Optimal control method and system for flexible endoscope robot based on image moment features

Abstract

The application discloses a flexible endoscope robot optimal control method and system based on image moment features, and belongs to the field of medical robots.The system comprises the following modules: an image data acquisition module, which acquires image information of a human body cavity by using a micro camera and pre-processes the image information; an image moment feature analysis module, which calculates image features and constructs an interaction matrix by using the image features; an optimal motion control module, which takes the image features as input, constructs an optimal control strategy, and outputs an optimal driving amount; and an extended state observation module, which can perform motion compensation for errors of a motion model and an environment, and an endoscope driving module, which receives the driving amount to adjust the position and posture of the endoscope, so that the difference between actual image features and expected features is reduced.The application innovatively proposes an optimal control method with image moments as features, has good visual anti-interference capability, can control the flexible endoscope with high precision and stability, and is helpful to reduce the surgical risk of endoscopy.

Description

Technical Field

[0001] This invention belongs to the field of medical robot technology, specifically relating to an optimal control method and system for a flexible endoscopic robot based on image moment features. Background Technology

[0002] Traditional flexible endoscopes are widely used in clinical surgery because they can enter the human body through natural holes or small incisions without the need for large incisions. Doctors can directly visualize lesions inside the patient's body, which has great advantages in routine diagnosis such as examination of intestinal diseases, endotracheal intubation, and gastric diseases. For patients, minimally invasive surgery causes less harm and allows for faster postoperative recovery.

[0003] However, the use of conventional flexible endoscopes has some problems. First, in some smaller cavities, flexible endoscopes have difficulty passing through large bends or narrow channels, which may result in some areas not being fully detected; second, doctors who perform flexible endoscopy often need to undergo specialized training and accumulate extensive experience, which increases costs; and finally, during the examination, human error may cause the target to be lost.

[0004] Therefore, to address these issues, it is necessary to provide an innovative method for controlling a flexible endoscope to overcome the limitations of traditional methods and improve the success rate of intubation procedures, patient comfort, and surgeon efficiency. This new technology should reduce trauma and discomfort during intubation while maintaining reliability in various complex situations. Furthermore, the technology should focus on improving healthcare professionals' control over the intubation process, thereby reducing operational risks.

[0005] The innovative endotracheal intubation technique introduced in this invention is designed to solve the above problems. It aims to reduce the risk of trauma during intubation, improve the success rate of the surgery, and make intubation surgery easier to perform by using new instruments or technical means. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention proposes an optimal control method and system for flexible endoscope robots based on image moment features, which enables the flexible endoscope robot to automatically adjust the posture of the endoscope.

[0007] To achieve the above objectives, the present invention provides the following solution:

[0008] The optimal control method for a flexible endoscopic robot based on image moment features includes the following steps:

[0009] S1: Acquire actual image information of the human airway and process the actual image information;

[0010] S2: Based on the processed actual image information, calculate the image features and interaction matrix of the actual image pose; based on the preset expected image information, calculate the image features and interaction matrix of the expected image pose.

[0011] S3: Based on the image features and interaction matrix of the desired image pose and the actual image pose, the optimal solution for the motion speed of each joint of the flexible endoscope robot is obtained;

[0012] S4: Based on the optimal solution of the joint movement speed of the flexible endoscope robot, motion compensation is performed on the error between the motion model and the environment during the actual control process.

[0013] S5: The optimal solution of the movement speed of each joint of the flexible endoscope robot after compensation is input to the drive motor of each joint of the flexible endoscope robot, thereby realizing the control of the flexible endoscope robot.

[0014] Preferably, in step S1, the method for processing the actual image information includes:

[0015] Deep learning is used to extract cavity information from the actual image information to obtain human body cavity image information;

[0016] An adaptive thresholding method is used to binarize the human body cavity image information to obtain a binarized image.

[0017] Preferably, in step S2, the method for calculating the image features and interaction matrix of the desired image pose based on preset desired image information includes:

[0018] s0=[x n0 y n0 a n0 s x0 s y0 α0] T

[0019]

[0020] In the formula, s0 represents the desired image features, and L0 represents the desired interaction matrix;

[0021] The method for calculating the image features and interaction matrix of the actual image pose based on the processed actual image information includes:

[0022] s1=[x n1 y n1 a n1 s x1 s y1 α1] T

[0023]

[0024] In the formula, s1 represents the actual image features, and L1 represents the actual interaction matrix.

[0025] Preferably, in step S3, the method for solving the optimal solution for the motion velocity of each joint of the flexible endoscope robot based on the image features and interaction matrix of the desired image pose and the actual image pose includes:

[0026] The image features of the actual image pose are compared with the image features of the desired image pose, and the image feature error and interaction matrix are calculated.

[0027] Based on the image feature error and the interaction matrix, the velocity of the camera's desired position and orientation is obtained;

[0028] Based on the camera's desired position and orientation velocity, and combined with the robot's Jacobian matrix, the camera's actual position and orientation velocity are obtained.

[0029] Based on the actual position and orientation of the camera, and using a nonlinear solver, the rotational speed, bending speed, and feed speed of the flexible endoscope robot are obtained, thus solving for the optimal solutions for the speeds of each joint of the flexible endoscope robot.

[0030] Preferably, in step S4, the method for motion compensation of the error between the motion model and the environment during the actual control process, based on the optimal solution of the motion velocity of each joint of the flexible endoscope robot, includes:

[0031] Establish the state-space equation of the robot: Δq out =Δq + W, where: Δq out Δq is the system output, W is the error between the model and the environment, and Δq is the joint velocity.

[0032] Establish an observer to define the actual state of the robot joints. The actual error between the model and the environment is The error between the actual value and the output value is

[0033] Establish equations and Where β1 and β2 are constants, calculate the compensation value of Z2. Where t is the time constant;

[0034] The calculation result of motion compensation is Δq. out =Δq+Z'2.

[0035] The present invention also discloses an optimal control system for a flexible endoscope robot based on image moment features, comprising: an image data acquisition module, an image moment feature analysis module, an optimal motion control module, an expansion state observation module, and an endoscope drive module;

[0036] The image data acquisition module is used to acquire actual image information of the human airway and process the actual image information;

[0037] The image moment feature analysis module is used to calculate the image features and interaction matrix of the actual image pose based on the processed actual image information; and to calculate the image features and interaction matrix of the desired image pose based on the preset desired image information.

[0038] The optimal motion control module is used to solve for the optimal solution of the motion speed of each joint of the flexible endoscope robot based on the image features and interaction matrix of the desired image pose and the actual image pose.

[0039] The expansion state observation module is used to perform motion compensation for errors between the motion model and the environment during the actual control process, based on the optimal solution of the motion speed of each joint of the flexible endoscope robot.

[0040] The endoscope drive module is used to input the optimal solution of the movement speed of each joint of the flexible endoscope robot after compensation to the drive motor of each joint of the flexible endoscope robot, thereby realizing the control of the flexible endoscope robot.

[0041] Preferably, the image data acquisition module includes: a cavity information extraction unit and a binarization unit;

[0042] The cavity information extraction unit is used to extract cavity information from the actual image information using deep learning to obtain human cavity image information.

[0043] The binarization unit is used to perform binarization processing on the human body cavity image information using an adaptive thresholding method to obtain a binarized image.

[0044] Preferably, in the image moment feature analysis module, the process of calculating the image features and interaction matrix of the desired image pose based on preset desired image information includes:

[0045] s0=[x n0 y n0 a n0 s x0 s y0 α0] T

[0046]

[0047] In the formula, s0 represents the desired image features, and L0 represents the desired interaction matrix;

[0048] The process of calculating the image features and interaction matrix of the actual image pose based on the processed actual image information includes:

[0049] s1=[x n1 y n1 a n1 s x1 s y1 α1] T

[0050]

[0051] In the formula, s1 represents the actual image features, and L1 represents the actual interaction matrix.

[0052] Preferably, the optimal motion control module includes: an error calculation unit, a desired speed calculation unit, an actual speed calculation unit, and an optimal solution unit;

[0053] The error calculation unit is used to compare the image features of the actual image pose with the image features of the desired image pose, and calculate the image feature error and the interaction matrix.

[0054] The desired velocity calculation unit is used to obtain the velocity of the camera's desired position and attitude based on the image feature error and the interaction matrix;

[0055] The actual velocity calculation unit is used to obtain the actual position and attitude velocity of the camera based on the velocity of the camera's expected position and attitude, combined with the robot's Jacobian matrix;

[0056] The optimal solution unit is used to obtain the rotational speed, bending speed and feed speed of the flexible endoscope robot based on the actual position and orientation of the camera, and using a nonlinear solver, thus solving for the optimal solution of the speed of each joint of the flexible endoscope robot.

[0057] Preferably, in the expansion state observation module, the process of motion compensation for errors between the motion model and the environment during actual control, based on the optimal solution of the joint motion velocity of the flexible endoscope robot, includes:

[0058] Establish the state-space equation of the robot: Δq out =Δq + W, where: Δq out Δq is the system output, W is the error between the model and the environment, and Δq is the joint velocity.

[0059] Establish an observer to define the actual state of the robot joints. The actual error between the model and the environment is The error between the actual value and the output value is

[0060] Establish equations and Where β1 and β2 are constants, calculate the compensation value of Z2. Where t is the time constant;

[0061] The calculation result of motion compensation is Δq. out =Δq+Z'2.

[0062] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0063] This invention calculates the actual features and interaction matrix after image processing, then performs calculations with the desired features and interaction matrix to obtain the desired state changes of the camera. A nonlinear solver is then used to solve for the optimized joint velocities, adjusting the state of each joint to reduce feature errors. This invention enables endoscope control even with incomplete images, improving the safety and efficiency of endoscope use. Attached Figure Description

[0064] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0065] Figure 1 This is a flowchart illustrating an optimal control method for a flexible endoscopic robot based on image moment features according to the present invention.

[0066] Figure 2 This is a schematic diagram of the structure of an optimal control system for a flexible endoscopic robot based on image moment features according to the present invention.

[0067] Figure 3 This is a schematic diagram of the motion state of the endoscope in an optimal control system for a flexible endoscope robot based on image moment features according to the present invention. Detailed Implementation

[0068] 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.

[0069] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0070] Example 1

[0071] like Figure 1 As shown, this invention provides an optimal control method for a flexible endoscopic robot based on image moment features, comprising the following steps:

[0072] S1: Acquire actual image information of the human airway and process the actual image information;

[0073] S2: Based on the processed actual image information, calculate the image features and interaction matrix of the actual image pose; based on the preset expected image information, calculate the image features and interaction matrix of the expected image pose.

[0074] S3: Based on the image features and interaction matrix of the desired image pose and the actual image pose, the optimal solution for the motion speed of each joint of the flexible endoscope robot is obtained;

[0075] S4: Based on the optimal solution of the joint motion speed of the flexible endoscope robot, motion compensation is performed on the error between the motion model and the environment during the actual control process.

[0076] S5: The optimal solution of the movement speed of each joint of the flexible endoscope robot after compensation is input to the drive motor of each joint of the flexible endoscope robot, thereby realizing the control of the flexible endoscope robot.

[0077] In this embodiment, S1: Acquiring actual image information of the human airway and processing the actual image information specifically includes:

[0078] First, the endoscope acquires real-time images within the human airway. Deep learning is used to extract cavity information from the images, and an adaptive thresholding method is employed to binarize the original images, resulting in a binarized image. For the binarized image, (where (x, y) are the image coordinates of the image plane).

[0079] In this embodiment, S2: Calculating the feature and interaction matrices of the desired image pose and the actual image pose respectively specifically includes:

[0080] Features are extracted from the cavity in the binarized image to obtain the actual features s1 and interaction matrix L1. Based on the desired cavity pose, feature extraction is performed to obtain the desired features s0 and interaction matrix L0. The real-time image features s1 are compared with the desired features s0, and the image feature error e = s1 - s0 and the interaction matrix L0 are calculated. When controlling the robot, when e shrinks to the set threshold, the robot's end effector reaches the desired position and completes the tracking operation.

[0081] In this embodiment, S3: Optimizing the velocity of each joint based on the desired pose of the camera specifically includes:

[0082] Using the image feature error and interaction matrix obtained in S2, the desired position and pose of the camera, v, can be obtained. cd Combined with the robot's Jacobian matrix J robot The position and orientation of the actual camera controlled by the robot are obtained. c Then, the robot's rotation speed is obtained using optimal control. Bending speed and feed rate

[0083] In this embodiment, S4: Based on the optimal control values ​​of each joint obtained in S3, motion compensation is performed to address the errors between the motion model and the environment during the actual control process.

[0084] In this embodiment, S5: Control the movement of each joint of the robot according to the compensated speed of each joint movement;

[0085] The compensated motion speeds of each joint of the robot in S4 are transmitted to each drive motor, thereby adjusting the robot's pose.

[0086] If the calculated image feature error is consistently not less than the threshold, repeat the above steps until the error is less than the threshold.

[0087] The specific definition is as follows:

[0088] The definition of image moments is: Where i = 0, 1, 2, 3..., j = 0, 1, 2, 3...

[0089] The center distance is defined as: Where i = 0, 1, 2, 3..., j = 0, 1, 2, 3..., x g =m 10 / m 00 With y g =m 01 / m 00 The center of the image moments.

[0090] Specifically, the corresponding interaction matrices are as follows:

[0091]

[0092] in:

[0093] m vx =-i(Am ij +Bm i-1,j+1 +Cm i-1,j )-Am ij

[0094] m vy =-j(Am i+1,j +Bm ij +Cm i,j-1 )-Bm ij

[0095] m vz =(i+j+3)(Am i+1,j +Bm i,j+1 +Cm ij )-Cm ij

[0096] m wx =(i+j+3)m i,j+1 +jm i,j-1

[0097] m wy =-(i+j+3)m i+1,j -im i-1,j

[0098] m wz =im i-1,j+1 -jm i+1,j-1

[0099] L μij =[μ vx m vy m vz m wx m wy m wz ]

[0100] Among them:

[0101] m vx =-(i+1)Aμ ij -iBμ i-1,j+1

[0102] m vy =-jAμ i+1,j-1 -(j+1)Bμ ij

[0103] m vz =-Am wy +Bμ wx +(i+j+2)Cμ ij

[0104] m wx= (i+j+3)μ i,j+1 +(i+2j+3)y g μ ij +ix g μ i-1,j+1 -in 11 μ i-1,j -jn 02 μ i,j-1

[0105] μ wy =-(i+j+3)μ i+1,j -(2i+j+3)x g μ ij -jy g μ i+1,j-1 +in 20 μ i-1,j +jn 11 μ i,j-1

[0106] μ wz =iμ i-1,j+1 -jμ i+1,j-1

[0107] Where A, B, and C are constants;

[0108] Specifically, the moment center of the image is:

[0109] x g =m 10 / m 00

[0110] y g =m 01 / m 00

[0111] Specifically, the features s and the interaction matrix L are calculated based on the image information:

[0112] Calculate the moment center (x) of the image region g ,y g The area of ​​the target region is a = m. 00 The orientation angle α of the target area, and the feature s that controls rotation around the X-axis. x The feature s that controls rotation around the Y-axis y .

[0113] To reduce motion coupling, the features are normalized. The normalization process is as follows: x n =a n x g ,y n =a n y g , where z *It is the expected depth, a and a * Let x be the area of ​​the actual region and the area of ​​the expected region, respectively. Therefore, let feature one be: x n Feature 2 is y n Feature three is: a n

[0114] The area of ​​the target image is calculated as follows:

[0115]

[0116] The direction angle of the target object Feature six.

[0117] The feature four that controls rotation about the X-axis x With control of rotation around the Y-axis, feature five s y for:

[0118] s x =(c2c3+s2s3) / K

[0119] s y =(s2c3-c2s3) / K

[0120] in:

[0121] c1=μ 20 -μ 02

[0122] c2=μ 03 -3μ 21

[0123]

[0124] s1=2μ 11

[0125] s2=μ 30 -3μ 12

[0126] s3=2s1c1

[0127]

[0128]

[0129] I3=μ 20 +μ 02

[0130]

[0131] The aforementioned feature is s = [x n y n a n sx s y α] T .

[0132] The interaction matrix of the target is calculated as follows:

[0133] in:

[0134]

[0135]

[0136]

[0137] L α =[0 0 0 α wx α wy -1]

[0138]

[0139]

[0140] n ij =μ ij / m 00 Where i = 0, 1, 2, 3..., j = 0, 1, 2, 3...

[0141] ε 11 =ε 22 =4n 11 -x g y g / 2;

[0142] ε 31 =3y g / 2;

[0143] ε 32 =3x g / 2;

[0144]

[0145]

[0146]

[0147]

[0148]

[0149]

[0150]

[0151]

[0152]

[0153]

[0154] Specifically, the expected features s0 and interaction matrix L0 of the desired location image information are calculated as follows:

[0155] s0=[x n0 y n0 a n0 s x0 s y0 α0] T

[0156]

[0157] Specifically, the actual feature s1 and interaction matrix L1 calculated from the actual location image information are: s1 = [x n1 y n1 a n1 s x1 s y1 α1] T

[0158]

[0159] Specifically, calculate the optimal value of the joint velocity:

[0160] Under constraints (λ is a constant, It is the interaction matrix L s (inverse matrix), v c =J robot Δq、 and ( The rotation angle of the rotary joint, the bending angle of the θ-segment, and J robot Given the Jacobian matrix of the robot, M = ||v cd -v c || 2 When the value is at its minimum, solve for the joint velocity Δq.

[0161] in:

[0162]

[0163] l is the length of the flexible segment. θ is the angle of rotation, and θ is the angle of curvature.

[0164] Solve for the joint velocity Δq, and adjust the robot's position and orientation in step S4.

[0165] in It is the expression for the actual plane, where A, B, and C are constants, Z is the depth of the actual plane in the camera coordinate system, and (x, y) are the coordinates of a point on the image plane.

[0166] Specifically, the motion compensation process for errors between the motion model and the environment during actual control is as follows:

[0167] First, establish the state-space equations of the robot:

[0168] Δq out =Δq+W

[0169] Where: Δq out This refers to the system's output.

[0170] W represents the error between the model and the environment;

[0171] Then establish an observer, setting the actual state of the robot joints as follows: The actual error between the model and the environment is The error between the actual value and the output value is

[0172] Establish equations and (β1, β2 are constants), calculate the compensation value of Z2.

[0173]

[0174] Finally, the motion compensation result can be calculated as Δq. out =Δq+Z'2.

[0175] When controlling the endotracheal intubation robot, the compensated movement speed of each joint is transmitted to each drive motor, thereby adjusting the robot's posture. S-S0 will decrease as the robot is controlled. At the same time, an endoscope can be inserted into the human airway.

[0176] Example 2

[0177] This invention provides an optimal motion control system for flexible endoscopes based on image moments, such as... Figure 2 As shown, it includes:

[0178] The image data acquisition module is used to acquire image information and perform binarization processing on the image information;

[0179] The image moment feature analysis module is used to calculate the feature and interaction matrices of the desired image pose and the actual image pose, respectively.

[0180] The optimal motion control module is used to optimize and solve for the optimal velocity of each joint based on the desired pose of the camera.

[0181] The extended state observation module is used to perform motion compensation for errors between the motion model and the environment during the actual control process.

[0182] The endoscope drive module is used to input the optimal solutions for the speeds of each joint obtained through optimization into the drive motors of each joint of the robot, thereby realizing the control of the flexible endoscope robot.

[0183] The image data acquisition module is used to acquire human body cavity image information in real time and perform binarization processing to obtain a binarized image.

[0184] The image moment feature analysis module is used to calculate the feature and interaction matrix of the desired image pose and the actual image pose based on the binarized image.

[0185] The optimal motion control module is used to calculate the desired camera speed using characteristic errors, and then use a nonlinear solver to solve for the optimal solution of the speed of each joint.

[0186] The module is capable of performing motion compensation for errors between the motion model and the environment during actual control processes.

[0187] The endoscope drive module is used to send the compensated speed to the motors of each joint of the endoscope to control the position and orientation of the endoscope.

[0188] In this embodiment, the image data acquisition module includes: a cavity information extraction unit and a binarization unit;

[0189] The cavity information extraction unit is used to extract cavity information from actual image information using deep learning to obtain human cavity image information;

[0190] The binarization unit is used to perform binarization processing on human cavity image information using an adaptive thresholding method to obtain a binarized image.

[0191] In this embodiment, the process of calculating the image features and interaction matrix of the desired image pose based on preset desired image information in the image moment feature analysis module includes:

[0192] s0=[x n0 y n0 a n0 s x0 s y0 α0] T

[0193]

[0194] In the formula, s0 represents the desired image features, and L0 represents the desired interaction matrix;

[0195] The process of calculating the image features and interaction matrix of the actual image pose based on the processed actual image information includes:

[0196] s1=[x n1 y n1 a n1 s x1 s y1 α1] T

[0197]

[0198] In the formula, s1 represents the actual image features, and L1 represents the actual interaction matrix.

[0199] In this embodiment, the optimal motion control module includes: an error calculation unit, a desired speed calculation unit, an actual speed calculation unit, and an optimal solution unit;

[0200] The error calculation unit is used to compare the image features of the actual image pose with the image features of the desired image pose, and calculate the image feature error and the interaction matrix;

[0201] The desired velocity calculation unit is used to obtain the velocity of the camera's desired position and orientation based on image feature errors and the interaction matrix;

[0202] The actual velocity calculation unit is used to obtain the actual velocity of the camera based on the camera's expected position and orientation, combined with the robot's Jacobian matrix;

[0203] The optimal solution unit is used to obtain the rotational speed, bending speed and feed speed of the flexible endoscope robot based on the actual position and orientation of the camera, and uses a nonlinear solver to obtain the optimal solution for the speed of each joint of the flexible endoscope robot.

[0204] In this embodiment, the process of motion compensation for errors between the motion model and the environment during actual control, based on the optimal solution of the joint motion velocity of the flexible endoscope robot in the expansion state observation module, includes:

[0205] Establish the state-space equation of the robot: Δq out =Δq + W, where: Δq out Δq is the system output, W is the error between the model and the environment, and Δq is the joint velocity.

[0206] Establish an observer to define the actual state of the robot joints. The actual error between the model and the environment is The error between the actual value and the output value is

[0207] Establish equations and Where β1 and β2 are constants, calculate the compensation value of Z2. Where t is the time constant;

[0208] The calculation result of motion compensation is Δq. out =Δq+Z'2.

[0209] Figure 3 This is a schematic diagram of the motion state of the endoscope in an optimal motion control system for a flexible endoscope based on image moments, as described in this invention. In actual control, adjusting the control handle can realize the bending motion of the endoscope. The camera's rotation is achieved by rotating the entire endoscope. When the feature error is less than the set threshold value, the advance movement can be completed by holding the endoscope.

[0210] The technical solution of this invention calculates the actual features and interaction matrix after image processing, then performs calculations with the desired features and interaction matrix to obtain the desired state changes of the camera. A nonlinear solver is then used to solve for the optimized joint velocities, adjusting the state of each joint to reduce feature errors. This invention enables endoscope control even with incomplete images, improving the safety and efficiency of endoscope use.

[0211] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. An optimal control system for a flexible endoscopic robot based on image moment features, characterized in that, include: Image data acquisition module, image moment feature analysis module, optimal motion control module, dilation state observation module, and endoscope drive module; The image data acquisition module is used to acquire actual image information of the human airway and process the actual image information; The image moment feature analysis module is used to calculate the image features and interaction matrix of the actual image pose based on the processed actual image information; and to calculate the image features and interaction matrix of the desired image pose based on the preset desired image information. The optimal motion control module is used to solve for the optimal solution of the motion speed of each joint of the flexible endoscope robot based on the image features and interaction matrix of the desired image pose and the actual image pose. The expansion state observation module is used to perform motion compensation for errors between the motion model and the environment during the actual control process, based on the optimal solution of the motion speed of each joint of the flexible endoscope robot. The endoscope drive module is used to input the optimal solution of the movement speed of each joint of the flexible endoscope robot after compensation to the drive motor of each joint of the flexible endoscope robot, thereby realizing the control of the flexible endoscope robot. In the image moment feature analysis module, the process of calculating the image features and interaction matrix of the desired image pose based on preset desired image information includes: ； ； In the formula, For the desired image features, Let be the desired interaction matrix; The process of calculating the image features and interaction matrix of the actual image pose based on the processed actual image information includes: ； ； In the formula, For actual image features, This is the actual interaction matrix; Calculate the optimal value of the joint velocity: Under constraints , , and Down, When the minimum value is reached, solve for the joint velocity. ,in, It is a constant. It is an interaction matrix inverse matrix, interaction matrix , Image feature error, The angle of rotation of the rotary joint, The angle of the bend in the curved section, For the Jacobian matrix of the robot, For the desired position and orientation of the camera, This refers to the actual position and orientation of the camera.

2. The optimal control system for a flexible endoscopic robot based on image moment features according to claim 1, characterized in that, The image data acquisition module includes: a cavity information extraction unit and a binarization unit; The cavity information extraction unit is used to extract cavity information from the actual image information using deep learning to obtain human cavity image information. The binarization unit is used to perform binarization processing on the human body cavity image information using an adaptive thresholding method to obtain a binarized image.

3. The optimal control system for a flexible endoscopic robot based on image moment features according to claim 1, characterized in that, The optimal motion control module includes: an error calculation unit, a desired speed calculation unit, an actual speed calculation unit, and an optimal solution unit; The error calculation unit is used to compare the image features of the actual image pose with the image features of the desired image pose, and calculate the image feature error and the interaction matrix. The desired velocity calculation unit is used to obtain the velocity of the camera's desired position and attitude based on the image feature error and the interaction matrix; The actual velocity calculation unit is used to obtain the actual position and attitude velocity of the camera based on the velocity of the camera's expected position and attitude, combined with the robot's Jacobian matrix; The optimal solution unit is used to obtain the rotational speed, bending speed and feed speed of the flexible endoscope robot based on the actual position and orientation of the camera, and using a nonlinear solver, thus solving for the optimal solution of the speed of each joint of the flexible endoscope robot.

Images