Marine narrow cabin automatic meter reading robot and control method thereof

Abstract

The invention provides a marine narrow cabin automatic meter reading robot and a control method thereof, and relates to the technical field of robot control. The robot comprises a shipborne robot control terminal installed in a cabin body; the control terminal cable is mounted on the inner wall of the cabin body; the relay station is mounted at the top, close to the wall, in the cabin body; the automatic meter reading robot identifies the terminal; the five-axis mechanical arm is used for adjusting the spatial position of the automatic meter reading robot identification terminal; the transmission cable is used for supplying power to the five-axis mechanical arm and transmitting data; the V-shaped guide rail is mounted at the top in the cabin body and is used for moving the five-axis mechanical arm; and the electric wheel module is used for fixing the five-axis mechanical arm on the guide rail and moving to change the position of the five-axis mechanical arm. According to the marine narrow cabin automatic meter reading robot, the meter reading efficiency and accuracy are improved, space occupation caused by arrangement of too many devices in a narrow cabin is avoided, and the whole meter reading system and the cabin environment are highly integrated and do not interfere with each other.

Description

Technical Field

[0001] This invention relates to the field of robot control technology, specifically to an automatic meter reading robot for narrow marine cabins and its control method. Background Technology

[0002] Existing marine meter reading methods have numerous limitations in confined space environments. Currently, conventional marine meter reading methods mainly fall into three categories: manual meter reading, semi-automatic meter reading, and meter reading using simple mechanical devices in certain areas. The drawbacks of traditional meter reading methods are becoming increasingly apparent. Current meter reading activities largely rely on dedicated personnel periodically entering confined spaces for manual reading. This method not only consumes significant manpower and time costs but is also inefficient, and the accuracy and timeliness of data are difficult to guarantee, failing to meet the demands of efficient operation and maintenance in modern ships. Furthermore, given the special environment of confined spaces and the high value of the equipment, equipment damage and data loss occur frequently, posing safety hazards to ship operations. Therefore, there is an urgent need for an intelligent automatic meter reading robot capable of accurately collecting equipment data in confined spaces while ensuring the security of both equipment and data. Summary of the Invention

[0003] In view of the fact that the traditional meter reading method in the narrow compartments of ships in the existing technology mostly relies on a dedicated person to enter the narrow compartments at regular intervals to manually read the meters, this method consumes a lot of manpower and time costs, has low meter reading efficiency, and makes it difficult to guarantee the accuracy and timeliness of the data, which is difficult to meet the needs of efficient operation and maintenance of modern ships, this invention proposes an automatic meter reading robot for narrow compartments of ships and its control method.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is: an automatic meter reading robot for narrow marine cabins, the robot comprising: The control terminal includes: a shipborne robot control terminal, a control terminal cable, and a relay station; wherein, the shipborne robot control terminal is installed inside the cabin; the control terminal cable is installed on the interior wall of the cabin; and the relay station is installed at the top of the cabin near the wall. Automatic meter reading robot identification terminal; A five-axis robotic arm is used to adjust the spatial position of the automatic meter reading robot's identification terminal; A transmission cable is provided to power the five-axis robotic arm and for data transmission. V-shaped guide rails are installed on the top inside the cabin to move the five-axis robotic arm; The electric wheel module is used to fix the five-axis robotic arm on the guide rail and move it to change the position of the five-axis robotic arm; The automatic meter reading robot identification terminal is installed at the end of the five-axis robotic arm.

[0005] Furthermore, the automatic meter reading robot identification terminal includes: an inertial sensor, a high-precision identification camera, a lidar, a torque sensor, and a 4G module; the inertial sensor is used to detect the acceleration and rotation speed of an object, and calculate the object's current speed and position by integration based on the initial speed and position; the high-precision identification camera is used to identify data from the screen and dashboard in real time, and uses image processing technology to accurately identify, convert, and upload the read data; the lidar is used to emit laser beams and receive reflected signals to achieve high-precision ranging and imaging of objects inside the cabin, as well as to determine the movement of people within a certain range; the torque sensor is used to sense the interaction force between the automatic meter reading robot and the external environment in real time; the 4G module is used for remote communication and to transmit information from the automatic meter reading robot and the inertial sensor to a remote integrated control center.

[0006] Furthermore, the control terminal is used to transmit device initialization signals, control electrical signals, and electrical energy to the relay station through the control terminal cable; the relay station is responsible for voltage adjustment and transmits electrical energy and electrical signals to the automatic meter reading robot identification terminal, the transmission cable, and the five-axis robotic arm through the transmission cable. After the automatic meter reading robot identification terminal starts working, it uses LiDAR to model the cabin map space and objects, and inertial sensors to calculate the spatial position of the five-axis robotic arm. The five-axis robotic arm is used to drive the electric wheel module to move according to the spatial position and command information, while adjusting its own posture and the orientation of the end-effector automatic meter reading robot identification terminal.

[0007] Furthermore, the torque sensor remains operational throughout the initialization phase, monitoring the interaction forces between the automatic meter reading robot and the external environment in real time to prevent collisions. Based on the spatial and positional information of the five-axis robotic arm, the shipborne robot control terminal confirms that it has reached the designated position. The high-precision recognition camera then identifies the data on the screen and instrument panel in real time and uploads it to the terminal.

[0008] Furthermore, the control modes of the five-axis robotic arm include: teach control mode, remote control mode, and autonomous scanning mode; The teaching control mode and the autonomous scanning mode transmit instruction information, spatial information, location information and scanning data bidirectionally through control terminal cable, relay station and transmission cable; The remote control mode transmits command information, spatial information, location information, and scan data bidirectionally with the remote integrated control center via the 4G module; when the 4G network signal is lost, the above information will be transmitted to the shipborne robot control terminal; in both remote control mode and teaching control mode, the posture of the electric wheel module and the five-axis robotic arm can be manually controlled. In the autonomous scanning mode, the shipborne robot control terminal autonomously controls the posture of the electric wheel module and the five-axis robotic arm based on sensor information and task requirements.

[0009] On the other hand, this application also provides a control method for the aforementioned automatic meter reading robot in a narrow marine cabin, the method comprising: Step 1: Initialize the terminal system, electric wheel module, and five-axis robotic arm posture. After completion, proceed to Step 2. Step 2: The five-axis robotic arm carries a lidar to scan the cabin and create a map. After the modeling is completed, it returns to its initial position, activates the inertial sensor, marks the origin, and then proceeds to Step 3. Step 3: Select the control mode; after running, proceed to Step 4. Step 4: Select the teaching control mode to activate the torque sensor. After completion, proceed to step 51. Step 51: The system determines whether a control command has been received. If received, proceed to step 61; otherwise, return to step 4. Step 61: The five-axis robotic arm starts the electric wheel module to adjust its position according to the instruction, adjusts the posture of the five-axis robotic arm, and records the spatial position. After completion, proceed to step 71. Step 71: The torque sensor determines whether there is an abnormal interaction force. If so, obstacle avoidance is performed and the process returns to step 61. If not, proceed to step 81. Step 81: Determine whether to end control. If not, proceed to step 61; if yes, proceed to step 91. Step 91, end the process.

[0010] Furthermore, if the autonomous scanning mode is selected in step 4, then proceed to step 52; Step 52: Activate the torque sensor. After completion, proceed to step 62. Step 62: Search for the target based on the map and object location modeled by the LiDAR. After completion, proceed to step 72. Step 72: Determine whether the target has been found. If yes, proceed to step 82; otherwise, return to step 62. Step 82: Compare the target position with the current position, start the electric wheel module to adjust the position automatically, adjust the posture of the five-axis robotic arm automatically, and record the spatial position. After completion, proceed to step 92. Step 92: The torque sensor determines whether there is an abnormal interaction force. If so, obstacle avoidance is performed and the process returns to step 82. If not, proceed to step 102. Step 102: The shipborne robot control terminal locates itself based on the data from the lidar and inertial sensors, and calls the autofocus algorithm to find the optimal focus. After completion, proceed to step 112. Step 112: The shipborne robot control terminal starts the electric wheel module to adjust its position according to the optimal focus position, adjusts the posture of the five-axis robotic arm, stops the action after reaching the optimal focus position, and then proceeds to step 122. Step 122: The shipborne robot control terminal calls the visual recognition algorithm, and after completion, proceeds to step 132; Step 132: The high-precision recognition camera uses a visual recognition algorithm to identify the screen and dashboard, read data and upload it. After completion, proceed to step 142. Step 142: The terminal system determines whether the device data has been read and uploaded. If yes, proceed to step 152; otherwise, return to step 132. Step 152, end the process.

[0011] Furthermore, if the remote control mode is selected in step 4, then proceed to step 53; Step 53: Activate the torque sensor; after completion, proceed to step 63. Step 63: Activate the 4G module and establish a communication handshake with the remote integrated control center. After completion, proceed to step 73. Step 73: The remote integrated control center determines whether it has received the status information and instruction signal of the automatic meter reading robot. If yes, proceed to step 83; otherwise, return to step 63. Step 83: The remote integrated control center obtains information from the lidar and inertial sensor to obtain the spatial location information of the automatic meter reading robot. After completion, proceed to step 93. Step 93: The remote control center starts the electric wheel module of the remote arm to adjust its position and adjust the posture of the five-axis robotic arm. After completion, proceed to step 103. Step 103: The torque sensor determines whether there is an abnormal interaction force. If so, obstacle avoidance is performed and the process returns to step 93. If not, proceed to step 113. Step 113: The remote integrated control center performs positioning based on the data from the lidar and inertial sensor transmitted by the 4G module, and calls the autofocus algorithm to find the best focus. After completion, it proceeds to step 123. Step 123: The remote control center starts the electric wheel module to remotely adjust the position according to the optimal focus position, adjusts the posture of the five-axis robotic arm, stops the action after reaching the optimal focus position, and then proceeds to step 133. Step 133: The remote integrated control center calls the visual recognition algorithm. After completion, proceed to step 143. Step 143: The high-precision recognition camera uses a visual recognition algorithm to identify the screen and dashboard, reads the data and uploads it to the remote integrated control center via the 4G network. After completion, proceed to step 153. Step 153: The remote integrated control center receives and stores the data transmitted by the 4G network. After completion, proceed to step 163. Step 163: Determine whether the device data has been read and uploaded. If yes, proceed to step 173; otherwise, return to step 153. Step 173, end the process.

[0012] Furthermore, the autofocus algorithm includes the following steps: S1, Initialize the deep learning model, sensor, and focus parameters; S2, acquire the current image frame, IMU data, and LiDAR data; S3, which fuses sensor data to estimate relative position and attitude; S4, Determine the focused search area based on the estimated information; S5, input the image into the deep learning model to predict focus quality; S6, Scan the focus position within the search range according to the current step size; S7, acquire the new image and calculate the actual focus quality; S8 compares the actual and predicted focus quality and adjusts the step size; S9 records the optimal focus position; S10, determine whether the stopping condition is met; S11 outputs the final focus position and controls the lens focus; Specifically, S5 includes: The input image is , where H, W, and C are the height, width, and number of channels of the image, respectively; After feature extraction using a lightweight CNN model, feature maps are obtained. ; in, Let be the eigenvalue at coordinates (k, l) in the feature map; h, w, and c are the height, width, and number of channels of the feature map, respectively. This indicates that the feature map is a dimensional map. A real matrix; The feature map is transformed into a one-dimensional vector using a global average pooling layer. , represents a one-dimensional vector obtained by transforming the feature map through a global average pooling layer. The dimension is c. This indicates that the vector is a real vector of length c; The formula for global average pooling is shown below: in, A one-dimensional vector The One element; This represents the total number of pixels in the feature map. Indicates the first feature map All pixel values ​​in each channel Summing and then dividing by the total number of pixels yields the average value for that channel, which is then used as a vector. The One element; Finally, the focus quality prediction value is output through a fully connected layer and a Sigmoid activation function. , is represented as: in, This represents the predicted focus quality value, ranging from [0,1]. The closer the value is to 1, the sharper the focus of the current image. The Sigmoid activation function maps the output of the fully connected layer to the [0,1] interval. Its calculation formula is: W represents the weight parameters of the fully connected layer, with dimensions of . Used for feature vectors Perform a linear transformation; v is a one-dimensional feature vector obtained by global average pooling; b is the bias parameter of the fully connected layer, used to adjust the baseline offset of the linear transformation.

[0013] Furthermore, the visual recognition algorithm includes the following steps: S1, Image Acquisition: Obtaining raw frames via a high-precision recognition camera. The frame rate is set to 30 FPS; S2, Image Preprocessing: Processing the original frame Perform Gaussian filtering Histogram equalization (H) yields enhanced frames. : S3, Object Detection: Enhanced Frames As input, a set of candidate boxes is obtained through a lightweight YOLO network. : ; Each candidate box It contains information about candidate target regions identified in target detection; express The center coordinates in the image coordinate system are used to determine the position of the candidate box in the image; Candidate boxes The width and height are used to describe the size of the candidate box; The logistic sigmoid function used to calculate the confidence score of candidate boxes converts the raw scores output by the network. Mapping to the [0,1] interval yields the probability that the candidate box belongs to the true target; is the activation function for sigmoid; This refers to the raw score output by the object detection network for the i-th candidate box, before it has been mapped by the Sigmoid function. Its numerical range is the real number domain. The confidence score is obtained after processing by the Sigmoid function. ; Non-maximum suppression of NMS is employed: in, This refers to nonmaximum suppression. After the operation, the remaining set of candidate boxes are the areas that can accurately correspond to the real target after filtering. This is a function for non-maximum suppression operations. For multiple candidate boxes containing the same target, the cross-union ratio (CUI) is calculated. If the CUI is greater than a set threshold, the candidate box is considered for suppression. If the confidence level is high, then the candidate boxes with high confidence are retained, and the candidate boxes with low confidence are deleted. S4, ROI perspective transformation: If a target bounding box is detected... Extract the region of interest (ROI) containing the target, and use... This indicates that for ROIs with skew, perspective transformation is needed for correction, and the homography matrix is ​​obtained through four-point mapping. in, It is a homography matrix used to realize perspective transformation of two-dimensional image planes. It can convert tilted ROIs into front views according to the set mapping rules, so that text areas are straightened. and A matrix constructed for the original four-point coordinates and the expected corrected four-point coordinates in the ROI; This is a front view obtained after perspective transformation and correction; the target area in the image is horizontal or upright. This represents the perspective transformation function, with the original tilted ROI image as input. Homography matrix The output is a front view after perspective transformation correction. ; S5, OCR recognition: Correcting the image Input OCR engine Obtain the character sequence: in This represents an optical character recognition (OCR) engine, which receives the corrected image. Output the identified characters and related confidence information; S6, Result Analysis and Output: (For...) Perform regularization parsing Output JSON structure: in, This indicates a regularization parsing operation; This represents the confidence threshold, which is used to filter the identification results during the regularization parsing process.

[0014] Compared with the prior art, the beneficial effects of this application are: 1. This application proposes an automated meter reading robot specifically designed for the confined environments of marine cabins. A guide rail, running from bow to stern within the narrow cabin, serves as the robot's sole path of movement. This guide rail is expandable or reconfigurable to adapt to different cabin layouts, enhancing system flexibility. This design makes full use of the limited cabin space, avoiding the space-consuming problems associated with arranging complex tracks or supports in confined environments. The robot moves along the guide rail, flexibly reaching various meter reading locations without interfering with other equipment and personnel within the cabin.

[0015] 2. The automatic meter reading robot proposed in this application is equipped with a five-degree-of-freedom robotic arm. With its flexible multi-axis motion capabilities, it can cover the entire narrow cabin space, easily reaching various instruments and equipment to achieve comprehensive data collection. This design not only improves the efficiency and accuracy of meter reading but also avoids the space encroachment caused by placing too much equipment in the narrow cabin, allowing the entire meter reading system to be highly integrated with the cabin environment without interference.

[0016] 3. This application employs a combination of LiDAR, an inertial measurement unit (IMU), and an autofocus algorithm. The LiDAR provides high-precision 3D point clouds, while the IMU compensates for motion jitter, enabling accurate pose estimation (position + attitude) of the robot within the cabin. The autofocus algorithm adjusts the LiDAR or camera parameters based on the target distance, improving the detection accuracy of distant or small targets.

[0017] 4. The technical solution of this application combines autofocus algorithm and visual recognition algorithm. Through multi-sensor fusion (LiDAR, IMU, camera) and deep learning technology, it realizes precise focusing, target detection and data reading of the robot in narrow cabin environment. This enables the robot to operate efficiently and accurately in narrow cabin environment, significantly improves the automation level of narrow cabin inspection, reduces labor costs, and is suitable for narrow space or high-precision tasks in industrial, military, and shipbuilding fields.

[0018] Other features and advantages of the embodiments of this application will be described in detail in the following detailed description section. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the structure of the automatic meter reading robot for narrow cabins on ships according to the present invention; Figure 2 This is a working structure diagram of the automatic meter reading robot for narrow marine cabins according to the present invention; Figure 3 This is a flowchart of the automatic meter reading robot for narrow marine cabins according to the present invention. Detailed Implementation

[0020] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings, so as to more clearly understand the purpose, features and advantages of this invention. It should be understood that the embodiments shown in the drawings are not intended to limit the scope of this invention, but are only for illustrating the essential spirit of the technical solutions of this invention. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this invention.

[0021] Unless the context requires otherwise, throughout the specification and claims, the word “comprising” and its variations, such as “including” and “having”, shall be understood to have an open, inclusive meaning, that is, to be interpreted as “including, but not limited to”.

[0022] Throughout this specification, references to "an embodiment" or "an embodiment" indicate that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. Therefore, the appearance of "in an embodiment" or "an embodiment" in various places throughout the specification does not necessarily refer to the same embodiment. Furthermore, a particular feature, structure, or characteristic may be combined in any manner in one or more embodiments.

[0023] The singular forms “a” and “the” used in this specification and the appended claims include plural references unless otherwise expressly stated herein. It should be noted that the term “or” is generally used to mean “and / or” unless otherwise expressly stated herein.

[0024] In the following description, in order to clearly demonstrate the structure and working method of the present invention, a number of directional terms will be used. However, terms such as "front", "back", "left", "right", "outside", "inside", "outward", "inward", "up", and "down" should be understood as convenient terms and not as limiting terms.

[0025] The implementation details of the embodiments of the present invention will be described in detail below with reference to the accompanying drawings. The following content is only for the convenience of understanding the implementation details and is not necessary for implementing the technical solution.

[0026] Combination Figure 1 A schematic diagram of the structure of the automatic meter reading robot in a narrow marine cabin. A2 is the shipborne robot control terminal, installed inside cabin A1, used for power distribution, control, and processing and storage of sensor data for the automatic meter reading robot's robotic arm; A3 is the control terminal cable, installed on the inner wall of cabin A1; A4 is the relay station, installed at the top of cabin A1 against the wall, used for voltage conversion and data transmission.

[0027] Continue to refer to Figure 1 B represents the robotic arm and related equipment of the automatic meter reading robot. B1 is the automatic meter reading robot's identification terminal, which includes: an inertial sensor B1-1, a high-precision identification camera B1-2, a lidar B1-3, a torque sensor B1-4, and a 4G module B1-5. The inertial sensor B1-1 detects the acceleration and rotational speed of an object, combines this with its initial velocity and position, and calculates the object's current velocity and position using integration. The high-precision identification camera B1-2 identifies data from the screen and dashboard in real time, using advanced image processing technology to accurately identify, convert, and upload the data. The lidar B1-3 emits a laser beam and receives reflected signals to achieve high-precision ranging and imaging of objects within the cabin, and can also determine the movement of people within a certain range. The torque sensor B1-4 senses the interaction forces between the automatic meter reading robot and the external environment in real time, enabling precise control and obstacle avoidance. The 4G module B1-5 enables remote communication and transmits information from the automatic meter reading robot and sensors to a remote control center.

[0028] B2 is a transmission cable used to power the automatic meter reading robot's robotic arm and transmit data. B3 is a V-shaped guide rail, installed on the top of the interior of compartment A1, used to move the automatic meter reading robot's robotic arm within compartment A1. B4 is an electric wheel module used to fix the automatic meter reading robot's robotic arm to the guide rail and to move and change the position of the robotic arm. B5 is a five-axis robotic arm used to adjust the spatial position of the identification terminal. The automatic meter reading robot's identification terminal B1 is installed at the end of the five-axis robotic arm B5.

[0029] The aforementioned automated meter reading robot achieves efficient movement, precise positioning, and autonomous operation within narrow cabins through the coordinated use of a guide rail layout on the top of the ship's cabin, integrated cables, multi-axis robotic arms, and electric wheels. It also takes into account safety, flexibility, and maintainability, making it significantly superior to traditional manual meter reading or fixed sensor solutions. It is especially suitable for scenarios with limited space and complex environments, such as power compartments and ship cabins.

[0030] In some embodiments, combined with Figure 2 The diagram illustrates the working structure of an automated meter reading robot for narrow marine cabins. It should be noted that, depending on the location and functional requirements of the equipment within the cabin, the sensors and data transmission modules mounted on the automated meter reading robot may be adjusted during actual installation. All electrical components of this application are installed in the available space within the cabin, without affecting the operator's task performance or normal movement.

[0031] like Figure 2 As shown, the working structure of the marine automatic meter reading robot in a narrow cabin includes: P1 is the control terminal, placed against the wall inside the cabin. Control terminal P1 includes shipborne robot control terminal A2, control terminal cable A3, and relay station A4.

[0032] P2 is the mobile part, installed at the top of the cabin, and includes V-shaped guide rail B3 and electric wheel module B4.

[0033] P3 is the main body of the automatic meter reading robot, mounted on the V-shaped guide rail B3. The main body P3 includes the automatic meter reading robot identification terminal B1, the transmission cable B2, and the five-axis robotic arm B5.

[0034] The control terminal P1 transmits device initialization signals, control signals, and electrical energy to the relay station A4 via control terminal cable A3. Relay station A4 handles voltage adjustment and transmits electrical energy and signals to the main body P3 of the automatic meter reading robot via transmission cable B2. The automatic meter reading robot identification terminal B1 begins operation, using lidar B1-3 to model the cabin space and objects, and inertial sensor B1-1 to calculate the spatial position of the automatic meter reading robot's robotic arm. The five-axis robotic arm B5 drives the electric wheel module B4 to move according to the spatial position and command information, while simultaneously adjusting its own posture and the orientation of the end effector automatic meter reading robot identification terminal B1.

[0035] It is important to note that the torque sensor B1-4 remains operational throughout the initialization phase, continuously monitoring the interaction forces between the automatic meter reading robot and its external environment to prevent collisions. Based on the robot arm's spatial and positional information, the shipborne robot control terminal A2 confirms that it has reached the designated location. Simultaneously, the high-precision recognition camera B1-2 identifies data from the screen and dashboard in real time and uploads it to the terminal.

[0036] In some embodiments, this application also provides three control modes for the automatic meter reading robot arm: teach-in control, remote control, and autonomous scanning. The three control modes differ in the transmission of signals and data.

[0037] The teaching control mode and the autonomous scanning mode transmit instruction information, spatial information, location information and scanning data bidirectionally through the control terminal cable A3, the relay station A4 and the transmission cable B2.

[0038] In remote control mode, the system transmits command information, spatial information, location information, and scan data bidirectionally with the remote integrated control center via a 4G module. When the 4G network signal is unavailable, the above information will be transmitted to the shipborne robot control terminal A2. In both remote control and teach-control modes, the posture of the electric wheel module B4 and the five-axis robotic arm B5 can be manually controlled.

[0039] In autonomous scanning mode, the shipborne robot control terminal A2 autonomously controls the posture of the electric wheel module B4 and the five-axis robotic arm B5 based on sensor information and task requirements.

[0040] Next, based on the aforementioned automated meter reading robot for narrow marine cabins, such as Figure 3 As shown, this application also provides a flowchart of the automatic meter reading robot for narrow shipboard compartments. To demonstrate its functionality, its operation is described here with a fully modular setup, which includes the aforementioned teaching control, remote control, and autonomous scanning mode. The steps are as follows: Step 1: Initialize the terminal system, electric wheel module, and robot arm posture. After completion, proceed to Step 2. Step 2: The robotic arm carries a lidar to scan the cabin and create a map. After the modeling is completed, it returns to its initial position, activates the inertial sensor, marks the origin, and then proceeds to Step 3. Step 3: Select the control mode. After running, proceed to Step 4 (Steps after Step 4 are at the same level and are independent of each other). Step 4: If you select the teach control mode, the torque sensor will be activated, and you will proceed to step 51 after completion; if you select the autonomous scanning mode, you will proceed to step 52; if you select the remote control mode, you will proceed to step 53. Step 51: The system determines whether a control command has been received. If received, proceed to step 61; otherwise, return to step 4. Step 61: The robotic arm activates the electric wheel module according to the command to adjust its position, adjust the robotic arm's posture, and record the spatial position. After completion, proceed to step 71; Step 71: The torque sensor determines whether there is an abnormal interaction force. If so, obstacle avoidance is performed and the process returns to step 61. If not, proceed to step 81. Step 81: Determine whether to end control. If not, proceed to step 61; if yes, proceed to step 91. Step 91, end the process.

[0041] Step 52: Activate the torque sensor. After completion, proceed to step 62. Step 62: Search for the target based on the map and object location modeled by the LiDAR. After completion, proceed to step 72. Step 72: Determine whether the target has been found. If yes, proceed to step 82; otherwise, return to step 62. Step 82: Compare the target position with the current position, activate the electric wheel module to adjust the position automatically, adjust the robot arm posture automatically, and record the spatial position. After completion, proceed to step 92. Step 92: The torque sensor determines whether there is an abnormal interaction force. If so, obstacle avoidance is performed and the process returns to step 82. If not, proceed to step 102. Step 102: The shipborne robot control terminal performs positioning based on data from the LiDAR and inertial sensors, and calls the autofocus algorithm to find the optimal focus. After completion, proceed to step 112. Step 112: The shipborne robot control terminal activates the electric wheel module to automatically adjust its position based on the optimal focal position, adjusting the robot arm's posture until it reaches the optimal focal position and then stops. After completion, proceed to step 122. Step 122: The shipborne robot control terminal calls the visual recognition algorithm, and after completion, proceeds to step 132; Step 132: The high-precision recognition camera uses a visual recognition algorithm to identify the screen and dashboard, read data and upload it. After completion, proceed to step 142. Step 142: The terminal system determines whether the device data has been read and uploaded. If yes, proceed to step 152; otherwise, return to step 132.

[0042] Step 152, end the process.

[0043] Step 53: Activate the torque sensor; after completion, proceed to step 63. Step 63: Activate the 4G module and establish a communication handshake with the remote integrated control center. After completion, proceed to step 73. Step 73: The remote control center determines whether it has received robot device status information and command signals. If yes, proceed to step 83; otherwise, return to step 63. Step 83: The remote control center acquires information from the lidar and inertial sensors to obtain the robot's spatial position information. After completion, proceed to step 93. Step 93: The remote control center starts the electric wheel module of the remote arm to adjust its position and adjust the robot arm's posture. After completion, proceed to step 103; Step 103: The torque sensor determines whether there is an abnormal interaction force. If so, obstacle avoidance is performed and the process returns to step 93. If not, proceed to step 113. Step 113: The remote integrated control center performs positioning based on the data from the lidar and inertial sensor transmitted by the 4G module, and calls the autofocus algorithm to find the best focus. After completion, it proceeds to step 123. Step 123: The remote control center activates the electric wheel module to remotely adjust the position based on the optimal focal position, adjusting the robot arm's posture until the optimal focal position is reached, at which point the movement stops. After completion, proceed to step 133. Step 133: The remote integrated control center calls the visual recognition algorithm. After completion, proceed to step 143. Step 143: The high-precision recognition camera uses a visual recognition algorithm to identify the screen and dashboard, reads the data and uploads it to the remote integrated control center via the 4G network. After completion, proceed to step 153. Step 153: The remote integrated control center receives and stores the data transmitted by the 4G network. After completion, proceed to step 163. Step 163: Determine whether the device data has been read and uploaded. If yes, proceed to step 173; otherwise, return to step 153. Step 173, end the process.

[0044] After system initialization, the control process of the aforementioned automatic meter reading robot constructs a cabin map and marks the origin point based on lidar and inertial sensors, providing a spatial reference for subsequent tasks. It supports three independent modes: teaching control, autonomous scanning, and remote control. These modes are respectively implemented through torque sensor interactive teaching control, guided autonomous scanning, and remote control via a 4G module, covering all scenarios from manual guidance to fully autonomous operation. The robot integrates lidar positioning, inertial sensor attitude compensation, and autofocus algorithms to ensure that the robotic arm is precisely adjusted to the optimal observation position. Combined with visual recognition algorithms, it achieves high-precision reading and uploading of dashboard data. At the same time, the torque sensor monitors abnormal interaction forces in real time, triggering obstacle avoidance mechanisms to ensure safety. Ultimately, this significantly improves the efficiency and accuracy of tasks such as cabin inspection and equipment monitoring, while reducing the risk of human intervention.

[0045] During the operation of the aforementioned automatic meter reading robot, a high-performance visual recognition algorithm is employed. The high-precision recognition camera B1-2 identifies data from the screen and dashboard, and uploads the data after reading. This application also features a high-performance automatic focusing algorithm. Relying on the inertial sensor B1-1, the high-precision recognition camera B1-2, and the torque sensor B1-4, the algorithm automatically calculates the optimal focus position and drives the electric wheel module B4 to change position, adjusting the posture of the five-axis robotic arm B5 to reach the optimal focus position.

[0046] In some embodiments, the autofocus algorithm includes the following steps: Step 1: Initialize the deep learning model, sensor, and focusing parameters; Step 2: Acquire the current image frame, IMU data, and LiDAR data; Step 3, fuse sensor data to estimate relative position and attitude: use Kalman filtering to fuse the IMU data and lidar data to estimate the relative position and attitude of the automatic meter reading robot and the target device; Step 4: Determine the focusing search range based on the estimated information: Determine the starting position range of the focusing search based on the estimated relative position and attitude; Step 5, Input the image into the deep learning model to predict focus quality: Input the current image frame into the lightweight CNN model to obtain the focus quality prediction value; Step 6: Scan the focus position within the search range at the current step size; Step 7: Acquire the new image and calculate the actual focus quality; Step 8: Compare the actual and predicted focus quality and adjust the step size accordingly; Step 9: Record the optimal focus position; Step 10: Determine if the stopping condition is met; Step 11: Output the final focus position and control the lens focus.

[0047] Further, in the above step 1, initialize the deep learning model, sensors and focusing parameters, including: loading a pre-trained lightweight CNN model; initializing the IMU and lidar data receivers; setting the initial focusing position, maximum search range, initial step size and threshold.

[0048] Further, in the above step 5, input the image into the deep learning model to predict the focus quality: input the current image frame into the lightweight CNN model to obtain the focus quality prediction value, specifically including: Let the input image be , where H, W, and C are the height, width, and number of channels of the image, respectively.

[0049] After extracting features through the lightweight CNN model, the feature map is obtained. Among them, is the feature value at the coordinate (k, l) in the feature map; h, w, and c are the height, width, and number of channels of the feature map (usually h < H, w < W are the results of dimensionality reduction after convolution operations, indicates that this feature map is a real number matrix with a dimension of .) Convert the feature map into a one-dimensional vector through the global average pooling layer , indicating the one-dimensional vector obtained by converting the feature map through the global average pooling layer. is the number of channels of the feature map, so the dimension of the vector v is c, indicating that this vector is a real number vector with a length of c.

[0050] The calculation formula of global average pooling is as follows: Among them, is the k-th element of the one-dimensional vector v; is the total number of pixels of the feature map; represents the sum of all pixel values (coordinate (i, j)) in the k-th channel of the feature map, and then divided by the total number of pixels to obtain the average value of this channel, which is used as the k-th element of the vector v.

[0051] Finally, output the focus quality prediction value through the fully connected layer and the Sigmoid activation function, which is expressed as: Among them, represents the focus quality prediction value, with a range of [0, 1]. The closer the value is to 1, the clearer the focus of the current image. is the Sigmoid activation function, whose role is to map the output of the fully connected layer to the interval [0, 1]. The calculation formula is: W represents the weight parameters of the fully connected layer, with dimensions of . (c is the dimension of the feature vector v), used to perform a linear transformation on the feature vector v. v is a one-dimensional feature vector obtained by global average pooling; b is the bias parameter of the fully connected layer, used to adjust the baseline offset of the linear transformation.

[0052] In some embodiments, the above visual recognition algorithm includes the following steps: S1, Image Acquisition: Obtaining raw frames via a high-precision recognition camera. The frame rate is set to 30 FPS. S2, Image Preprocessing: Processing the original frame Perform Gaussian filtering Histogram equalization Get enhanced frames : S3, Object Detection: Enhanced Frames As input, a set of candidate boxes is obtained through a lightweight YOLO network. : ; Each candidate box It contains information about regions that may be targets identified during object detection.

[0053] express The center coordinates in the image coordinate system are used to determine the position of the candidate box in the image. Candidate boxes The width and height are used to describe the size of the candidate box. The sigmoid function used to calculate the confidence score of candidate boxes converts the raw scores output by the network. Mapping to the [0,1] interval yields the probability that the candidate box belongs to the true target. This is the activation function for the Sigmoid function. This refers to the raw score output by the object detection network for the i-th candidate box, before it has been mapped by the Sigmoid function. Its numerical range is generally the real number domain. The confidence score is obtained after processing by the Sigmoid function. .

[0054] Non-maximum suppression of NMS is employed: in, This refers to nonmaximum suppression ( After the operation, the final set of candidate boxes is retained. These candidate boxes are the areas that are most likely to accurately correspond to the real target after filtering. Non-maximum suppression (NMS) is a function representation of the non-maximum suppression operation and a key operation in the post-processing stage of object detection. The basic idea is: for multiple candidate boxes that may exist for the same object, calculate their intersection-union ratio (IUR). If the IUR is greater than a set threshold... We retain candidate boxes with high confidence and delete candidate boxes with low confidence.

[0055] S4, ROI perspective transformation: If a target bounding box is detected... Extract the region of interest (ROI) containing the target, using... This indicates that for ROIs with skew, perspective transformation is needed for correction, and the homography matrix is ​​obtained through four-point mapping. in, It is a homography matrix, such as a 3×3 matrix, which is used to realize the perspective transformation of a two-dimensional image plane. It can convert a tilted ROI into a front view according to the set mapping rules, so that the text area is straightened. and A matrix is ​​constructed for the "original four-point coordinates" and the "expected corrected four-point coordinates" in the ROI. This is the front view obtained after perspective transformation correction. The target area (such as text) in the image is horizontal / upright, which can improve OCR recognition performance. W' represents the perspective transformation function, and the input is the original tilted ROI image. Homography matrix The output is a front view after perspective transformation correction. .

[0056] S5, OCR recognition: Correcting the image Input OCR engine Obtain the character sequence: in OCR (Optical Character Recognition) engines are algorithms or tools that perform character recognition tasks; they receive corrected images. It outputs the recognized characters and related confidence information.

[0057] S6, Result Analysis and Output: (For...) Perform regularization parsing Output JSON structure: in, Regularization parsing is a process of processing the character sequence T obtained by OCR according to specific rules (such as regular expressions) to extract valid information and standardize the format for easier subsequent use. This represents the confidence threshold, used to filter recognition results during the regularization parsing process. For example, a character is considered valid and included in the final parsing output only if its confidence score is greater than or equal to 0.8; characters below this threshold will be filtered or marked to improve the reliability of the output results.

[0058] This application proposes an automated meter reading robot for narrow marine cabins and its control method. A guide rail, running from bow to stern, serves as the robot's sole path within the confined space. This design fully utilizes the limited cabin space, avoiding the space-consuming problems associated with complex tracks or supports in narrow environments. Moving along the guide rail, the robot can flexibly reach various meter reading locations without interfering with other equipment and personnel within the cabin. Simultaneously, the robot is equipped with a five-degree-of-freedom robotic arm, whose flexible multi-axis motion capabilities allow it to cover the entire cabin space, easily reaching various instruments and equipment for comprehensive data acquisition. Furthermore, this technical solution combines autofocus and visual recognition algorithms, employing multi-sensor fusion and deep learning technologies to achieve precise focusing, target detection, and data reading within the narrow cabin environment. This design not only improves the efficiency and accuracy of meter reading but also avoids the space encroachment caused by excessive equipment within the narrow cabin, ensuring the entire meter reading system is highly integrated with the cabin environment without interference.

[0059] Although the present invention has been described in detail with reference to the accompanying drawings and preferred embodiments, the invention is not limited thereto. Various equivalent modifications or substitutions can be made to the embodiments of the invention by those skilled in the art without departing from the spirit and essence of the invention. Such modifications or substitutions should all fall within the scope of the invention, or any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the invention should be covered within the protection scope of the invention. Therefore, the protection scope of the invention should be determined by the scope of the claims.

Claims

1. An automatic meter reading robot for narrow marine cabins, characterized in that, The robot includes: The control terminal includes: a shipborne robot control terminal, a control terminal cable, and a relay station; wherein, the shipborne robot control terminal is installed inside the cabin; the control terminal cable is installed on the interior wall of the cabin; and the relay station is installed at the top of the cabin near the wall. Automatic meter reading robot identification terminal; A five-axis robotic arm is used to adjust the spatial position of the automatic meter reading robot's identification terminal; A transmission cable is provided to power the five-axis robotic arm and for data transmission. V-shaped guide rails are installed on the top inside the cabin to move the five-axis robotic arm; The electric wheel module is used to fix the five-axis robotic arm on the guide rail and move it to change the position of the five-axis robotic arm; The automatic meter reading robot identification terminal is installed at the end of the five-axis robotic arm.

2. The robot according to claim 1, characterized in that, The automatic meter reading robot identification terminal includes: an inertial sensor, a high-precision identification camera, a lidar, a torque sensor, and a 4G module. The inertial sensor detects the acceleration and rotation speed of an object, combines the initial velocity and position, and calculates the object's current velocity and position using integration. The high-precision identification camera identifies data from the screen and dashboard in real time, and uses image processing technology to accurately identify, convert, and upload the read data. The lidar emits a laser beam and receives reflected signals to achieve high-precision ranging and imaging of objects inside the cabin, as well as to determine the movement of people within a certain range. The torque sensor senses the interaction force between the automatic meter reading robot and the external environment in real time. The 4G module enables remote communication and transmits information from the automatic meter reading robot and the inertial sensor to a remote integrated control center.

3. The robot according to claim 2, characterized in that, The control terminal is used to transmit device initialization signals, control electrical signals, and electrical energy to the relay station through the control terminal cable; the relay station is responsible for voltage adjustment and transmits electrical energy and electrical signals to the automatic meter reading robot identification terminal, the transmission cable, and the five-axis robotic arm through the transmission cable. After the automatic meter reading robot identification terminal starts working, it uses LiDAR to model the cabin map space and objects, and inertial sensors to calculate the spatial position of the five-axis robotic arm. The five-axis robotic arm is used to drive the electric wheel module to move according to the spatial position and command information, while adjusting its own posture and the orientation of the end-effector automatic meter reading robot identification terminal.

4. The robot according to claim 2, characterized in that, The torque sensor is in working condition throughout the initialization phase, monitoring the interaction force between the automatic meter reading robot and the external environment in real time to prevent collisions. Based on the spatial and positional information of the five-axis robotic arm, the shipborne robot control terminal confirms that it has reached the designated position. The high-precision recognition camera then identifies the data on the screen and instrument panel in real time and uploads it to the terminal.

5. The robot according to claim 1, characterized in that, The control modes of the five-axis robotic arm include: teach control mode, remote control mode, and autonomous scanning mode; The teaching control mode and the autonomous scanning mode transmit instruction information, spatial information, location information and scanning data bidirectionally through control terminal cable, relay station and transmission cable; The remote control mode transmits command information, spatial information, location information, and scan data bidirectionally with the remote integrated control center via the 4G module; when the 4G network signal is lost, the above information will be transmitted to the shipborne robot control terminal; in both remote control mode and teaching control mode, the posture of the electric wheel module and the five-axis robotic arm can be manually controlled. In the autonomous scanning mode, the shipborne robot control terminal autonomously controls the posture of the electric wheel module and the five-axis robotic arm based on sensor information and task requirements.

6. A control method for an automatic meter reading robot for a narrow marine cabin as described in any one of claims 1-5, characterized in that, The method includes: Step 1: Initialize the terminal system, electric wheel module, and five-axis robotic arm posture. After completion, proceed to Step 2. Step 2: The five-axis robotic arm carries a lidar to scan the cabin and create a map. After the modeling is completed, it returns to its initial position, activates the inertial sensor, marks the origin, and then proceeds to Step 3. Step 3: Select the control mode; after running, proceed to Step 4. Step 4: Select the teaching control mode to activate the torque sensor. After completion, proceed to step 51. Step 51: The system determines whether a control command has been received. If received, proceed to step 61; otherwise, return to step 4. Step 61: The five-axis robotic arm starts the electric wheel module to adjust its position according to the instruction, adjusts the posture of the five-axis robotic arm, and records the spatial position. After completion, proceed to step 71. Step 71: The torque sensor determines whether there is an abnormal interaction force. If so, obstacle avoidance is performed and the process returns to step 61. If not, proceed to step 81. Step 81: Determine whether to end control. If not, proceed to step 61; if yes, proceed to step 91. Step 91, end the process.

7. The method according to claim 6, characterized in that, If the autonomous scanning mode is selected in step 4, then proceed to step 52; Step 52: Activate the torque sensor. After completion, proceed to step 62. Step 62: Search for the target based on the map and object location modeled by the LiDAR. After completion, proceed to step 72. Step 72: Determine whether the target has been found. If yes, proceed to step 82; otherwise, return to step 62. Step 82: Compare the target position with the current position, start the electric wheel module to adjust the position automatically, adjust the posture of the five-axis robotic arm automatically, and record the spatial position. After completion, proceed to step 92. Step 92: The torque sensor determines whether there is an abnormal interaction force. If so, obstacle avoidance is performed and the process returns to step 82. If not, proceed to step 102. Step 102: The shipborne robot control terminal locates itself based on the data from the lidar and inertial sensors, and calls the autofocus algorithm to find the optimal focus. After completion, proceed to step 112. Step 112: The shipborne robot control terminal starts the electric wheel module to adjust its position according to the optimal focus position, adjusts the posture of the five-axis robotic arm, stops the action after reaching the optimal focus position, and then proceeds to step 122. Step 122: The shipborne robot control terminal calls the visual recognition algorithm, and after completion, proceeds to step 132; Step 132: The high-precision recognition camera uses a visual recognition algorithm to identify the screen and dashboard, read data and upload it. After completion, proceed to step 142. Step 142: The terminal system determines whether the device data has been read and uploaded. If yes, proceed to step 152; otherwise, return to step 132. Step 152, end the process.

8. The method according to claim 6, characterized in that, If the remote control mode is selected in step 4, then proceed to step 53; Step 53: Activate the torque sensor; after completion, proceed to step 63. Step 63: Activate the 4G module and establish a communication handshake with the remote integrated control center. After completion, proceed to step 73. Step 73: The remote integrated control center determines whether it has received the status information and instruction signal of the automatic meter reading robot. If yes, proceed to step 83; otherwise, return to step 63. Step 83: The remote integrated control center obtains information from the lidar and inertial sensor to obtain the spatial location information of the automatic meter reading robot. After completion, proceed to step 93. Step 93: The remote control center starts the electric wheel module of the remote arm to adjust its position and adjust the posture of the five-axis robotic arm. After completion, proceed to step 103. Step 103: The torque sensor determines whether there is an abnormal interaction force. If so, obstacle avoidance is performed and the process returns to step 93. If not, proceed to step 113. Step 113: The remote integrated control center performs positioning based on the data from the lidar and inertial sensor transmitted by the 4G module, and calls the autofocus algorithm to find the best focus. After completion, it proceeds to step 123. Step 123: The remote control center starts the electric wheel module to remotely adjust the position according to the optimal focus position, adjusts the posture of the five-axis robotic arm, stops the action after reaching the optimal focus position, and then proceeds to step 133. Step 133: The remote integrated control center calls the visual recognition algorithm. After completion, proceed to step 143. Step 143: The high-precision recognition camera uses a visual recognition algorithm to identify the screen and dashboard, reads the data and uploads it to the remote integrated control center via the 4G network. After completion, proceed to step 153. Step 153: The remote integrated control center receives and stores the data transmitted by the 4G network. After completion, proceed to step 163. Step 163: Determine whether the device data has been read and uploaded. If yes, proceed to step 173; otherwise, return to step 153. Step 173, end the process.

9. The method according to claim 7 or 8, characterized in that, The autofocus algorithm includes the following steps: S1, Initialize the deep learning model, sensor, and focus parameters; S2, acquire the current image frame, IMU data, and LiDAR data; S3, which fuses sensor data to estimate relative position and attitude; S4, Determine the focused search area based on the estimated information; S5, input the image into the deep learning model to predict focus quality; S6, Scan the focus position within the search range according to the current step size; S7, acquire the new image and calculate the actual focus quality; S8 compares the actual and predicted focus quality and adjusts the step size; S9 records the optimal focus position; S10, determine whether the stopping condition is met; S11 outputs the final focus position and controls the lens focus; Specifically, S5 includes: The input image is , where H, W, and C are the height, width, and number of channels of the image, respectively; After feature extraction using a lightweight CNN model, feature maps are obtained. ; in, Let be the eigenvalue at coordinates (k, l) in the feature map; h, w, and c are the height, width, and number of channels of the feature map, respectively. This indicates that the feature map is a dimensional map. A real matrix; The feature map is transformed into a one-dimensional vector using a global average pooling layer. , represents a one-dimensional vector obtained by transforming the feature map through a global average pooling layer. The dimension is c. This indicates that the vector is a real vector of length c; The formula for global average pooling is shown below: in, A one-dimensional vector The One element; This represents the total number of pixels in the feature map. Indicates the first feature map All pixel values ​​in each channel Summing and then dividing by the total number of pixels yields the average value for that channel, which is then used as a vector. The One element; Finally, the focus quality prediction value is output through a fully connected layer and a Sigmoid activation function. , is represented as: in, This represents the predicted focus quality value, ranging from [0,1]. The closer the value is to 1, the sharper the focus of the current image. The Sigmoid activation function maps the output of the fully connected layer to the [0,1] interval. Its calculation formula is: W represents the weight parameters of the fully connected layer, with dimensions of . Used for feature vectors Perform a linear transformation; v is a one-dimensional feature vector obtained by global average pooling; b is the bias parameter of the fully connected layer, used to adjust the baseline offset of the linear transformation.

10. The method according to claim 7 or 8, characterized in that, The visual recognition algorithm includes the following steps: S1, Image Acquisition: Obtaining raw frames via a high-precision recognition camera. The frame rate is set to 30 FPS; S2, Image Preprocessing: Processing the original frame Perform Gaussian filtering Histogram equalization (H) yields enhanced frames. : S3, Object Detection: Enhanced Frames As input, a set of candidate boxes is obtained through a lightweight YOLO network. : ； Each candidate box It contains information about candidate target regions identified in target detection; express The center coordinates in the image coordinate system are used to determine the position of the candidate box in the image; Candidate boxes The width and height are used to describe the size of the candidate box; The logistic sigmoid function used to calculate the confidence score of candidate boxes converts the raw scores output by the network. Mapping to the [0,1] interval yields the probability that the candidate box belongs to the true target; is the activation function for sigmoid; This refers to the raw score output by the object detection network for the i-th candidate box, before it has been mapped by the Sigmoid function. Its numerical range is the real number domain. The confidence score is obtained after processing by the Sigmoid function. ; Non-maximum suppression of NMS is employed: in, This refers to nonmaximum suppression. After the operation, the remaining set of candidate boxes are the areas that can accurately correspond to the real target after filtering. This is a function for non-maximum suppression operations. For multiple candidate boxes containing the same target, the cross-union ratio (CUI) is calculated. If the CUI is greater than a set threshold, the candidate box is considered for suppression. If the confidence level is high, then the candidate boxes with high confidence are retained, and the candidate boxes with low confidence are deleted. S4, ROI perspective transformation: If a target bounding box is detected... Extract the region of interest (ROI) containing the target, and use... This indicates that for ROIs with skew, perspective transformation is needed for correction, and the homography matrix is ​​obtained through four-point mapping. in, It is a homography matrix used to realize perspective transformation of two-dimensional image planes. It can convert tilted ROIs into front views according to the set mapping rules, so that text areas are straightened. and A matrix constructed for the original four-point coordinates and the expected corrected four-point coordinates in the ROI; This is a front view obtained after perspective transformation and correction; the target area in the image is horizontal or upright. This represents the perspective transformation function, with the original tilted ROI image as input. Homography matrix The output is a front view after perspective transformation correction. ; S5, OCR recognition: Correcting the image Input OCR engine Obtain the character sequence: in This represents an optical character recognition (OCR) engine, which receives the corrected image. Output the identified characters and related confidence information; S6, Result Analysis and Output: (For...) Perform regularization parsing Output JSON structure: in, This indicates a regularization parsing operation; This represents the confidence threshold, which is used to filter the identification results during the regularization parsing process.

Images