Control method based on multifunctional electric control mechanical arm

Abstract

The invention relates to a control method based on a multifunctional electric control mechanical arm, and belongs to the technical field of power system automation control. Comprising the following steps that a remote control terminal sends a switching operation control instruction to a mechanical arm control system; an image acquisition module carried by the mechanical arm acquires an operation environment image, and the operation environment image is transmitted to the remote control terminal in real time through an image transmission unit; the mechanical arm control system receives the operation environment image, an object recognition unit is used for conducting recognition and three-dimensional positioning on a target operation object in the image, and space coordinates and posture information of the target operation object are obtained; and the mechanical arm control system controls a mechanical arm execution mechanism to perform multi-degree-of-freedom coordination action according to the control instruction and the space coordinates and attitude information of the target operation object. By integrating deep learning target recognition, multi-closed-loop PID control and other technologies, the remote, automatic and precise switching operation of the power system is realized, the safety of operators is guaranteed, and the operation efficiency and reliability are improved.

Description

Technical Field

[0001] This invention relates to a control method for a multifunctional electrically controlled robotic arm, belonging to the field of power system automation control technology. Background Technology

[0002] In power system operation and maintenance, the switching operations (opening / closing) of equipment such as pole-mounted switches, drop-out fuses, and disconnectors are critical to ensuring the safe operation of the power grid. Traditional switching operations rely on manual on-site operation, which has the following technical drawbacks: Low safety: Operators need to be close to high-voltage equipment and are susceptible to risks such as electric shock and arc burns; Inefficient: Manual operation requires climbing and manual positioning, and the operation time for a single set of equipment usually exceeds 15 minutes; Poor environmental adaptability: Severe weather (such as heavy rain, dense fog, strong electromagnetic interference) or complex terrain may cause work to be interrupted; Insufficient precision: Manually grasping the handle can easily lead to equipment damage due to visual errors or improper operating force.

[0003] To address the aforementioned issues, some remote-controlled robotic arm solutions have emerged in the existing technology. However, these solutions generally suffer from weak target recognition robustness (significantly affected by lighting and occlusion), low control precision, and high transmission delay, making it difficult to meet the stringent requirements of power systems for operational safety and reliability.

[0004] Therefore, developing a multi-functional electric robotic arm control method that integrates high-precision target recognition, low-latency image transmission, multi-degree-of-freedom precise control, and closed-loop state feedback has become the key to replacing manual switching operations. Summary of the Invention

[0005] To address the problems of low safety and efficiency in manual switching operations and insufficient precision and poor environmental adaptability of traditional robotic arm control methods in existing technologies, this invention provides a control method for a multi-functional electrically controlled robotic arm. By integrating deep learning target recognition, anti-interference transmission, and multi-closed-loop PID control, it enables remote, automated, and precise switching operations in power systems, ensuring operator safety and improving operational efficiency and reliability.

[0006] The present invention discloses a control method for a multifunctional electrically controlled robotic arm, comprising the following steps: S1: The remote control terminal sends a switching operation control command to the robotic arm control system. The control command includes the target operation object type and operation action parameters. S2: The image acquisition module mounted on the robotic arm acquires images of the working environment and transmits the images of the working environment to the remote control terminal in real time through the image transmission unit; S3: The robotic arm control system receives the image of the working environment, identifies and three-dimensionally locates the target working object in the image through the object recognition unit, and obtains the spatial coordinates and posture information of the target working object. The target working object includes pole-mounted switches, drop-out fuses or disconnect switches in the power system. S4: The robotic arm control system controls the robotic arm actuator to perform multi-degree-of-freedom coordinated actions, including grasping, rotating and lifting actions, according to the control command and the spatial coordinates and posture information of the target work object, so as to perform a switching operation on the target work object. S5: During the switching operation, the robotic arm actuator collects the operation status parameters in real time through the status detection unit and feeds them back to the remote control terminal to form a closed-loop control until the switching operation is completed.

[0007] Preferably, the object recognition unit employs a target recognition and localization algorithm that combines a YOLOv5 deep learning model with ORB feature matching, specifically including: S31: The image of the work environment is preprocessed. After removing noise by Gaussian filtering, edge features are extracted by Canny operator and color features of the target work object are extracted by HSV color space segmentation. S32: Input the preprocessed image into the YOLOv5 model. The model enhances the feature weights of the target region by adding an attention mechanism module, and outputs the type of the target object and the coordinates of the two-dimensional bounding box. S33: Based on the ORB feature matching algorithm, feature points are extracted and matched for targets within a two-dimensional bounding box. Combined with the disparity map of the binocular vision system, the three-dimensional spatial coordinates (X,Y,Z) and yaw angles of the target object are calculated using the triangulation principle: pitch angle α and yaw angle β.

[0008] Preferably, the image transmission unit adopts a 4G / 5G communication protocol and integrates an anti-interference transmission algorithm based on forward error correction coding and automatic retransmission request, specifically including: The images of the work environment are compressed using H.265 encoding, with a compression ratio set to 30:1~50:1; The FEC algorithm is used to add redundant check codes to the encoded data, and the redundancy is dynamically adjusted according to the channel quality. When the receiving end detects that the data packet loss rate exceeds 5%, it triggers the ARQ retransmission mechanism to ensure that the transmission delay does not exceed 200ms and the image resolution is not less than 1920×1080 pixels.

[0009] Preferably, in step S4, the multi-degree-of-freedom cooperative motion control of the robotic arm actuator adopts a forward and inverse kinematics solution and cubic spline trajectory planning algorithm based on the DH parameter method, specifically including: S41: Grasping motion control: Based on the size and contour of the target work object, the end effector pose is calculated by inverse kinematics, the opening and closing trajectory of the gripping mechanism is planned by cubic spline interpolation, and the gripping force signal fed back by the force sensor is combined with the fuzzy PID algorithm to dynamically adjust the output torque of the servo motor. S42: Rotational motion control: Based on the deflection angle (α,β) of the target object, the kinematic model of the robotic arm is established through the DH parameter method, the coordinated rotation angle of the waist rotation joint and wrist rotation joint is calculated, and the joint motion is controlled by trapezoidal velocity planning. The rotation angle adjustment range is 0°~360°. S43: Lifting motion control: The lifting mechanism drives the robotic arm to lift and lower, and adopts a position-speed dual closed-loop PID control algorithm to draw a lifting speed curve according to the target lifting height.

[0010] Preferably, in step S5, the closed-loop control employs a fusion algorithm of three-loop PID and adaptive fuzzy PID, specifically including: Position loop: Taking the target spatial coordinates (X,Y,Z) as input, the real-time position of the joint is collected by the photoelectric encoder, and the joint motor rotation angle is adjusted by incremental PID control; Speed ​​loop: Taking joint movement speed commands as input, the motor speed is collected through Hall sensors, and PI control is used to suppress speed fluctuations; Force ring: Taking the target clamping force as input, the strain gauge force sensor collects the end clamping force and uses an adaptive fuzzy PID algorithm to dynamically adjust the proportional coefficient (Kp), integral coefficient (Ki), and derivative coefficient (Kd). When the deviation between the operation status parameters and the control command exceeds a preset threshold, the fuzzy PID parameter self-tuning is triggered to correct the motor drive signal until the deviation is less than the threshold.

[0011] Preferably, in step S5, the preset thresholds include: joint position deviation threshold ±1mm, clamping force deviation threshold ±2N, and lifting height deviation threshold ±0.5cm.

[0012] Preferably, the remote control terminal is equipped with a touch-screen human-machine interface for displaying images of the working environment, positioning results of the target object, and working status parameters, and supports manual adjustment of the motion parameters of the robotic arm actuator, including gripping force, rotation speed, and lifting speed.

[0013] Preferably, in step S3, the object recognition unit further includes an environment adaptability enhancement algorithm module, specifically: When the ambient light intensity is below 300 lux, the multi-scale Retinex algorithm is used for illumination compensation. The illuminance component and reflectance component of the image are separated by the Gaussian wrap function to improve the contrast of the target area. When rain and snow interference is detected, a wavelet transform-based denoising algorithm is used: the image is decomposed into wavelets, hard thresholding is used for high-frequency sub-bands, and non-local mean filtering is used for low-frequency sub-bands.

[0014] Preferably, in step S4, the switching operation includes a tripping operation and a closing operation, and the specific action control algorithm is as follows: Opening operation: The robotic arm actuator grasps the operating handle of the target object according to the size of the operating handle, and then rotates to the opening position according to the preset trajectory based on the rotation algorithm in step S42. Closing operation: Rotate to the closed position; After the action is completed, the force sensor detects the change in the reaction force of the handle, maintains the clamping state for 0.5s~1s to confirm that the operation is in place, and then controls the clamping mechanism to slowly release through the gripping algorithm in step S41.

[0015] Preferably, the process also includes a work completion verification step: after the switching operation is completed, the image acquisition module acquires the status image of the target work object again, the object recognition unit identifies its current status through the YOLOv5 model and performs feature matching with the preset target status. If the match is consistent, the operation is determined to be successful; otherwise, manual intervention is prompted through the remote control terminal.

[0016] Compared with existing technologies, the present invention provides a control method for a multifunctional electrically controlled robotic arm, which has the following advantages: Significantly enhanced safety: By replacing manual on-site operation with remote control, operators are protected from exposure to high voltage and harsh environments, fundamentally eliminating safety risks such as electric shock and falls.

[0017] Significantly improved work efficiency: Automated target recognition and positioning, multi-degree-of-freedom collaborative actions (single set of equipment operation time < 5 minutes), which improves efficiency compared to manual operation.

[0018] High control precision and reliability: The YOLOv5+ORB fusion recognition algorithm and three-closed-loop PID control are adopted to ensure accurate switching operations; anti-interference transmission and environmental adaptability algorithms ensure stable operation in complex scenarios.

[0019] Multifunctional and highly expandable: In addition to power switching operations, by adjusting the target recognition model and motion parameters, it can be adapted to scenarios such as goods handling on shelves of different heights and industrial assembly, and has strong versatility. Attached Figure Description

[0020] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0021] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0022] Example 1: like Figure 1 As shown, this embodiment discloses a control method for a multi-functional electronically controlled robotic arm, including the following steps: The remote control terminal sends switching operation control commands to the robotic arm control system. The control commands include the target operation object type and operation motion parameters. The robotic arm is equipped with an image acquisition module that acquires images of the working environment and transmits these images to the remote control terminal in real time via an image transmission unit. The robotic arm control system receives images of the working environment and uses an object recognition unit to identify and three-dimensionally locate the target work object in the image, thereby obtaining the spatial coordinates and posture information of the target work object. The target work object includes pole-mounted switches, drop-out fuses, or disconnect switches in the power system. The robotic arm control system controls the robotic arm actuator to perform multi-degree-of-freedom coordinated actions, including grasping, rotating and lifting actions, based on control commands and the spatial coordinates and posture information of the target object, so as to perform switching operations on the target object. During the switching operation, the robotic arm actuator collects the operation status parameters in real time through the status detection unit and feeds them back to the remote control terminal to form a closed-loop control until the switching operation is completed.

[0023] The control method of this invention is implemented based on the following hardware system: Remote control terminal: Equipped with a 10.1-inch touch screen, dual joystick controller and emergency stop button, supporting control command sending and status parameter display; Robotic arm body: 6-DOF serial structure (waist J1, upper arm J2, forearm J3, wrist J4 / J5 / J6), end gripping mechanism is two-finger pneumatic gripper, lifting device adopts ball screw drive (stroke 0.5m~5m, maximum load 50kg). Image acquisition module: Basler acA2500-14uc binocular camera, equipped with an 8mm fixed-focus lens and ISP preprocessing chip; Object recognition module: NVIDIA Jetson AGX Xavier processor (integrated 512-core Volta GPU), storing YOLOv5 model (training dataset contains 100,000+ images of power equipment) and ORB feature matching algorithm; Condition detection module: ATI Mini45 six-axis force sensor (range ±50N, accuracy ±0.1%FS), Heidenhain ROD 486 photoelectric encoder (resolution 10000 lines), STMicro LSM6DSOX tilt sensor; Main controller: ARM processor, equipped with 5G dual-mode communication module (supports SA / NSA networking) and EtherCAT industrial bus.

[0024] Example 2: Based on Example 1, the specific implementation process of the object recognition unit is as follows: Image preprocessing: Gaussian filtering (convolution kernel 5×5, σ=1.2) was applied to the RGB images acquired by the stereo camera to remove salt-and-pepper noise. Edge features were extracted using the Canny operator (threshold 50 / 150). The target region was segmented using the HSV color space (on-pole switch: H=100~120, S=50~255, V=50~255). The formula for the Gaussian function is:

[0025] Where σ = 1.2, the normalized convolution kernel weight matrix (partial example):

[0026] Gradient calculation: Using the Sobel operator, horizontal gradient Vertical gradient Gradient magnitude gradient direction (Quantized into four directions: 0°, 45°, 90°, and 135°); Non-maximum suppression: If the pixel gradient magnitude M(x,y) is greater than the magnitudes of the two adjacent pixels along the gradient direction, it is retained as an edge point.

[0027] Dual threshold filtering: low threshold 50 (retains 90% of edge pixels), high threshold 150 (suppresses 80% of noisy edges).

[0028] For pole-mounted switches (red / yellow), based on statistical analysis of 200 images of equipment from different manufacturers, the HSV range was determined as follows: H=100~120 (blue-series equipment), S=50~255, V=50~255. The conversion formula is as follows: Let R,G,B∈[0,255], max=max(R,G,B), min=min(R,G,B), Δ=max min:

[0029] YOLOv5 model: The CBAM attention mechanism module (channel attention + spatial attention) is added to the Neck layer of YOLOv5 to enhance the feature weights of the target region; the model input size is 640×640, and the CIoU loss function is used.

[0030] 3D localization: The ORB algorithm (FAST corner detection, BRIEF descriptor) is used to extract feature points for targets within the 2D bounding box, with ≥500 feature points extracted per frame.

[0031] The disparity map is calculated using the calibration parameters of the binocular camera (intrinsic parameter matrix, extrinsic parameter matrix, distortion coefficient), and the three-dimensional coordinates (X, Y, Z) are calculated using the triangulation principle (baseline distance 120mm, focal length 8mm). The deflection angles α and β are solved using the PnP algorithm.

[0032] (1) Binocular camera parameters Baseline distance B = 120mm (based on working distance of 1~5m, target device size of 50~300mm, ensuring parallax range of 50~300 pixels), focal length f = 8mm (physical focal length), pixel size d = 3.45μm, pixel focal length fp = f / d = 2319 pixels, intrinsic parameter matrix. .

[0033] (2) Triangulation (three-dimensional coordinate calculation) For the left camera image point (xl, yl) and the right camera image point (xr, yr) (row aligned, yl = yr), the parallax d = xl xr (pixels), depth 3D coordinates:

[0034] (3) PnP algorithm (deflection angle calculation) The target deflection angles α (pitch angle) and β (yaw angle) are solved using the PnP algorithm, with the 3D target point as input. and corresponding 2D pixels Output rotation matrix R, angle extraction formula:

[0035] Angle calculation accuracy ±0.1°.

[0036] Example 3: The motion control of the robotic arm actuator is specifically implemented as follows: Kinematic modeling: The kinematic model of the robotic arm was established using the DH parameter method. The DH parameters of each joint are shown in Table 1. Table 1 DH parameters for each joint

[0037] Trajectory planning: Cubic spline interpolation is used to plan the trajectory of the end effector to ensure the continuity of position, velocity and acceleration; for example, in the grasping action, the opening and closing trajectory of the clamping mechanism is: t∈[0,1s], the opening degree s(t)=s0 + at³ + bt² + ct, where s0 is the initial opening degree, and the coefficients a, b, c are obtained by solving the boundary conditions s(0)=s0, s(1)=s_target, s'(0)=0, s'(1)=0; Taking the gripping of a 50mm diameter disconnect switch handle as an example: Initial opening degree s0 = 100mm (greater than the handle diameter to avoid collision), target opening degree s_target = 50mm (clamping the handle). Boundary conditions: s(0)=100, s(1)=50, s'(0)=0 (initial velocity 0), s'(1)=0 (final velocity 0); Solving for the coefficients: Substituting into the equation, we get a = 60, b = -90, c = 0. The locus of opening and closing angles is s(t) = 100 + 60t. 3 -90t 2 t∈[0,1s], trajectory acceleration is continuous (maximum acceleration 180mm / s²). 2 No impact.

[0038] Force control adjustment: The closed-loop control employs a fusion algorithm of three-loop PID and adaptive fuzzy PID, specifically including: Position loop: Taking the target spatial coordinates (X,Y,Z) as input, the real-time position of the joint is collected by the photoelectric encoder, and the joint motor rotation angle is adjusted by incremental PID control; Speed ​​loop: Taking joint movement speed commands as input, the motor speed is collected through Hall sensors, and PI control is used to suppress speed fluctuations; Force ring: Taking the target clamping force as input, the strain gauge force sensor collects the end clamping force and uses an adaptive fuzzy PID algorithm to dynamically adjust the proportional coefficient Kp, integral coefficient Ki, and derivative coefficient Kd; When the deviation between the operation status parameters and the control command exceeds a preset threshold, the fuzzy PID parameter self-tuning is triggered to correct the motor drive signal until the deviation is less than the threshold.

[0039] The preset thresholds include: joint position deviation threshold ±1mm, clamping force deviation threshold ±2N, and lifting height deviation threshold ±0.5cm.

[0040] The specific implementation of the image transmission unit is as follows: H.265 encoding compression: H.265 encoding is implemented using JM reference software, with an I-frame interval of 30 frames, a P-frame interval of 5 frames, a quantization parameter QP=25~35, and a compression ratio dynamically adjusted to 30:1~50:1 (depending on image complexity). FEC+ARQ anti-interference: FEC uses LDPC code (code length 64800, code rate 0.8), and the redundancy is dynamically adjusted according to the 5G channel CQI value (redundancy of 30% when CQI=1~5, redundancy of 10% when CQI=6~10); the receiver detects packet loss through CRC check, and when the packet loss rate is >5%, a retransmission request is sent through the ARQ protocol to ensure that the transmission delay is ≤200ms (the measured average delay is 180ms±20ms).

[0041] Example 4: The specific process for verifying the completion of the assignment is as follows: After the switching operation is completed, the image acquisition module acquires the target status image a second time (such as the open state of the disconnecting switch after the circuit breaker is tripped). The object recognition unit calls the YOLOv5 model to identify the current state (open / closed) and extracts ORB feature points (≥300) of the target area. The system matches the circuit breaker points against a pre-defined target state feature library, using pre-stored standard feature templates for open / closed states. A FLANN matcher is then used to calculate the feature point matching rate.

[0042] Failure handling: When the matching rate is <95%, the remote control terminal will issue an audible and visual alarm to prompt manual intervention; The operation is considered successful when the matching rate is ≥95%; otherwise, an audible and visual alarm is issued via the remote control terminal to prompt manual intervention. The alarm sound frequency is 2kHz and lasts for 2 seconds.

[0043] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A control method for a multifunctional electrically controlled robotic arm, characterized in that, Includes the following steps: S1: The remote control terminal sends a switching operation control command to the robotic arm control system. The control command includes the target operation object type and operation action parameters. S2: The image acquisition module mounted on the robotic arm acquires images of the working environment and transmits the images of the working environment to the remote control terminal in real time through the image transmission unit; S3: The robotic arm control system receives the image of the working environment, identifies and three-dimensionally locates the target working object in the image through the object recognition unit, and obtains the spatial coordinates and posture information of the target working object. The target working object includes pole-mounted switches, drop-out fuses or disconnect switches in the power system. S4: The robotic arm control system controls the robotic arm actuator to perform multi-degree-of-freedom coordinated actions, including grasping, rotating and lifting actions, according to the control command and the spatial coordinates and posture information of the target work object, so as to perform a switching operation on the target work object. S5: During the switching operation, the robotic arm actuator collects the operation status parameters in real time through the status detection unit and feeds them back to the remote control terminal to form a closed-loop control until the switching operation is completed.

2. The control method for a multifunctional electrically controlled robotic arm according to claim 1, characterized in that, The object recognition unit employs a target recognition and localization algorithm that combines a YOLOv5 deep learning model with ORB feature matching, specifically including: S31: The image of the work environment is preprocessed. After removing noise by Gaussian filtering, edge features are extracted by Canny operator and color features of the target work object are extracted by HSV color space segmentation. S32: Input the preprocessed image into the YOLOv5 model. The model enhances the feature weights of the target region by adding an attention mechanism module, and outputs the type of the target object and the coordinates of the two-dimensional bounding box. S33: Based on the ORB feature matching algorithm, feature points are extracted and matched for targets within a two-dimensional bounding box. Combined with the disparity map of the binocular vision system, the three-dimensional spatial coordinates (X,Y,Z) and yaw angles of the target object are calculated using the triangulation principle: pitch angle α and yaw angle β.

3. The control method for a multifunctional electrically controlled robotic arm according to claim 1, characterized in that, The image transmission unit adopts the 4G / 5G communication protocol and integrates an anti-interference transmission algorithm based on forward error correction coding and automatic retransmission request, specifically including: The images of the work environment are compressed using H.265 encoding, with a compression ratio set to 30:1~50:1; The FEC algorithm is used to add redundant check codes to the encoded data, and the redundancy is dynamically adjusted according to the channel quality. When the receiving end detects that the data packet loss rate exceeds 5%, it triggers the ARQ retransmission mechanism to ensure that the transmission delay does not exceed 200ms and the image resolution is not less than 1920×1080 pixels.

4. The control method for a multifunctional electrically controlled robotic arm according to claim 2, characterized in that, In step S4, the multi-degree-of-freedom cooperative motion control of the robotic arm actuator adopts a forward and inverse kinematics solution and cubic spline trajectory planning algorithm based on the DH parameter method, specifically including: S41: Grasping motion control: Based on the size and contour of the target work object, the end effector pose is calculated by inverse kinematics, the opening and closing trajectory of the gripping mechanism is planned by cubic spline interpolation, and the gripping force signal fed back by the force sensor is combined with the fuzzy PID algorithm to dynamically adjust the output torque of the servo motor. S42: Rotational motion control: Based on the deflection angle (α,β) of the target object, the kinematic model of the robotic arm is established through the DH parameter method, the coordinated rotation angle of the waist rotation joint and wrist rotation joint is calculated, and the joint motion is controlled by trapezoidal velocity planning. The rotation angle adjustment range is 0°~360°. S43: Lifting motion control: The lifting mechanism drives the robotic arm to lift and lower, and adopts a position-speed dual closed-loop PID control algorithm to draw a lifting speed curve according to the target lifting height.

5. The control method for a multifunctional electrically controlled robotic arm according to claim 4, characterized in that, In step S5, the closed-loop control employs a fusion algorithm of three-loop PID and adaptive fuzzy PID, specifically including: Position loop: Taking the target spatial coordinates (X,Y,Z) as input, the real-time position of the joint is collected by the photoelectric encoder, and the joint motor rotation angle is adjusted by incremental PID control; Speed ​​loop: Taking joint movement speed commands as input, the motor speed is collected through Hall sensors, and PI control is used to suppress speed fluctuations; Force ring: Taking the target clamping force as input, the strain gauge force sensor collects the end clamping force and uses an adaptive fuzzy PID algorithm to dynamically adjust the proportional coefficient Kp, integral coefficient Ki, and derivative coefficient Kd; When the deviation between the operation status parameters and the control command exceeds a preset threshold, the fuzzy PID parameter self-tuning is triggered to correct the motor drive signal until the deviation is less than the threshold.

6. The control method for a multifunctional electrically controlled robotic arm according to claim 5, characterized in that, In step S5, the preset thresholds include: joint position deviation threshold ±1mm, clamping force deviation threshold ±2N, and lifting height deviation threshold ±0.5cm.

7. The control method for a multifunctional electrically controlled robotic arm according to claim 1, characterized in that, The remote control terminal is equipped with a touch-screen human-machine interface for displaying images of the working environment, positioning results of the target object, and working status parameters. It also supports manual adjustment of the motion parameters of the robotic arm actuator, including gripping force, rotation speed, and lifting speed.

8. The control method for a multifunctional electrically controlled robotic arm according to claim 1, characterized in that, In step S3, the object recognition unit further includes an environment adaptability enhancement algorithm module, specifically: When the ambient light intensity is below 300 lux, the multi-scale Retinex algorithm is used for illumination compensation. The illuminance component and reflectance component of the image are separated by the Gaussian wrap function to improve the contrast of the target area. When rain and snow interference is detected, a wavelet transform-based denoising algorithm is used: the image is decomposed into wavelets, hard thresholding is used for high-frequency sub-bands, and non-local mean filtering is used for low-frequency sub-bands.

9. The control method for a multifunctional electrically controlled robotic arm according to claim 4, characterized in that, In step S4, the switching operation includes opening and closing operations, and the specific action control algorithm is as follows: Opening operation: The robotic arm actuator grasps the operating handle of the target object according to the size of the operating handle, and then rotates to the opening position according to the preset trajectory based on the rotation algorithm in step S42. Closing operation: Rotate to the closed position; After the action is completed, the force sensor detects the change in the reaction force of the handle, maintains the clamping state for 0.5s~1s to confirm that the operation is in place, and then controls the clamping mechanism to slowly release through the gripping algorithm in step S41.

10. The control method for a multifunctional electrically controlled robotic arm according to claim 1, characterized in that, It also includes a work completion verification step: after the switching operation is completed, the image acquisition module acquires the status image of the target work object again, the object recognition unit identifies its current status through the YOLOv5 model and performs feature matching with the preset target status. If the match is consistent, the operation is determined to be successful; otherwise, manual intervention is prompted through the remote control terminal.

Images