Inspection method, device and equipment for inspection robot and storage medium

Abstract

The application relates to the technical field of electric power, and provides a method and device for determining an inspection mode of an inspection robot, computer equipment, a storage medium and a computer program product. The application can improve the accuracy and efficiency of determining the inspection mode of the inspection robot. The method comprises the following steps: acquiring a current image of a target device photographed by an inspection robot in a substation, and a current photographing position and a current photographing posture of the inspection robot; identifying the current image to obtain an identification result of the current image; identifying the current photographing position and the current photographing posture according to the identification result, to obtain a target photographing position and a target photographing posture of the inspection robot; and determining the inspection mode of the inspection robot in the substation according to the target photographing position and the target photographing posture.

Description

Technical Field

[0001] This application relates to the field of power technology, and in particular to a method, apparatus, computer equipment, storage medium and computer program product for determining the inspection mode of an inspection robot. Background Technology

[0002] With the development of power technology, substations, as an important component of the power system, play a crucial role in regulating voltage, converting power types, and distributing power. To ensure the safe, stable, and efficient operation of substations, inspection robots are needed to acquire images of the equipment within the substation. Through this image data, potentially defective or faulty equipment can be detected in a timely manner, preventing serious accidents. Therefore, determining the inspection method for inspection robots has become an important research direction.

[0003] Traditional technology typically involves manually setting the inspection mode of inspection robots. However, this technology relies heavily on subjective judgment, resulting in low accuracy in determining the inspection mode of the inspection robot. Summary of the Invention

[0004] Based on this, it is necessary to provide a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for determining the inspection mode of an inspection robot, in order to address the above-mentioned technical problems.

[0005] Firstly, this application provides a method for determining the inspection mode of an inspection robot. The method includes:

[0006] Acquire the current image of the inspection robot taking pictures of the target equipment in the substation, as well as the current shooting position and current shooting posture of the inspection robot;

[0007] The current image is identified to obtain the identification result of the current image;

[0008] Based on the recognition results, the current shooting position and the current shooting posture are identified to obtain the target shooting position and target shooting posture of the inspection robot.

[0009] Based on the target shooting position and the target shooting posture, the inspection method of the inspection robot in the substation is determined.

[0010] In one embodiment, determining the inspection method of the inspection robot in the substation based on the target shooting position and the target shooting posture includes:

[0011] The inspection robot acquires target images of the target equipment captured in the substation; the target images correspond to the target shooting position and the target shooting posture.

[0012] The target image is identified to obtain the target recognition result of the target image;

[0013] If the target recognition result indicates that the target is approved, the target shooting position and the target shooting posture are identified as the inspection point information of the inspection robot.

[0014] Based on the inspection point information, the inspection method of the inspection robot in the substation is determined.

[0015] In one embodiment, determining the inspection method of the inspection robot in the substation based on the inspection point information includes:

[0016] Based on the inspection point information, the inspection robot's inspection route, inspection point location, and inspection point shooting posture in the substation are determined.

[0017] The inspection route, inspection point location, and inspection point shooting posture of the inspection robot in the substation are identified as the inspection mode of the inspection robot in the substation.

[0018] In one embodiment, the step of recognizing the current image to obtain the recognition result of the current image includes:

[0019] The target occlusion rate of the current image is identified to obtain a first identification result of the current image;

[0020] The target size is identified in the current image to obtain a second identification result for the current image;

[0021] The current image is subjected to shooting angle recognition to obtain a third recognition result for the current image;

[0022] Perform backlight shooting rate recognition on the current image to obtain a fourth recognition result for the current image;

[0023] The current image is subjected to illumination angle recognition to obtain the fifth recognition result of the current image;

[0024] The recognition result of the current image is determined based on the first recognition result, the second recognition result, the third recognition result, the fourth recognition result, and the fifth recognition result.

[0025] In one embodiment, the step of identifying the current shooting position and the current shooting posture based on the identification result to obtain the target shooting position and target shooting posture of the inspection robot includes:

[0026] The recognition result, the current shooting position, and the current shooting posture are input into the shooting prediction model to obtain the candidate shooting position, the probability value of the candidate shooting position, the candidate shooting posture, and the probability value of the candidate shooting posture of the inspection robot.

[0027] Based on the probability values ​​of the candidate shooting positions and the probability values ​​of the candidate shooting postures, the target shooting position of the inspection robot is selected from the candidate shooting positions, and the target shooting posture of the inspection robot is selected from the candidate shooting postures.

[0028] In one embodiment, obtaining the current shooting posture of the inspection robot includes:

[0029] Obtain the current shooting angle, current shooting focal length, and current shooting distance of the inspection robot;

[0030] The current shooting angle, the current shooting focal length, and the current shooting distance are all identified as the current shooting posture of the inspection robot.

[0031] Secondly, this application also provides a device for determining the inspection method of an inspection robot. The device includes:

[0032] The image acquisition module is used to acquire the current image of the inspection robot taking pictures of the target equipment in the substation, as well as the current shooting position and current shooting posture of the inspection robot.

[0033] An image recognition module is used to recognize the current image and obtain the recognition result of the current image;

[0034] The location recognition module is used to identify the current shooting position and the current shooting posture based on the recognition result, so as to obtain the target shooting position and target shooting posture of the inspection robot.

[0035] The method determination module is used to determine the inspection method of the inspection robot in the substation based on the target shooting position and the target shooting posture.

[0036] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0037] The system acquires current images of the inspection robot taking pictures of target equipment in the substation, as well as the current shooting position and shooting posture of the inspection robot; it identifies the current image to obtain the identification result; based on the identification result, it identifies the current shooting position and the current shooting posture to obtain the target shooting position and target shooting posture of the inspection robot; and based on the target shooting position and the target shooting posture, it determines the inspection method of the inspection robot in the substation.

[0038] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:

[0039] The system acquires current images of the inspection robot taking pictures of target equipment in the substation, as well as the current shooting position and shooting posture of the inspection robot; it identifies the current image to obtain the identification result; based on the identification result, it identifies the current shooting position and the current shooting posture to obtain the target shooting position and target shooting posture of the inspection robot; and based on the target shooting position and the target shooting posture, it determines the inspection method of the inspection robot in the substation.

[0040] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:

[0041] The system acquires current images of the inspection robot taking pictures of target equipment in the substation, as well as the current shooting position and shooting posture of the inspection robot; it identifies the current image to obtain the identification result; based on the identification result, it identifies the current shooting position and the current shooting posture to obtain the target shooting position and target shooting posture of the inspection robot; and based on the target shooting position and the target shooting posture, it determines the inspection method of the inspection robot in the substation.

[0042] The aforementioned inspection robot inspection mode determination method, device, computer equipment, storage medium, and computer program product acquire current images of the inspection robot taking pictures of target equipment in the substation, as well as the current shooting position and current shooting posture of the inspection robot; identify the current image to obtain the identification result of the current image; based on the identification result, identify the current shooting position and the current shooting posture to obtain the target shooting position and target shooting posture of the inspection robot; and determine the inspection mode of the inspection robot in the substation based on the target shooting position and the target shooting posture. This scheme acquires current images of the inspection robot taking pictures of target equipment in the substation, along with the robot's current shooting position and posture, thus obtaining the current image captured by the robot at the current shooting position and posture. The current image is then recognized to determine if it meets the specified criteria. Based on the recognition results, the current shooting position and posture are further identified to determine the robot's target shooting position and posture. These are then adjusted to obtain a more optimal target shooting position and posture. Finally, based on the target shooting position and posture, the inspection method of the inspection robot in the substation is determined, thereby improving the accuracy and efficiency of determining the inspection method. Attached Figure Description

[0043] Figure 1 This is a flowchart illustrating the method for determining the inspection mode of an inspection robot in one embodiment.

[0044] Figure 2 This is a flowchart illustrating the method for determining the inspection mode of the inspection robot in another embodiment;

[0045] Figure 3 This is a logical schematic diagram of the Q-learning algorithm in one embodiment;

[0046] Figure 4 This is a schematic diagram illustrating the experimental results of different Q-learning algorithms in one embodiment;

[0047] Figure 5 This is a structural block diagram of a device for determining the inspection method of an inspection robot in one embodiment;

[0048] Figure 6 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0050] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data shall comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0051] In one embodiment, such as Figure 1 As shown, a method for determining the inspection mode of an inspection robot is provided. This embodiment illustrates the application of this method to a terminal, including the following steps:

[0052] Step S101: Obtain the current image of the inspection robot taking pictures of the target equipment in the substation, as well as the current shooting position and current shooting posture of the inspection robot.

[0053] In this step, the inspection robot can be located within the substation; the target equipment can be the substation's electrical equipment; the current image can be a picture; the current shooting position can be the location where the inspection robot is when shooting the current image; and the current shooting posture can be the shooting posture corresponding to the inspection robot when shooting the current image.

[0054] Specifically, the terminal acquires the current image of the inspection robot taking pictures of the target equipment in the substation, as well as the current shooting position and shooting posture of the inspection robot when taking the current image.

[0055] Step S102: Recognize the current image to obtain the recognition result of the current image.

[0056] In this step, the recognition result can be a result related to the image quality of the current image (such as the image capture quality).

[0057] Specifically, the terminal performs image quality recognition on the current image to obtain the recognition result of the current image.

[0058] Step S103: Based on the recognition results, identify the current shooting position and the current shooting posture to obtain the target shooting position and target shooting posture of the inspection robot.

[0059] In this step, the Q-learning algorithm can be a reinforcement learning algorithm, such as a value-based iterative reinforcement learning algorithm; the target shooting position can be the shooting position after the current shooting position is updated (e.g., after optimization); the target shooting pose can be the shooting pose after the current shooting pose is updated (e.g., after optimization).

[0060] Specifically, based on the recognition results, the terminal uses the Q-learning algorithm to identify the current shooting position and the current shooting posture, and obtains the updated current shooting position and the updated current shooting posture. The updated current shooting position is used as the target shooting position of the inspection robot, and the updated current shooting posture is used as the target shooting posture of the inspection robot.

[0061] Step S104: Determine the inspection method of the inspection robot in the substation based on the target shooting position and target shooting posture.

[0062] In this step, the target shooting posture can be the target shooting posture corresponding to the target shooting position.

[0063] Specifically, after determining the shooting positions of each target and the corresponding shooting postures, the terminal determines the inspection method of the inspection robot in the substation based on the shooting positions of each target and the corresponding shooting postures.

[0064] In the above-mentioned method for determining the inspection mode of the inspection robot, the following steps are taken: acquiring the current image of the inspection robot taking pictures of the target equipment in the substation, as well as the current shooting position and current shooting posture of the inspection robot; recognizing the current image to obtain the recognition result of the current image; recognizing the current shooting position and current shooting posture based on the recognition result to obtain the target shooting position and target shooting posture of the inspection robot; and determining the inspection mode of the inspection robot in the substation based on the target shooting position and target shooting posture. This scheme acquires current images of the inspection robot taking pictures of target equipment in the substation, along with the robot's current shooting position and posture, thus obtaining the current image captured by the robot at the current shooting position and posture. The current image is then recognized to determine if it meets the specified criteria. Based on the recognition results, the current shooting position and posture are further identified to determine the robot's target shooting position and posture. These are then adjusted to obtain a more optimal target shooting position and posture. Finally, based on the target shooting position and posture, the inspection method of the inspection robot in the substation is determined, thereby improving the accuracy and efficiency of determining the inspection method.

[0065] In one embodiment, in step S104, the inspection method of the inspection robot in the substation is determined according to the target shooting position and the target shooting posture. Specifically, this includes: acquiring the target image of the target equipment captured by the inspection robot in the substation; the target image corresponding to the target shooting position and the target shooting posture; recognizing the target image to obtain the target recognition result; if the target recognition result indicates that the target image is valid, recognizing the target shooting position and the target shooting posture as the inspection point information of the inspection robot; and determining the inspection method of the inspection robot in the substation based on the inspection point information.

[0066] In this embodiment, the target image can be an image captured by the inspection robot at the target shooting position and in the target shooting posture; the target recognition result can be a recognition result used to represent the image quality of the target image, which can be represented as pass or fail; the inspection point information can be used to represent the position of the inspection point in the inspection route of the inspection robot's inspection method and the corresponding shooting posture.

[0067] Specifically, the terminal acquires the target image of the inspection robot at the target shooting position and in the target shooting posture in the substation; performs image recognition on the target image to obtain the target recognition result; if the target recognition result indicates that it passes (or the target recognition result meets the preset image conditions), the target shooting position and target shooting posture are identified as an inspection point information of the inspection robot; based on the obtained inspection point information, the inspection method of the inspection robot in the substation is determined.

[0068] The technical solution of this embodiment, by determining the inspection method of the inspection robot in the substation based on the target shooting position and the target shooting posture when the target recognition result indicates that the target recognition is successful, helps to improve the accuracy of determining the inspection method of the inspection robot.

[0069] In one embodiment, the above steps, determining the inspection method of the inspection robot in the substation based on the inspection point information, specifically includes the following: determining the inspection route, inspection point location, and inspection point shooting posture of the inspection robot in the substation based on the inspection point information; and identifying the inspection route, inspection point location, and inspection point shooting posture of the inspection robot in the substation as the inspection method of the inspection robot in the substation.

[0070] In this embodiment, the inspection route can be the inspection route of the inspection robot in the substation; the inspection point location can be the location of the target equipment photographed by the inspection robot; and the inspection point shooting posture can be the shooting posture corresponding to the inspection robot photographing the target equipment at the inspection point.

[0071] Specifically, the terminal determines the inspection route, inspection point location, and inspection point shooting posture of the inspection robot in the substation based on the information of each inspection point; the inspection route, inspection point location, and inspection point shooting posture of the inspection robot in the substation are used as the inspection method of the inspection robot in the substation.

[0072] The technical solution of this embodiment identifies the inspection route, inspection point location, and inspection point shooting posture of the inspection robot in the substation as the inspection mode of the inspection robot in the substation, thereby improving the accuracy of determining the inspection mode of the inspection robot.

[0073] In one embodiment, step S102 involves recognizing the current image to obtain a recognition result, specifically including: recognizing the target occlusion rate of the current image to obtain a first recognition result; recognizing the target size of the current image to obtain a second recognition result; recognizing the shooting angle of the current image to obtain a third recognition result; recognizing the backlight shooting rate of the current image to obtain a fourth recognition result; recognizing the illumination angle of the current image to obtain a fifth recognition result; and determining the recognition result of the current image based on the first, second, third, fourth, and fifth recognition results.

[0074] In this embodiment, the first recognition result may be the recognition result of the occlusion rate of the target device in the current image; the second recognition result may be the recognition result of the size of the target device in the current image; the third recognition result may be the recognition result of the angle at which the current image was captured; the fourth recognition result may be the recognition result of the backlight shooting rate (such as the degree of backlight) of the current image; and the fifth recognition result may be the recognition result of the illumination angle (or illumination intensity) in the current image.

[0075] Specifically, the terminal performs target occlusion rate recognition on the current image to obtain a first recognition result regarding the target occlusion rate; performs target size recognition on the current image to obtain a second recognition result regarding the target size; performs shooting angle recognition on the current image to obtain a third recognition result regarding the shooting angle; performs backlight shooting rate recognition on the current image to obtain a fourth recognition result regarding the backlight shooting rate; performs illumination angle recognition on the current image to obtain a fifth recognition result regarding the illumination angle; and fuses the first, second, third, fourth, and fifth recognition results to obtain the recognition result of the current image.

[0076] The technical solution of this embodiment, by recognizing the target occlusion rate, target size, shooting angle, backlight shooting rate, and illumination angle of the current image, helps to obtain more accurate recognition results of the current image, thereby improving the accuracy of determining the inspection method of the inspection robot in the future.

[0077] In one embodiment, in step S103, based on the recognition result, the current shooting position and the current shooting posture are identified to obtain the target shooting position and target shooting posture of the inspection robot. Specifically, this includes the following: inputting the recognition result, the current shooting position, and the current shooting posture into the shooting prediction model to obtain the candidate shooting positions, the probability values ​​of the candidate shooting positions, the candidate shooting postures, and the probability values ​​of the candidate shooting postures of the inspection robot; selecting the target shooting position of the inspection robot from the candidate shooting positions and the target shooting posture of the inspection robot from the candidate shooting postures based on the probability values ​​of the candidate shooting positions and the probability values ​​of the candidate shooting postures.

[0078] In this embodiment, the shooting prediction model can be a model built based on the Q-learning algorithm, which can be a model used to predict the shooting position and shooting posture; there can be one or more candidate shooting positions; each candidate shooting position can have one or more corresponding candidate shooting postures.

[0079] Specifically, the terminal inputs the recognition results, current shooting position, and current shooting posture into the shooting prediction model. The shooting prediction model identifies the current shooting position and current shooting posture based on the recognition results, and obtains the candidate shooting position, probability value of the candidate shooting position, candidate shooting posture, and probability value of the candidate shooting posture of the inspection robot (based on the recognition results, the Q-learning algorithm is used to identify the current shooting position and current shooting posture to obtain the candidate shooting position, probability value of the candidate shooting position, candidate shooting posture, and probability value of the candidate shooting posture of the inspection robot). Based on the probability values ​​of the candidate shooting position and the candidate shooting posture, the candidate shooting position with the highest probability value is selected as the target shooting position of the inspection robot, and the candidate shooting posture corresponding to the target shooting position is determined. From the candidate shooting postures corresponding to the target shooting position, the candidate shooting posture with the highest probability value is determined as the target shooting posture of the inspection robot.

[0080] The technical solution of this embodiment determines the target shooting position and target shooting posture of the inspection robot based on the probability values ​​of the candidate shooting positions and the probability values ​​of the candidate shooting postures. This helps to determine a more accurate target shooting position and target shooting posture, thereby improving the accuracy of determining the inspection method of the inspection robot in the future.

[0081] In one embodiment, in step S101, obtaining the current shooting posture of the inspection robot specifically includes the following: obtaining the current shooting angle, current shooting focal length, and current shooting distance of the inspection robot; and identifying the current shooting angle, current shooting focal length, and current shooting distance as the current shooting posture of the inspection robot.

[0082] In this embodiment, the current shooting angle can be the shooting angle corresponding to when the inspection robot takes the current image; the current shooting focal length can be the shooting focal length corresponding to when the inspection robot takes the current image; and the current shooting distance can be the shooting distance corresponding to when the inspection robot takes the current image (which can be the distance between the inspection robot and the target device).

[0083] Specifically, the terminal acquires the current shooting angle, current shooting focal length, and current shooting distance of the inspection robot; and combines the current shooting angle, current shooting focal length, and current shooting distance to obtain the current shooting posture of the inspection robot.

[0084] The technical solution of this embodiment identifies the current shooting angle, current shooting focal length, and current shooting distance as the current shooting posture of the inspection robot, which helps to obtain a more accurate current shooting posture of the inspection robot, thereby improving the accuracy of determining the inspection method of the inspection robot in the future.

[0085] The following example illustrates the inspection method for the inspection robot provided in this application. This example demonstrates the application of this method to a terminal, and the main steps include:

[0086] The first step is for the terminal to acquire the current image of the inspection robot taking pictures of the target equipment in the substation, as well as the current shooting position of the inspection robot; to acquire the current shooting angle, current shooting focal length, and current shooting distance of the inspection robot; and to identify the current shooting angle, current shooting focal length, and current shooting distance as the current shooting posture of the inspection robot.

[0087] The second step involves the terminal identifying the target occlusion rate of the current image to obtain a first identification result; identifying the target size of the current image to obtain a second identification result; identifying the shooting angle of the current image to obtain a third identification result; identifying the backlight shooting rate of the current image to obtain a fourth identification result; and identifying the illumination angle of the current image to obtain a fifth identification result. Based on the first, second, third, fourth, and fifth identification results, the terminal determines the identification result of the current image.

[0088] The third step involves the terminal inputting the recognition results, current shooting position, and current shooting posture into the shooting prediction model to obtain the candidate shooting positions, probability values ​​of the candidate shooting positions, candidate shooting postures, and probability values ​​of the candidate shooting postures of the inspection robot. Based on the probability values ​​of the candidate shooting positions and the probability values ​​of the candidate shooting postures, the target shooting position of the inspection robot is selected from the candidate shooting positions, and the target shooting posture of the inspection robot is selected from the candidate shooting postures.

[0089] The fourth step is to acquire the target image captured by the inspection robot on the target equipment in the substation; the target image corresponds to the target shooting position and the target shooting posture; the target image is recognized to obtain the target recognition result; if the target recognition result indicates that it passes, the target shooting position and the target shooting posture are identified as the inspection point information of the inspection robot.

[0090] The fifth step involves the terminal determining the inspection route, inspection point location, and inspection point shooting posture of the inspection robot in the substation based on the inspection point information; and identifying the inspection route, inspection point location, and inspection point shooting posture of the inspection robot in the substation as the inspection method of the inspection robot in the substation.

[0091] The technical solution of this embodiment obtains the current image captured by the inspection robot at the target equipment in the substation, as well as the current shooting position and shooting posture of the inspection robot, thereby obtaining the current image captured by the inspection robot at the current shooting position and shooting posture; the current image is identified to obtain the identification result, thereby determining whether the captured current image meets the conditions; based on the identification result, the current shooting position and current shooting posture are identified to obtain the target shooting position and target shooting posture of the inspection robot, thereby adjusting the current shooting position and current shooting posture according to the image identification result to obtain a better target shooting position and target shooting posture; based on the target shooting position and target shooting posture, the inspection method of the inspection robot in the substation is determined, thereby improving the accuracy and efficiency of determining the inspection method of the inspection robot.

[0092] The following application example illustrates the inspection method determination method for the inspection robot provided in this application. This application example demonstrates the application of this method to a terminal. Figure 2 and Figure 3 As shown, the main steps include:

[0093] Step 1, Start, S1: The inspection robot takes pictures of the target equipment in the substation, acquires photos, and records its current location and pose. The terminal receives the photos, current location, and current pose sent by the inspection robot.

[0094] The inspection robot consists of a camera pose transfer (PTZ) system (omnidirectional movement and lens zoom / zoom control), a three-axis turntable, and a mobile vehicle or drone. The camera pose transfer includes both infrared and high-definition cameras. When the inspection robot takes pictures, it records the robot's current position and the camera's current pose. The current pose includes the shooting angle, focal length, and shooting distance. Specifically, the robot's current position is determined and used as a reference point; then, the camera photographs the target device and saves the acquired image; finally, the robot's current pose is recorded in real time using a robot pose sensing system.

[0095] The second step, S2: The terminal evaluates the photo and obtains the evaluation results.

[0096] Specifically, the terminal evaluates the photos, including the target occlusion rate, average target size, average shooting angle, backlighting rate, and average illumination angle. Among these, the target occlusion rate and average target size affect whether the photo can be used, while the average shooting angle, backlighting rate, and average illumination angle affect the quality of the photo.

[0097] The third step, S3: Based on the evaluation results, the terminal uses the Q-learning algorithm to provide feedback on the current position and pose of the inspection robot in S1.

[0098] In traditional Q-learning algorithms, the inspection robot randomly selects actions from possible choices, resulting in slow convergence. Therefore, the Q-learning algorithm proposed in this application selects actions based on probability.

[0099]

[0100] In the formula: P(ai|s) represents the selection of action a based on state s. i The probability is given by k, where k is a constant greater than 0. In this case, actions with high Q values ​​will have high selection probabilities, while the selection probabilities of all actions are greater than zero.

[0101] Therefore, the Q-learning algorithm of this application has higher learning efficiency than the traditional Q-learning algorithm. The k value reflects the degree of preference for actions with high Q values. A larger k value will assign a higher probability to actions with above-average Q values, which makes the inspection robot use its learned knowledge to seek actions that maximize rewards. Conversely, a smaller k value will allow other actions to have a higher selection probability, which causes the inspection robot to favor exploring actions that do not currently have high Q values.

[0102] To further improve the convergence rate, the inspection robot learns based on experience, building upon the optimized Q-learning algorithm. This further enhances the efficiency of finding the optimal path. Generally, in both traditional and optimized Q-learning, Q(s,a) is initialized to 0 or other random values, which tends to lead the inspection robot to perform more trials in each round. This affects the learning speed, resulting in slow convergence. Therefore, if Q(s,a) is initialized appropriately, the learning efficiency of the inspection robot can be further improved.

[0103] Step 4, S4: The terminal adjusts the current position and pose of the inspection robot in S1 based on the feedback information (adjusting the current position and pose).

[0104] In the problem of joint inspection route planning for multiple inspection robots in a substation environment, the original action space includes the entire airspace and road surface. Directly using this space would make the solution process extremely complex. To solve this problem, the continuous space can be discretized by using inspection points, which greatly improves the search efficiency of the algorithm. Regarding the factors affecting the imaging effect of substation inspection robots at inspection points, the spatial relationship between inspection points and equipment, the adaptability of inspection points to the environment, and the efficiency of the inspection robot in executing inspection points are comprehensively considered to evaluate the quality of the spatial relationship between inspection points and equipment. First, the spatial relationship between inspection points and equipment is considered, including distance, the tilt angle of the imaging equipment, and the degree of occlusion by other equipment. Second, the adaptability to ambient lighting is considered. Given a position (x... p y p , z p And the shooting target S, the shooting posture at this time is:

[0105]

[0106] In the formula: ρ S Let U be the area information density to characterize the importance of S, where S is the area of ​​the target region, σS is the area differential, and θ is the angle between the camera's line of sight and the surface normal vector. The smaller θ is, the smaller the deformation of the target in the image. When the observation of S from point P is obscured by other devices, U PS =0. When considering ambient lighting, let the angle between the line of sight and the lighting vector be θ. Introducing the illumination influence coefficient β, i.e.

[0107]

[0108]

[0109] In the formula: U P ′ STo account for the potential for image capture under ambient lighting conditions, the suitability of using various inspection robots for image capture and data collection at each point can be determined by traversing and calculating spatial points within the substation. Furthermore, by introducing specific constraints for each type of inspection robot, the automatic deployment of inspection points for substation inspection robots can be achieved, thereby promoting efficient and effective monitoring of substation equipment and significantly improving the overall maintenance and management level of the substation.

[0110] Step 5, S5: Repeat S1 to S4 until the inspection robot reaches the optimal shooting position and optimal pose or reaches the maximum number of learning iterations.

[0111] During the inspection process, to ensure effective monitoring of the equipment under inspection, it is necessary to identify inspection points that can capture images of the target without obstruction. The feasibility of these spatial points can be determined by calculating their imaging potential. To make the calculation of the imaging potential more accurate, the imaging potential of these points is normalized, and points with potential values ​​below a set threshold are removed. The remaining points are added to the candidate point set. If the candidate point set is insufficient to meet the inspection requirements, other detection methods need to be used for monitoring and evaluation.

[0112] Step 6, S6: The terminal obtains the optimal shooting path based on the optimal shooting position and optimal pose.

[0113] In actual inspection processes, the operational space of inspection robots is limited. For example, drones must maintain a safe distance from substation equipment and cannot cross the area where the equipment is located; robots can only move on existing roads. When multiple inspection robots are used, there may be situations where two or more robots can inspect the same target. In this case, the advantages of different equipment should be leveraged. For example, drones are more suitable for high-altitude equipment, while robots have better capabilities for photographing ground equipment. The utilization rate of inspection robots should be maximized by considering various factors. The inspection points established above constitute the action space corresponding to the Q-learning algorithm. Specifically, when an inspection robot is at a certain inspection point, the nearest m (set value) alternative inspection points can be selected as the action space at that moment.

[0114] The basic idea of ​​the Q-algorithm is that the inspection robot explores the environment while continuously estimating the Q-value and learning the optimal strategy, enabling the agent to obtain the maximum long-term reward from the environment. The Q-learning process can be summarized as follows: the agent takes a single action in the environment, the environment provides a corresponding reward, the agent updates its Q-value based on the current state and the next state, and the agent selects the next action based on the Q-value. Improving the action space design of the potential function and the reward design as improvement strategies will significantly enhance model performance.

[0115] In both traditional and optimized Q-learning, the goal of intelligent device inspection route planning is to capture images of all devices while avoiding collisions. Specifically, a subset of inspection points can be randomly selected as destinations, and the substation can be divided into different zones based on their distance to the equipment being inspected: a danger zone, a safe zone, and a target zone. Inspected equipment receives different rewards depending on its location within the zone, thus achieving obstacle avoidance. Figure 3 As shown, the proposed improved Q-learning method employs an action space design based on potential functions and a reward function design based on the analytic hierarchy process (AHP) and entropy weighting. This allows for updating the Q-learning algorithm and the policy through rewards. Simultaneously, the inspection robot observes and provides feedback on the substation environment through different actions. The determination of the comprehensive weights of the indicators is as follows: First, the subjective weights of the nine indicators in the indicator layer to the target layer are determined using the AHP; the elements of the judgment matrix can be set arbitrarily. Then, the objective weights and combined weights of each indicator are determined using the entropy weighting method. Experimental results for different Q-learning algorithms are shown below. Figure 4 As shown, with the same number of learning iterations (ranging from 0 to 400), the proposed improved Q-learning algorithm scores significantly higher than the traditional Q-learning algorithm and the optimized Q-learning algorithm. This is because the improved Q-learning algorithm optimizes the action space, effectively utilizing prior knowledge to reduce the search space and overcoming the problem that reinforcement learning typically requires a large number of trials and explorations. Furthermore, the proposed model also significantly outperforms the other two models in terms of the final score (ranging from 0 to 1). This is because the proposed model integrates the evaluation system into the reward function. Unlike traditional obstacle avoidance problems, substation inspection also needs to consider image quality, lighting conditions, and inspection efficiency. Therefore, when using artificial intelligence methods such as reinforcement learning to improve performance, these requirements must be organically integrated into the algorithm design.

[0116] The technical solution in this application example, for the complex task of substation inspection, effectively improves learning performance and convergence speed by adopting an action space design based on potential function and a reward function design based on analytic hierarchy process and entropy weight method compared to the traditional Q-learning algorithm. The improved algorithm can reduce the number of learning iterations required for convergence by 55.7% and increase the final score by 0.292. Compared to existing inspection route generation strategies, reinforcement learning can solve complex multi-objective optimization problems and has stronger algorithm adaptability. Compared to the energy-optimal deployment scheme and the deployment scheme considering spatial relationships, the final scores were improved by 0.558 and 0.624 respectively. Meanwhile, due to the use of joint inspections to fully leverage the advantages of different equipment, the final score increased by 0.439 compared to using a single device. It can generate inspection routes for inspection robots more effectively, thus playing an important role in the operation and maintenance of smart substations. The action space design based on potential functions and the reward function design based on the analytic hierarchy process and entropy weight method effectively improve learning performance and convergence speed, thus improving the Q-algorithm. It can automatically evaluate the deployment scheme of substation inspection points and routes, improving environmental adaptability and reducing the waste of human resources. It also improves the accuracy and efficiency of determining the inspection method for inspection robots.

[0117] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0118] Based on the same inventive concept, this application also provides an inspection mode determination device for implementing the inspection mode determination method of the inspection robot described above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the inspection mode determination device for inspection robots provided below can be found in the limitations of the inspection mode determination method for inspection robots described above, and will not be repeated here.

[0119] In one embodiment, such as Figure 5 As shown, an inspection method determination device for an inspection robot is provided. The device 500 may include:

[0120] The image acquisition module 501 is used to acquire the current image of the inspection robot taking pictures of the target equipment in the substation, as well as the current shooting position and current shooting posture of the inspection robot.

[0121] The image recognition module 502 is used to recognize the current image and obtain the recognition result of the current image;

[0122] The position recognition module 503 is used to identify the current shooting position and the current shooting posture based on the recognition result, so as to obtain the target shooting position and target shooting posture of the inspection robot.

[0123] The mode determination module 504 is used to determine the inspection mode of the inspection robot in the substation based on the target shooting position and the target shooting posture.

[0124] In one embodiment, the method determination module 504 is further configured to acquire target images taken by the inspection robot of the target equipment in the substation; the target images correspond to the target shooting position and the target shooting posture; the target images are identified to obtain the target recognition result of the target images; if the target recognition result indicates that the target image is passed, the target shooting position and the target shooting posture are identified as the inspection point information of the inspection robot; and the inspection method of the inspection robot in the substation is determined based on the inspection point information.

[0125] In one embodiment, the mode determination module 504 is further configured to determine the inspection route, inspection point location, and inspection point shooting posture of the inspection robot in the substation based on the inspection point information; and to identify the inspection route, inspection point location, and inspection point shooting posture of the inspection robot in the substation as the inspection mode of the inspection robot in the substation.

[0126] In one embodiment, the image recognition module 502 is further configured to: recognize the target occlusion rate of the current image to obtain a first recognition result of the current image; recognize the target size of the current image to obtain a second recognition result of the current image; recognize the shooting angle of the current image to obtain a third recognition result of the current image; recognize the backlight shooting rate of the current image to obtain a fourth recognition result of the current image; recognize the illumination angle of the current image to obtain a fifth recognition result of the current image; and determine the recognition result of the current image based on the first recognition result, the second recognition result, the third recognition result, the fourth recognition result, and the fifth recognition result.

[0127] In one embodiment, the position recognition module 503 is further configured to input the recognition result, the current shooting position, and the current shooting posture into the shooting prediction model to obtain the candidate shooting position, the probability value of the candidate shooting position, the candidate shooting posture, and the probability value of the candidate shooting posture of the inspection robot; and select the target shooting position of the inspection robot from the candidate shooting positions and the target shooting posture of the inspection robot from the candidate shooting postures based on the probability values ​​of the candidate shooting positions and the probability values ​​of the candidate shooting postures.

[0128] In one embodiment, the image acquisition module 501 is further configured to acquire the current shooting angle, current shooting focal length, and current shooting distance of the inspection robot; and to identify the current shooting angle, current shooting focal length, and current shooting distance as the current shooting posture of the inspection robot.

[0129] The inspection method determination device for the aforementioned inspection robot can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0130] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 6 As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements a method for determining the inspection mode of an inspection robot. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0131] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0132] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0133] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0134] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0135] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0136] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0137] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for determining the inspection mode of an inspection robot, characterized in that, The method includes: The system acquires current images of the inspection robot capturing images of target equipment in the substation, as well as the current shooting position and shooting posture of the inspection robot. The inspection robot consists of a camera pose measurement and zoom (PTZ) camera, a three-axis turntable, and a mobile vehicle or drone. The current image is identified to obtain the identification result of the current image; Based on the recognition results, the Q-learning algorithm is used to identify the current shooting position and the current shooting posture to obtain the target shooting position and target shooting posture of the inspection robot; the target shooting position is the updated current shooting position; the target shooting posture is the updated current shooting posture; wherein, the Q-learning algorithm represents the selection of actions based on probability, i.e. In the formula: This indicates that an action is selected based on state s. The probability, where k is a constant greater than 0; The inspection robot acquires target images of the target equipment captured in the substation; the target images correspond to the target shooting position and the target shooting posture. The target image is identified to obtain a target recognition result; the target recognition result is the image quality recognition result of the target image. If the target recognition result indicates that the target is recognized as passed, the target shooting position and the target shooting posture are identified as an inspection point information of the inspection robot. Based on the information obtained from each inspection point, the inspection route, inspection point location, and inspection point shooting posture of the inspection robot in the substation are determined; the inspection point location is the location where the inspection robot shoots the target equipment; the inspection point shooting posture is the shooting posture corresponding to the inspection robot shooting the target equipment at the inspection point. The inspection route, inspection point location, and inspection point shooting posture of the inspection robot in the substation are identified as the inspection mode of the inspection robot in the substation.

2. The method according to claim 1, characterized in that, The target equipment is the power equipment of the substation.

3. The method according to claim 1, characterized in that, The step of recognizing the current image to obtain the recognition result of the current image includes: The target occlusion rate of the current image is identified to obtain a first identification result of the current image; The target size is identified in the current image to obtain a second identification result for the current image; The current image is subjected to shooting angle recognition to obtain a third recognition result for the current image; Perform backlight shooting rate recognition on the current image to obtain a fourth recognition result for the current image; The current image is subjected to illumination angle recognition to obtain the fifth recognition result of the current image; The recognition result of the current image is determined based on the first recognition result, the second recognition result, the third recognition result, the fourth recognition result, and the fifth recognition result.

4. The method according to claim 1, characterized in that, The step of identifying the current shooting position and the current shooting posture based on the identification result to obtain the target shooting position and target shooting posture of the inspection robot includes: The recognition result, the current shooting position, and the current shooting posture are input into the shooting prediction model to obtain the candidate shooting position, the probability value of the candidate shooting position, the candidate shooting posture, and the probability value of the candidate shooting posture of the inspection robot. Based on the probability values ​​of the candidate shooting positions and the probability values ​​of the candidate shooting postures, the target shooting position of the inspection robot is selected from the candidate shooting positions, and the target shooting posture of the inspection robot is selected from the candidate shooting postures.

5. The method according to claim 1, characterized in that, Obtaining the current shooting posture of the inspection robot includes: Obtain the current shooting angle, current shooting focal length, and current shooting distance of the inspection robot; The current shooting angle, the current shooting focal length, and the current shooting distance are all identified as the current shooting posture of the inspection robot.

6. A device for determining the inspection mode of an inspection robot, characterized in that, The device includes: The image acquisition module is used to acquire the current image of the inspection robot taking pictures of the target equipment in the substation, as well as the current shooting position and current shooting posture of the inspection robot; the inspection robot consists of a camera pose measurement and zwittering (PTZ), a three-axis turntable, and a mobile trolley or drone. An image recognition module is used to recognize the current image and obtain the recognition result of the current image; The location recognition module is used to identify the current shooting position and the current shooting posture based on the recognition result using a Q-learning algorithm, thereby obtaining the target shooting position and target shooting posture of the inspection robot; the target shooting position is the updated current shooting position; the target shooting posture is the updated current shooting posture; wherein, the Q-learning algorithm represents selecting actions based on probability, i.e. In the formula: This indicates that an action is selected based on state s. The probability, where k is a constant greater than 0; The method determination module is used to acquire target images captured by the inspection robot on the target equipment in the substation; the target images correspond to the target shooting position and the target shooting posture; the target images are identified to obtain target recognition results; the target recognition results are the image quality recognition results of the target images; if the target recognition results indicate that the target image is acceptable, the target shooting position and the target shooting posture are identified as inspection point information of the inspection robot; based on the obtained inspection point information, the inspection route, inspection point positions, and inspection point shooting postures of the inspection robot in the substation are determined; the inspection point position is the position where the inspection robot captures the target equipment; the inspection point shooting posture is the shooting posture corresponding to the inspection robot capturing the target equipment at the inspection point; the inspection route, inspection point positions, and inspection point shooting postures of the inspection robot in the substation are identified as the inspection method of the inspection robot in the substation.

7. The apparatus according to claim 6, characterized in that, The image recognition module is further configured to perform target occlusion rate recognition on the current image to obtain a first recognition result of the current image; The target size is identified in the current image to obtain a second identification result for the current image; The current image is subjected to shooting angle recognition to obtain a third recognition result for the current image; Perform backlight shooting rate recognition on the current image to obtain a fourth recognition result for the current image; The current image is subjected to illumination angle recognition to obtain the fifth recognition result of the current image; The recognition result of the current image is determined based on the first recognition result, the second recognition result, the third recognition result, the fourth recognition result, and the fifth recognition result.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

Images