An intelligent wiring control system for an electrical cabinet robot based on deep learning

Abstract

The application discloses a kind of based on deep learning electrical cabinet robot intelligent wiring control system, and the application relates to electrical engineering field, data acquisition module, image processing module, data processing module, data storage module and data control module, data acquisition module is by the camera installed on mechanical arm to the inside of electrical cabinet is lighted and photographed, the advantage of the application is that: when moving, mechanical arm drives camera to move by mechanical arm, and then camera shoots the internal data of electrical cabinet under different visual angles in the moving process, then the photo data of different angles is handled using DeepMask model, so that DeepMask model can obtain better electronic component data, to facilitate subsequent system according to the data of identification to carry out wiring processing.

Description

TECHNICAL FIELD

[0001] The application relates to the field of electrical engineering, in particular to an electrical cabinet robot intelligent wiring control system based on deep learning. BACKGROUND

[0002] With the continuous progress of science and technology and the development of industrial automation, the electrical cabinet plays a vital role in the field of power systems and industrial production. The electrical cabinet contains various electrical components such as circuit breakers, contactors, relays, etc. The wiring work between them directly affects the normal operation of the electrical system. In the process of applying robots to electrical cabinet wiring work, deep learning technology plays a key role. Deep learning can make robots have stronger intelligent perception and decision-making ability. Through learning a large amount of electrical cabinet image data, robots can quickly and accurately identify the type, position and posture of electrical components, thereby realizing precise wiring operation. In addition, robots also need to be equipped with advanced sensors and control systems to ensure their safe and stable work in the electrical cabinet.

[0003] However, when processing the inside of the electrical cabinet on the processing line, the traditional method is to fix the visual sensor on the top of the processing rack to obtain the internal information of the electrical cabinet, which limits the image data processed by the system and affects the recognition effect of the circuit breakers, contactors, terminals and other electronic components in the electrical cabinet. Therefore, we propose an electrical cabinet robot intelligent wiring control system based on deep learning. SUMMARY

[0004] The purpose of the present application is to provide an electrical cabinet robot intelligent wiring control system based on deep learning.

[0005] To solve the problems raised in the above background technology, the present application provides the following technical solution: an electrical cabinet robot intelligent wiring control system based on deep learning, comprising a data acquisition module, an image processing module, a data processing module, a data storage module and a data control module.

[0006] The data acquisition module supplements light and takes pictures of the inside of the electrical cabinet through the camera installed on the mechanical arm.

[0007] The image processing module is used for processing image data photographed by the data acquisition module, the data acquisition module photographs the inside of the electrical cabinet through the camera when the mechanical arm moves, and then transmits the photographed image data to the image processing module, the image processing module is used for denoising, enhancing and background segmentation of the image, and the image processing module is used for position recognition of the recognized electronic components, the boundary position of the electrical cabinet is recognized by using the image processing module to obtain the two-dimensional plane coordinate system of the electrical cabinet, a DeepMask model is arranged in the image processing module, the gray scale, color, brightness and texture of the electronic components in the electrical cabinet are extracted by using the DeepMask model, and the electronic components are segmented by using the extracted features, a temporary storage unit is arranged in the DeepMask model, the temporary storage unit is used for storing electronic component model data, after the data of the gray scale, color, brightness and texture are extracted, the DeepMask model calls the data in the temporary storage unit for matching, the model data matched successfully is modeled according to the coordinate position of the electronic components recognized by the image processing module, and the modeling result data is transmitted to the data processing module for processing.

[0008] The data processing module is used for acquiring the installation data of the electronic components and generating corresponding instruction data.

[0009] The data storage module is used for storing the feature data of the image processing module and the instruction data generated by the data processing module.

[0010] The data control module is used for controlling the mechanical arm to process the circuit in the electrical cabinet.

[0011] As a further scheme of the application, the DeepMask model obtains the loss function of the electronic components by a formula, and the specific formula is as follows:

[0012]

[0013] Wherein, L1 (θ) represents the loss function of the electronic component L1, R represents the weight coefficient of the loss function, and the weight coefficient is affected by the definition of image shooting, Indicates the segmentation data in the DeepMask model, ij represents the pixel data of the electronic component, Indicates the segmentation data of the electronic component in the pixel ij, e represents a constant, and the approximate value of the constant is about 2.71828, Indicates the prediction data of the electronic component in the actual state, X represents the segmentation result data, and Y represents the prediction result data.

[0014] As a further scheme of the present application: the image data of the electronic components at different angles in the image is obtained through the movement of the mechanical arm, and the image processing module and the loss function are used to continuously optimize the electronic component image data.

[0015] As a further scheme of the present application: a model matching unit is arranged in the image processing module, which is used to call the modeling data in the temporary storage unit, and simultaneously obtain the electrical cabinet model data constructed in the DeepMask model, and match the data in the DeepMask model with the data in the temporary storage unit.

[0016] As a further scheme of the present application: when the electrical cabinet model data constructed in the DeepMask model is consistent with the data in the temporary storage unit, the model matching unit generates a calling instruction and transmits the instruction to the data processing module, which calls the instruction data in the data storage module and optimizes the movement path of the mechanical arm according to the circuit in the image data;

[0017] When the electrical cabinet model data constructed in the DeepMask model is inconsistent with the data in the temporary storage unit, the data processing module re-plans the movement path of the mechanical arm, simultaneously obtains the circuit route in the electrical cabinet through the data processing module, and then adjusts the height and angle of the mechanical arm during movement, so that the mechanical arm bypasses the circuit shielding part, and updates the model data in the temporary storage unit and eliminates the original model data.

[0018] As a further scheme of the present application: a camera for shooting the electrical cabinet is fixedly installed on the mechanical arm base, and the camera and the camera cooperate with each other to obtain the electronic component coordinate data at different angles, and the electronic component coordinate data is obtained through a formula, and the specific formula is as follows:

[0019]

[0020] Wherein, x represents the horizontal coordinate of the electronic component coordinate data, x1 and y1 represent the camera coordinates, x2 and y2 represent the camera head coordinates, a represents the included angle between the line connecting the camera shooting point and the target point and the horizontal direction, and b represents the included angle between the line connecting the camera head shooting point and the target point and the horizontal direction.

[0021] As a further scheme of the present application: the data processing module matches the calculated horizontal coordinate data with the horizontal coordinate data calculated by the DeepMask model, when the horizontal coordinate data is consistent, the data processing module enters the next step, and when the horizontal coordinate data deviates, x is brought into y=y1+(x-x1)tan(a) to re-verify the electronic component coordinate data.

[0022] As a further scheme of the present application: after receiving the instruction data, the data control module controls the mechanical arm to perform wiring processing on the circuit in the electrical cabinet, simultaneously captures the working process of the mechanical arm through the camera, and transmits the working process image to the DeepMask model in real time, and uses the DeepMask model to evaluate the wiring quality of the mechanical arm.

[0023] Compared with the prior art, the present application has the following advantages:

[0024] 1. The mechanical arm moves, and the camera moves with the mechanical arm, thereby capturing the internal data of the electrical cabinet at different angles during movement, and then processing the photo data at different angles by using the DeepMask model, so that the DeepMask model can obtain better electronic component data, and the subsequent system can perform wiring processing according to the recognized data.

[0025] 2. The present application obtains the segmentation data of the electronic component by the loss function of the formula, and constantly optimizes the electronic component by the data captured at different angles, so that the DeepMask model obtains more accurate electronic components, and improves the processing effect of the data processing module on the electronic components.

[0026] 3. The present application quickly obtains whether the modeling data of the DeepMask model is consistent with the model data in the temporary storage unit by the model matching unit, so that the subsequent data processing module obtains historical instruction data, the data processing module optimizes the historical instruction data of the mechanical arm, thereby improving the optimization computing power of the data processing module, and achieving the effect of constantly optimizing the moving path of the mechanical arm.

[0027] 4. The present application forms an included angle between the camera and the camera, thereby identifying the electronic components in the electrical cabinet, and analyzing the position of the electronic components again, thereby preventing errors in the coordinates and resulting in poor subsequent wiring effect. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 The present application is a wiring control system flowchart in the embodiment.

[0029] Figure 2 The present application is a model matching unit flowchart in the embodiment. DETAILED DESCRIPTION

[0030] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0031] Embodiment one:

[0032] A kind of intelligent wiring control system of robot based on deep learning for electrical cabinet, with the continuous progress of science and technology and the development of industrial automation, electrical cabinet plays a vital role in electric power system, industrial production and other fields, electrical cabinet contains various electrical components, such as circuit breaker, contactor, relay, etc., wiring work between them directly affects the normal operation of electrical system, in the process of robot application in electrical cabinet wiring work, deep learning technology plays a key role, deep learning can make robot have stronger intelligent perception and decision-making ability, through the learning of a large number of electrical cabinet image data, robot can quickly and accurately identify the type, position and posture of electrical component, to realize accurate wiring operation, in addition, robot also needs to be equipped with advanced sensor and control system, to ensure that it can work safely and stably in electrical cabinet;

[0033] But when processing the inside of electrical cabinet on processing production line, because traditional is through visual sensor fixedly installed on the top of processing frame to the inside information of electrical cabinet, and then lead to the image data processed by system is extremely limited, influence the recognition effect of circuit breaker, contactor, terminal and other electronic components in electrical cabinet.

[0034] Please refer to the accompanying Figure 1 -attached Figure 2 The present application is a kind of intelligent wiring control system of robot based on deep learning for electrical cabinet, data acquisition module, image processing module, data processing module, data storage module and data control module;

[0035] Data acquisition module carries out light supplement and shoots to the inside of electrical cabinet by camera installed on mechanical arm;

[0036] The image processing module is used for processing image data photographed by the data acquisition module. The data acquisition module photographs the inside of the electrical cabinet through the camera when the mechanical arm moves, and then transmits the photographed image data to the image processing module. The image processing module processes the image by denoising, enhancing and background segmentation. At the same time, the image processing module identifies the position of the recognized electronic components. The image processing module identifies the boundary position of the electrical cabinet to obtain the two-dimensional plane coordinate system of the electrical cabinet. A DeepMask model is set in the image processing module. The DeepMask model extracts the gray scale, color, brightness and texture of the electronic components in the electrical cabinet, and performs image segmentation on the electronic components using the extracted features. A temporary storage unit is set in the DeepMask model. The temporary storage unit is used to store electronic component model data. After the DeepMask model extracts the gray scale, color, brightness and texture data, the DeepMask model retrieves the data in the temporary storage unit for matching. The model data that matches successfully is modeled according to the coordinate position of the electronic component recognized by the image processing module, and the modeled result data is transmitted to the data processing module for processing.

[0037] The data processing module is used for planning the installation data of the mechanical arm through the obtained electronic components, and generating corresponding instruction data.

[0038] The data storage module is used for storing the feature data of the image processing module and the instruction data generated by the data processing module.

[0039] The data control module is used for controlling the mechanical arm to process the circuit in the electrical cabinet.

[0040] Specific working process: install the mechanical arm on the robot, then the mechanical arm obtains image data during movement, and the data acquisition module transmits the photographed image data to the image processing module.

[0041] The image processing module processes the image by denoising, enhancing and background segmentation.

[0042] The DeepMask model is used to extract the gray scale, color, brightness and texture of the electronic components in the electrical cabinet. The DeepMask model retrieves the data in the temporary storage unit and matches with the extracted features. The model data that matches successfully is modeled according to the coordinate position of the electronic component recognized by the image processing module. The image processing module transmits the modeled data to the data processing module.

[0043] The data processing module plans installation data of the mechanical arm according to the acquired electronic components, and generates corresponding instruction data; data storage: the data storage module stores feature data of the image processing module and instruction data generated by the data processing module; mechanical arm control: the data control module controls the mechanical arm to process the circuit in the electrical cabinet according to the instruction data.

[0044] Further, when the mechanical arm moves, the mechanical arm drives the camera to move, and then the camera shoots the internal data of the electrical cabinet at different angles during the movement, and then the DeepMask model processes the photo data at different angles, so that the DeepMask model can obtain better electronic component data, which is convenient for the subsequent system to process wiring according to the recognized data.

[0045] Embodiment two:

[0046] Based on embodiment one, please refer to the drawings, the DeepMask model obtains the loss function of the electronic component through the formula, and the specific formula is as follows:

[0047]

[0048] Wherein, L1(θ) represents the loss function of the electronic component L1, R represents the weight coefficient of the loss function, and the weight coefficient is affected by the definition of image shooting, represents the segmentation data in the DeepMask model, ij represents the pixel data of the electronic component, represents the segmentation data of the electronic component in the pixel ij, e represents a constant, and the approximate value of the constant is about 2.71828, represents the predicted data of the electronic component in the actual state, X represents the segmentation result data, and Y represents the predicted result data, the image data of the electronic component at different angles in the image is obtained through the movement of the mechanical arm, and the image processing module and the loss function are used to continuously optimize the electronic component image data.

[0049] Specific working process: according to the coordinate position of the electronic component recognized by the image processing module, the model data matched successfully is modeled, the loss function of the electronic component is calculated by using the formula, the loss function includes the part related to the segmentation data and the prediction data, and the part related to the intersection of the segmentation result data and the prediction result data, according to the result of the loss function, the DeepMask model is continuously optimized, so as to improve the accuracy of the electronic component image segmentation.

[0050] Further, the segmentation data of the electronic component is obtained through the loss function of the formula, and the electronic component is continuously optimized through the data taken at different angles, so that the DeepMask model obtains more accurate electronic components, and the processing effect of the data processing module on the electronic components is improved.

[0051] Embodiment three:

[0052] Based on the basis of embodiment two, please refer to the attached Figure 1 As shown in the accompanying drawings, the model matching unit is arranged in the image processing module, which is used to call the modeling data in the temporary storage unit, and at the same time, the electrical cabinet model data constructed in the DeepMask model is obtained, and the data in the DeepMask model is matched with the data in the temporary storage unit. When the electrical cabinet model data constructed in the DeepMask model is consistent with the data in the temporary storage unit, the model matching unit generates a calling instruction, and transmits the instruction to the data processing module. The data processing module calls the instruction data in the data storage module, and optimizes the moving path of the mechanical arm according to the line in the image data;

[0053] When the electrical cabinet model data constructed in the DeepMask model is inconsistent with the data in the temporary storage unit, the data processing module re-plans the moving path of the mechanical arm, and at the same time, the line route in the electrical cabinet is obtained through the data processing module, and then the height and angle of the mechanical arm during movement are adjusted, so that the mechanical arm bypasses the line shielding part, and at the same time, the model data in the temporary storage unit is updated, and the original model data is eliminated.

[0054] Specific working process: the image processing module uses the DeepMask model to extract features and model the electronic components in the electrical cabinet, and stores the modeling data in the temporary storage unit;

[0055] The model matching unit calls the modeling data from the temporary storage unit, and at the same time, the latest electrical cabinet model data constructed in the DeepMask model is obtained. The model matching unit compares the two data, if the data is consistent, the model matching unit generates a calling instruction, and transmits the instruction to the data processing module, if the data is inconsistent, the next step is entered;

[0056] After receiving the calling instruction, the data processing module calls the instruction data from the data storage module. The data processing module optimizes the moving path of the mechanical arm according to the line information in the image data, and the data control module controls the mechanical arm to move according to the optimized moving path;

[0057] The data processing module re-plans the moving path of the mechanical arm. The data processing module acquires the circuit route in the electrical cabinet, adjusts the height and angle of the mechanical arm when moving according to the circuit route, so that it can bypass the line shielding part, updates the model data in the temporary storage unit, and eliminates the original model data. The data control module controls the mechanical arm to move according to the new moving path and the adjusted height and angle.

[0058] Further, through the model matching unit, it is quickly determined whether the modeling data of the DeepMask model is consistent with the model data in the temporary storage unit, so that the subsequent data processing module acquires historical instruction data, so that the data processing module optimizes the historical instruction data of the mechanical arm, thereby improving the optimization computing power of the data processing module, and achieving the effect of continuously optimizing the moving path of the mechanical arm.

[0059] Embodiment Four:

[0060] Based on the basis of embodiment three, please refer to the drawings, the camera for shooting the electrical cabinet is fixedly installed on the mechanical arm base, the camera and the camera cooperate with each other, and then the electronic component coordinate data under different angles is obtained through mutual cooperation, and the electronic component coordinate data is obtained through formula, and the specific formula is as follows:

[0061]

[0062] Wherein, x represents the horizontal coordinate of the electronic component coordinate data, x1, y1 represents the camera coordinate, x2, y2 represents the camera coordinate, a represents the included angle between the line connecting the camera shooting point and the target point and the horizontal direction, b represents the included angle between the line connecting the camera shooting point and the target point and the horizontal direction, the data processing module matches the calculated horizontal coordinate data with the horizontal coordinate data calculated by the DeepMask model, when the horizontal coordinate data is consistent, then the data processing module enters the next step, when the horizontal coordinate data deviates, then x is brought into y=y1+(x-x1)tan(a), the electronic component coordinate data is re-verified, and the data control module controls the mechanical arm to process the circuit in the electrical cabinet after receiving the instruction data, and the working process of the mechanical arm is photographed through the camera, and the working process image is transmitted to the DeepMask model in real time, and the DeepMask model is used to evaluate the wiring quality of the mechanical arm.

[0063] The specific working process is that the camera and the camera in the data acquisition module on the mechanical arm cooperate with each other, and then the data processing module calculates whether the electronic component coordinate data is wrong, the horizontal coordinate data of the target is confirmed through the formula, and then it is determined whether the data of the DeepMask model is consistent.

[0064] Further, the camera and the camera head form an included angle, and then the electronic components in the electrical cabinet are identified, and the positions of the electronic components are analyzed again, so that the error of the coordinates is prevented, and the subsequent wiring effect is poor.

[0065] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above-mentioned embodiments, and the above-mentioned embodiments and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A deep learning-based intelligent wiring control system for a power cabinet robot, characterized by: The data acquisition module, the image processing module, the data processing module, the data storage module and the data control module; The data acquisition module supplements light and takes pictures of the inside of the electrical cabinet through the camera installed on the mechanical arm; The image processing module is used for processing the image data taken by the data acquisition module. When the mechanical arm moves, the data acquisition module takes pictures of the inside of the electrical cabinet through the camera, and then transmits the taken image data to the image processing module. The image processing module processes the image by denoising, enhancing and background segmentation. At the same time, the image processing module identifies the position of the recognized electronic components. The image processing module identifies the boundary position of the electrical cabinet to obtain the two-dimensional plane coordinate system of the electrical cabinet. A DeepMask model is set in the image processing module. The DeepMask model extracts the gray scale, color, brightness and texture of the electronic components in the electrical cabinet, and performs image segmentation on the electronic components using the extracted features. A temporary storage unit is set in the DeepMask model. The temporary storage unit is used to store electronic component model data. After the DeepMask model extracts the gray scale, color, brightness and texture data, the DeepMask model retrieves the data in the temporary storage unit for matching. The model data that matches successfully is modeled according to the coordinate position of the electronic components recognized by the image processing module, and the modeled result data is transmitted to the data processing module for processing. The data processing module is used for acquiring the installation data of the electronic components to plan the mechanical arm and generate corresponding instruction data. The data storage module is used for storing the feature data of the image processing module and the instruction data generated by the data processing module. The data control module is used for controlling the mechanical arm to process the circuit in the electrical cabinet. The DeepMask model obtains the loss function of the electronic components through a formula, and the specific formula is as follows: wherein, represents an electronic component loss function, represents a weight coefficient of the loss function, and the weight coefficient is affected by the sharpness of image shooting, represents segmentation data in a DeepMask model, represents pixel data of an electronic component, represents segmentation data about the electronic component in pixel data, represents a constant, and the constant is approximately 2.71828, represents electronic component prediction data in an actual state, represents segmentation result data, represents prediction result data.

2. The intelligent wiring control system for a robot for electrical cabinets based on deep learning according to claim 1, characterized in that: The image data of the electronic components at different angles in the image is obtained through the movement of the mechanical arm, and the image data of the electronic components is continuously optimized using the image processing module and the loss function.

3. The intelligent wiring control system for a robot based on deep learning for electrical cabinets according to claim 2, characterized in that: A model matching unit is set in the image processing module. The model matching unit is used to retrieve the modeling data in the temporary storage unit, and to obtain the electrical cabinet model data constructed in the DeepMask model, and to match the data in the DeepMask model with the data in the temporary storage unit.

4. The intelligent wiring control system for a robot based on deep learning for electrical cabinets according to claim 3, characterized in that: When the electrical cabinet model data constructed in the DeepMask model is consistent with the data in the temporary storage unit, the model matching unit generates a retrieval instruction and transmits it to the data processing module. The data processing module retrieves the instruction data in the data storage module, and optimizes the movement path of the mechanical arm according to the circuit in the image data; When the electrical cabinet model data built in the DeepMask model is inconsistent with the data in the temporary storage unit, the data processing module re-plans the mechanical arm movement path, and simultaneously obtains the circuit route in the electrical cabinet through the data processing module, and then adjusts the height and angle of the mechanical arm during movement, so that the mechanical arm bypasses the line blocking part, and updates the model data in the temporary storage unit and eliminates the original model data.

5. The intelligent wiring control system for a robot based on deep learning for electrical cabinets according to claim 4, characterized in that: The camera is fixedly installed on the mechanical arm base and is used for shooting the electrical cabinet, the camera and the camera cooperate with each other, then the electronic component coordinate data under different angles is obtained through cooperation, and the electronic component coordinate data is obtained through a formula, and the specific formula is as follows: Wherein, X represents the horizontal coordinate of the electronic component coordinate data, , Y represents the camera coordinate, , Z represents the camera coordinate, The angle between the line from the camera shooting point to the target point and the horizontal direction is represented by alpha, The angle between the line from the camera shooting point to the target point and the horizontal direction is represented by alpha.

6. The intelligent wiring control system for a robot based on deep learning for electrical cabinets according to claim 5, characterized in that: The data processing module matches the calculated abscissa data with the abscissa data calculated by the DeepMask model, when the abscissa data are consistent, the data processing module enters the next step, when the abscissa data deviate, the data processing module will brings in The electronic component coordinate data is re-verified.

7. The intelligent wiring control system for a robot based on deep learning for electrical cabinets according to claim 1, characterized in that: After receiving the instruction data, the data control module controls the mechanical arm to process the circuit connection in the electrical cabinet, simultaneously shoots the working process of the mechanical arm through the camera, and transmits the working process image to the DeepMask model in real time, and evaluates the connection quality of the mechanical arm by using the DeepMask model.

Images