (en) Pose recognition method, device and system for an object of interest to human eyes
CN110480634B (zh) 一种针对机械臂运动控制的手臂引导运动控制方法
US7997730B2 (en) Explicit raytracing for gimbal-based gazepoint trackers
CN107253192A (zh) 一种基于Kinect的无标定人机交互控制系统及方法
JP2022542241A (ja) ロボット装置からの視覚的出力を補強するシステム及び方法
WO2009043927A1 (en) Apparatus for acquiring and processing information relating to human eye movements
CN106625673A (zh) 狭小空间装配系统及装配方法
CN110136202A (zh) 一种基于ssd与双摄像头的多目标识别与定位方法
CN109297413A (zh) 一种大型筒体结构视觉测量方法
CN109345542A (zh) 一种穿戴式视觉注视目标定位装置及方法
Seo et al. A comparative study of in-field motion capture approaches for body kinematics measurement in construction
Droeschel et al. 3D body pose estimation using an adaptive person model for articulated ICP
CN116222543B (zh) 用于机器人环境感知的多传感器融合地图构建方法及系统
CN111712780A (zh) 用于增强现实的系统和方法
CN111524174A (zh) 一种动平台动目标双目视觉三维构建方法
Grewal et al. Autonomous wheelchair navigation in unmapped indoor environments
CN113505694A (zh) 一种基于视线追踪的人机交互方法、装置及计算机设备
CN114310957A (zh) 用于医疗检测的机器人系统及检测方法
Khan et al. On the calibration of active binocular and RGBD vision systems for dual-arm robots
CN108621164A (zh) 基于深度相机的太极推手机器人
Jin et al. Human-robot interaction for assisted object grasping by a wearable robotic object manipulation aid for the blind
Jin et al. A wearable robotic device for assistive navigation and object manipulation
Liu et al. 3D gaze estimation for head-mounted devices based on visual saliency

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20849509

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20849509

Country of ref document: EP

Kind code of ref document: A1