Power transmission channel external damage hidden danger detection method and device and computer equipment

Abstract

The invention relates to a power transmission channel external damage hidden danger detection method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the steps of obtaining a scene image which comprises a to-be-detected power transmission channel, performing target detection and depth estimation on the scene image to obtain category information and depth information of each pixel point in the scene image, and detecting whether the to-be-detected power transmission channel has an external damage hidden danger or not according to the category information and the depth information of the pixel points. By adopting the method, pixel-level target detection and depth estimation can be carried out on the scene image, and more detail information in category and space in the scene image can be captured, so that whether an external damage hidden danger exists in the power transmission channel to be detected in the scene image or not can be detected more accurately.

Description

Technical Field

[0001] The present application relates to the technical field of power equipment detection, and in particular to a method, device, computer equipment, computer-readable storage medium and computer program product for detecting hidden dangers of external damage in a power transmission channel. Background Art

[0002] With the development of power systems, transmission channels have become an important way for power system operation and power transmission. However, in the process of daily inspection and maintenance of transmission channels, engineering vehicles such as excavators and cranes are often used. There is a risk of improper operation of these engineering vehicles near transmission channels, which may damage the transmission channels with external forces and bring hidden dangers of external damage. Therefore, how to detect the hidden dangers of external damage in transmission channels has become one of the problems to be solved.

[0003] In traditional solutions, multiple lidar ranging devices are usually deployed near the transmission channel. The lidar ranging devices rotate to scan the environment near the transmission channel to identify whether there are engineering vehicles (such as excavators, cranes) near the transmission channel. When the presence of engineering vehicles is identified, whether the distance between the engineering vehicles and the transmission channel has exceeded the safe distance, thereby detecting whether there is a risk of external damage to the transmission channel.

[0004] However, in traditional solutions, the measurement effect of the lidar ranging equipment is easily affected by bad weather such as rain, snow, fog and haze, resulting in low detection accuracy. In addition, the three-dimensional point cloud data generated by the lidar ranging equipment has a reduced point cloud density at long distances or on non-vertical surfaces, resulting in a decrease in the ability to capture details, which will also lead to a decrease in the detection accuracy of external damage hazards. Summary of the invention

[0005] Based on this, it is necessary to provide a method, device, computer equipment, computer-readable storage medium and computer program product for detecting external damage hazards of a power transmission channel, which can improve the detection accuracy of external damage hazards, in order to address the above technical problems.

[0006] In a first aspect, the present application provides a method for detecting hidden dangers of external damage in a power transmission channel, comprising:

[0007] Acquire a scene image, wherein the scene image includes a power transmission channel to be tested;

[0008] Performing target detection and depth estimation on the scene image to obtain category information and depth information of each pixel in the scene image;

[0009] According to the category information and depth information of the pixel points, it is detected whether the power transmission channel to be tested has a hidden danger of external damage.

[0010] In one embodiment, the scene image includes engineering equipment, and detecting whether there is a hidden danger of external damage in the power transmission channel to be tested in the scene image according to the category information and depth information of the pixel points includes:

[0011] Determine the distance between the power transmission channel to be tested and the engineering equipment in the scene image based on the category information and depth information of the pixel points;

[0012] When the distance is less than or equal to a preset distance threshold, it is determined that there is a hidden danger of external damage in the power transmission channel to be tested.

[0013] In one embodiment, determining the distance between the power transmission channel to be tested and the engineering equipment in the scene image based on the category information and depth information of the pixel points includes:

[0014] Based on the category information, the scene image is divided into a first pixel area including the power transmission channel to be tested and a second pixel area including the engineering equipment;

[0015] Based on the depth information of the pixel points in the first pixel area and the depth information of the pixel points in the second pixel area, the distance between the power transmission channel to be measured and the engineering equipment is determined.

[0016] In one of the embodiments, performing target detection and depth estimation on the scene image to obtain category information and depth information of each pixel in the scene image includes:

[0017] Taking the scene image as input, calling the trained external damage hidden danger detection model to obtain category information and depth information of each pixel in the scene image;

[0018] The trained external damage hazard detection model is trained based on historical scene images carrying category information and synthetic data carrying category information and depth information, and the synthetic data is automatically generated by a server.

[0019] In one of the embodiments, the external damage hidden danger detection model includes a classification model and a depth estimation model, and the classification model and the depth estimation model share a decoder;

[0020] When the external damage hidden danger detection model is called, the following steps are performed:

[0021] Taking the scene image as input, respectively calling the classification model and the depth estimation model to obtain a category feature vector output by the trained classification model and a depth feature vector output by the trained depth estimation model;

[0022] Merging the category feature vector and the depth feature vector to obtain a merged target feature vector;

[0023] Calling the decoder to perform a decoding operation on the target feature vector to obtain a decoding result;

[0024] An upsampling operation is performed on the decoding result to obtain category information and depth information of each pixel in the scene image.

[0025] In one embodiment, the classification model is trained by the following steps:

[0026] Acquire a historical scene image set, wherein the historical scene images in the historical scene image set include power transmission channels and engineering equipment;

[0027] Marking the power transmission channel and the engineering equipment in each of the historical scene images by category, to obtain a plurality of historical scene images carrying category information;

[0028] Based on multiple historical scene images carrying category information, model parameters of a preset semantic segmentation model are adjusted to obtain a trained classification model, where the preset semantic segmentation model is trained based on synthetic data carrying category information.

[0029] In one embodiment, the process of labeling the historical scene images in the historical scene image set by categories to obtain a plurality of historical scene images carrying category information includes:

[0030] The historical scene image is labeled with categories by using a polygon labeling method to obtain a historical scene image carrying category information.

[0031] In a second aspect, the present application also provides a device for detecting hidden dangers of external damage to a power transmission channel, comprising:

[0032] A data acquisition module, used to acquire a scene image, wherein the scene image includes a power transmission channel to be tested;

[0033] A data processing module, used to perform target detection and depth estimation on the scene image to obtain category information and depth information of each pixel in the scene image;

[0034] The hidden danger detection module is used to detect whether there is an external damage hidden danger in the power transmission channel to be tested according to the category information and depth information of the pixel points.

[0035] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps in any one of the above-mentioned embodiments of the method for detecting hidden dangers of external damage to a power transmission channel are implemented.

[0036] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in any one of the above-mentioned embodiments of the method for detecting external damage hazards in a power transmission channel.

[0037] In a fifth aspect, the present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the steps in any of the above-mentioned embodiments of the method for detecting hidden dangers of external damage to a power transmission channel.

[0038] The above-mentioned power transmission channel external damage hidden danger detection method, device, computer equipment, computer readable storage medium and computer program product are different from the traditional scheme. The scene image containing the power transmission channel to be tested is combined with the preset external damage hidden danger detection model. Since the acquisition process of two-dimensional scene images is less affected by rain, snow, fog and haze than three-dimensional point cloud data, it is easier to collect clear and high-precision two-dimensional scene images, which can provide a high-precision detection basis for external damage hidden danger detection. Furthermore, pixel-level target detection and depth estimation are performed on the scene image. This pixel-level image processing can capture more category and spatial detail information in the scene image, overcoming the defect of the laser radar ranging equipment's weak detail capture ability on complex surfaces. Finally, based on the category information and depth information of the pixel points in the scene image, it can more accurately detect whether there is an external damage hidden danger in the power transmission channel to be tested in the scene image. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0040] FIG1 is a diagram showing an application environment of a method for detecting hidden dangers of external damage to a power transmission channel in one embodiment;

[0041] Figure 2 A schematic diagram of a flow chart of a method for detecting hidden dangers of external damage to a power transmission channel in one embodiment;

[0042] Figure 3 It is a flow chart of a method for detecting hidden dangers of external damage to a power transmission channel in another embodiment;

[0043] Figure 4 It is a flowchart of a method for detecting hidden dangers of external damage to a power transmission channel in another embodiment;

[0044] Figure 5A schematic diagram of a flow chart of a method for detecting hidden dangers of external damage to a power transmission channel in a detailed embodiment;

[0045] Figure 6 It is a structural block diagram of a device for detecting hidden dangers of external damage to a power transmission channel in one embodiment;

[0046] FIG. 7 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0048] The method for detecting hidden dangers of external damage to a power transmission channel provided in the embodiment of the present application can be applied in the application environment shown in FIG1 . In which, the terminal 102 communicates with the server 104 through a network. The data storage system can store data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers.

[0049] Specifically, the operator uploads the scene image containing the power transmission channel to be tested to the server 104 through the terminal 102, or the server 104 directly obtains the scene image containing the power transmission channel to be tested from the data storage system. Then, the server 104 performs target detection and depth estimation on the scene image to obtain the category information and depth information of each pixel in the scene image, and detects whether there is a hidden danger of external damage in the power transmission channel to be tested based on the category information and depth information of the pixel.

[0050] The terminal 102 may be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, IoT devices, and portable wearable devices. The IoT devices may be smart speakers, smart TVs, smart air conditioners, smart car devices, projection devices, etc. Portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. Head-mounted devices may be virtual reality (VR) devices, augmented reality (AR) devices, smart glasses, etc. The server 104 may be an independent physical server, a server cluster or distributed system consisting of multiple physical servers, or a cloud server that provides cloud computing services.

[0051] In an exemplary embodiment, Figure 2 As shown, a method for detecting hidden dangers of external damage to a power transmission channel is provided, and the method is applied to the server 104 in FIG. 1 as an example for description, and includes the following steps:

[0052] S100, obtaining a scene image, where the scene image includes a power transmission channel to be tested.

[0053] Among them, the transmission channel refers to the path of power transportation, such as the high-voltage transmission lines for transmitting electricity and the surrounding facilities, such as towers and wires. The transmission channel to be tested is the transmission channel that needs to be tested for hidden dangers of external damage, and the scene image is a two-dimensional image data containing the transmission channel to be tested.

[0054] Exemplarily, a camera or other device may be used to capture scene images, such as using a drone to inspect a transmission channel. The drone may carry a high-definition camera to photograph the transmission channel and generate scene images including transmission towers, wires, and their surroundings.

[0055] S200, performing target detection and depth estimation on the scene image to obtain category information and depth information of each pixel in the scene image.

[0056] Among them, object detection refers to identifying specific objects in an image and determining the bounding boxes and category labels of these objects. For example, object detection on a scene image can identify different objects in the scene image, such as excavators, cranes, and power transmission channels. Depth estimation refers to estimating the distance from an object in an image to a shooting device (such as a camera), that is, inferring the three-dimensional depth information of an object in a scene image by analyzing a two-dimensional image.

[0057] For example, taking a scene image taken by a drone containing a power transmission channel to be tested as an example, a target detection algorithm is first applied to identify objects such as the power transmission channel to be tested, excavators, cranes, etc. in the scene image, and then a depth estimation algorithm, such as a monocular depth estimation algorithm, is used to estimate the depth information of the object detected by the target. It should be noted that the target detection and depth estimation of the scene image in this embodiment are both at the pixel level, so that more accurate target detection results and depth estimation results can be obtained, and the category information and depth information of each pixel in the scene image can be obtained.

[0058] It should be noted that the above-mentioned target detection process and depth estimation process can be performed simultaneously or sequentially. For example, target detection and depth estimation are performed on a scene image simultaneously to obtain target detection results and depth estimation results, and finally the target detection results and depth estimation results are combined to obtain the category information and depth information of each pixel in the scene image. This parallel processing multi-task working mode can improve processing efficiency.

[0059] S300, detecting whether there is a hidden danger of external damage in the power transmission channel to be tested according to the category information and depth information of the pixel points.

[0060] Among them, in the power system, external damage hazards usually refer to external factors that may cause damage to the transmission channel, such as fallen trees, building construction, kites or other objects touching the wires, etc., which may trigger power failures, equipment damage, or even more serious accidents. In this embodiment, the main consideration is whether there is a hidden danger of damaging the transmission channel during the operation of engineering equipment such as excavators and cranes during construction near the transmission channel.

[0061] Following the above steps, after obtaining the category information and depth information of the pixel points of the scene image, if further analysis shows that the scene image includes an excavator, and it is identified that the excavator is very close to the power transmission channel when working, the server will mark the excavator, indicating that the excavator is a potential external damage hazard that may cause the power transmission channel. In addition to analyzing the engineering equipment near the power transmission channel, in the process of target detection of the scene image, trees, buildings, etc. near the power transmission channel can also be detected. For example, the category information and depth information of the pixel points of the scene image indicate that there is a tree near the power transmission channel, and the distance between the two is very close. When there is a storm, the tree may fall towards the power transmission channel and cause an external damage hazard. The server can also mark the potential external damage hazard here.

[0062] Furthermore, after identifying potential external damage hazards, the server can also promptly execute an early warning mechanism to remind staff to pay attention to external damage hazards and take corresponding measures, such as driving away the excavator and pruning branches in time.

[0063] The above-mentioned method for detecting hidden dangers of external damage in power transmission channels is different from the traditional scheme. The above-mentioned scheme combines the scene image containing the power transmission channel to be tested with the preset hidden danger detection model of external damage. Since the acquisition process of two-dimensional scene images is less affected by rain, snow, fog and haze than three-dimensional point cloud data, it is easier to collect clear and high-precision two-dimensional scene images, which can provide a high-precision detection basis for the detection of hidden dangers of external damage. Furthermore, pixel-level target detection and depth estimation are performed on the scene image. This pixel-level image processing can capture more detailed information in terms of category and space in the scene image, overcoming the defect of the weak detail capture ability of the laser radar ranging equipment on complex surfaces. Finally, based on the category information and depth information of the pixel points in the scene image, it can be more accurately detected whether there is a hidden danger of external damage in the power transmission channel to be tested in the scene image.

[0064] Based on the above embodiment, the distances of different objects in the scene image (such as engineering equipment and power transmission channels) can be quantified by using the category information and depth information of each pixel in the scene image, and external damage hazards can be detected accordingly. In an exemplary embodiment, the scene image includes engineering equipment, such as Figure 3 As shown, S300 includes:

[0065] S310, determining the distance between the power transmission channel to be measured and the engineering equipment in the scene image based on the category information and depth information of the pixel points.

[0066] S320: When the distance is less than or equal to a preset distance threshold, it is determined that there is a risk of external damage in the power transmission channel to be tested.

[0067] The category information refers to the object category to which each pixel belongs. For example, a pixel in a scene image may be classified as belonging to a "power transmission channel" or "engineering equipment". The depth information can represent the distance data from each pixel to the camera. Engineering equipment refers to mechanical equipment that is constructed or operated near a power transmission channel, such as excavators, cranes, and construction vehicles.

[0068] For example, if a scene image contains a power transmission channel and an excavator, through object detection, the server can identify which pixels in the scene image belong to the power transmission channel and which belong to the excavator, and through depth estimation, the server can determine the depth information of each pixel. Based on the category information and depth information, the server can determine the distance between the excavator and the camera in the scene image, and the distance between the power transmission channel and the camera, and then determine the minimum distance between the excavator and the power transmission channel.

[0069] Furthermore, if the analysis of the minimum distance between the excavator and the transmission channel shows that the excavator arm of the excavator is extended too high and close to the transmission channel, that is, the minimum distance between the excavator and the transmission channel is less than or equal to the preset distance threshold, the server can determine that the presence of the excavator has brought a risk of external damage to the transmission channel to be tested. At this time, the server can send an alarm message to the staff, prompting the staff to adjust the position of the excavator or stop construction.

[0070] In this embodiment, through pixel-level category information and depth information, the server can accurately measure the distance between the transmission channel and nearby engineering equipment, thereby accurately identifying external damage hazards. This distance calculation based on depth information is more reliable, especially in complex environments or multi-obstacle scenarios, and can significantly improve detection accuracy and real-time performance.

[0071] Continuing from the above embodiments, after obtaining the category information and depth information at the pixel level, the distance between the engineering equipment and the channel to be tested can be accurately determined. In an exemplary embodiment, S310 includes: based on the category information, dividing the scene image into a first pixel area including the power transmission channel to be tested, and a second pixel area including the engineering equipment, and determining the distance between the power transmission channel to be tested and the engineering equipment based on the depth information of the pixel points in the first pixel area and the depth information of the pixel points in the second pixel area.

[0072] The first pixel area refers to the pixel area containing the power transmission channel to be tested, and the second pixel area refers to the pixel area containing engineering equipment (such as construction machinery, construction equipment, etc.). By analyzing the depth information of the pixels in these two areas, the server can calculate the distance between the power transmission channel to be tested and the engineering equipment.

[0073] For example, suppose a drone takes a scene image of a power transmission channel under construction. The server performs target detection on the scene image and divides the scene image into a first pixel area containing the power transmission channel to be tested and a second pixel area containing engineering equipment. By analyzing the depth information of the pixels in these two pixel areas (i.e., the distance from the power transmission channel and the excavator to the shooting equipment), the server can calculate the distance between the excavator and the power transmission channel. If the calculated distance is less than the preset safety threshold, the server will determine that there is a risk of external damage to the power transmission channel to be tested and issue an alarm, prompting the staff to take measures, such as adjusting the position of the engineering equipment or temporarily stopping construction.

[0074] In addition, the above scheme can be carried out in real time. For example, the scene distance of the power transmission channel to be tested can be photographed in real time by the shooting equipment carried by the drone. At the same time, the server can synchronously analyze whether there is any hidden danger of external damage in the power transmission channel to be tested in each scene image, and update the distance between the power transmission channel to be tested and the engineering equipment in real time, so as to dynamically adapt to the changes in the scene, timely discover the hidden danger of external damage and issue an early warning.

[0075] In this embodiment, by dividing the scene image into different pixel areas and analyzing the depth information in these areas, compared with relying solely on the bounding box of the object, the pixel-level depth analysis provides a more accurate basis for distance measurement, and can more accurately determine the distance between the transmission channel and the engineering equipment, thereby improving the detection accuracy of external damage hazards.

[0076] In an exemplary embodiment, Figure 3 As shown, S200 includes:

[0077] S210, taking the scene image as input, calling the trained external damage hidden danger detection model to obtain the category information and depth information of each pixel in the scene image.

[0078] Among them, the external damage hidden danger detection model can be trained based on a convolutional neural network (CNN) or a similar architecture, and the training data is trained based on historical scene images carrying category information and synthetic data carrying category information and depth information, and the synthetic data is automatically generated by the server. Among them, synthetic data refers to data carrying category information and depth information automatically generated by a computer. This is because when the demand for training data is large, it takes a long time to manually label all of it, so a computer can be used to automatically generate large-scale training data for training the model. In addition, after the external damage hidden danger detection model is trained based on large-scale synthetic data, some manually labeled training data can be used to train the external damage hidden danger detection model based on the application scenario of this embodiment, and the model parameters can be fine-tuned to make the trained external damage hidden danger detection model more suitable for actual application scenarios.

[0079] Specifically, the synthetic data carrying category information and depth information is simulated data automatically generated by the server, which contains the category and depth information of various objects that may be encountered in real scenes. Each pixel is assigned a correct category label, such as power lines, trees, construction equipment, etc. Through these data, the model gradually learns how to identify and classify different objects in the scene, and how to accurately estimate their depths during the training process. Furthermore, after preliminary training with synthetic data, historical scene images carrying category information can also be used. For example, historical scene images are annotated with labels such as power transmission channels, excavators, and cranes. Based on these historical scene images, the model is retrained, and the model parameters are fine-tuned to obtain a trained external damage hidden danger detection model. Finally, the scene image is used as input, and the trained external damage hidden danger detection model is called to obtain the category information and depth information of each pixel in the scene image.

[0080] In this embodiment, the trained external damage hidden danger detection model can handle complex and changeable scenes, so as to quickly process the input scene images and generate pixel-level category information and depth information. In addition, the external damage hidden danger detection model has been trained with a large amount of synthetic data and has strong generalization ability. It can adapt to scene images in different environments and improve the stability of external damage hidden danger detection.

[0081] Following the above embodiment, the model architecture of the external damage hidden danger detection model may include a classification model and a depth estimation model, wherein the classification model and the depth estimation model share a decoder. In an exemplary embodiment, Figure 4 As shown, when the external damage potential detection model is called, the following steps are performed:

[0082] S211, taking the scene image as input, respectively calling the classification model and the depth estimation model to obtain the category feature vector output by the trained classification model and the depth feature vector output by the trained depth estimation model.

[0083] S212, merging the category feature vector and the depth feature vector to obtain a merged target feature vector.

[0084] S213, calling the decoder to perform a decoding operation on the target feature vector to obtain a decoding result.

[0085] S214, performing an up-sampling operation on the decoding result to obtain category information and depth information of each pixel in the scene image.

[0086] Among them, the main task of the classification model is to identify objects of different categories in the scene image. In this embodiment, the classification model can identify objects such as power transmission channels and engineering equipment, and output this information in the form of a category feature vector. The category feature vector is high-dimensional data that contains the category information of each object in the scene image. The main task of the depth estimation model is to predict the depth information of each pixel in the scene image and output this information in the form of a depth feature vector. The main task of the decoder is to convert the high-dimensional feature vector into a more intuitive image feature. For example, the decoder can include a multi-layer convolutional neural network structure for gradually restoring and extracting features. The upsampling operation is the process of restoring a low-resolution feature map to the resolution of the input image.

[0087] Furthermore, the category feature vector and the depth feature vector are merged, and the merging method includes but is not limited to simple vector concatenation, feature mapping, or processing these features through a more complex fusion network to obtain a merged target feature vector, so that the server can simultaneously process category information and depth information in a unified feature space. Then, the classification model and the depth estimation model share a decoder, and perform a decoding operation on the target feature vector to obtain a decoding result. Finally, the decoding result is upsampled to restore the low-resolution feature map in the decoding result to the resolution of the input image, so that the category information and depth information of each pixel in the scene image can be accurately located.

[0088] In this embodiment, through the collaborative work of the classification model, the depth estimation model, and the shared decoder, the external damage hidden danger detection model can efficiently process the scene image and generate pixel-level category information and depth information, providing a reliable data basis for subsequent external damage hidden danger detection. In addition, the classification model and the depth estimation model simultaneously process the scene image and the model architecture of the shared decoder, which optimizes resource utilization and enables complex data processing tasks to be efficiently completed in a real-time environment.

[0089] Among them, the classification model in the external damage hidden danger detection model is trained based on a historical scene image set with labeled category information. In an exemplary embodiment, the classification model is trained by the following steps: obtaining a historical scene image set, the historical scene images in the historical scene image set include power transmission channels and engineering equipment, classifying the power transmission channels and engineering equipment in each historical scene image to obtain multiple historical scene images carrying category information, adjusting the model parameters of a preset semantic segmentation model based on the multiple historical scene images carrying category information, and obtaining a trained classification model, wherein the preset semantic segmentation model is trained based on synthetic data carrying category information.

[0090] Among them, the historical scene image set contains multiple historical scene images. In order to improve the performance of the classification model, these historical scene images should be as extensive and multi-source as possible. For example, the historical scene image set includes historical scene images under different lighting, weather conditions, angles, distances, different target object sizes, and different distance distributions between engineering equipment and transmission channels. Category labeling refers to classifying pixels or objects in an image and identifying their categories. The semantic segmentation model is a deep learning model that aims to segment the input image into different semantic regions, such as transmission channels, engineering equipment, and background.

[0091] Exemplarily, after acquiring a set of historical scene images, category labels can be manually or semi-automatically assigned to the power transmission channels, engineering equipment, and other related objects in each historical scene image. After completing the category labeling, the next step is to use these historical scene images with category information to train a preset semantic segmentation model. During the training process, the semantic segmentation model gradually adjusts its internal parameters (such as the weight and bias of the convolution kernel) to improve the ability to distinguish objects of different categories. Through the above steps, the semantic segmentation model is continuously optimized and adjusted to eventually form a model with high-precision classification capabilities, that is, a trained classification model. The trained classification model can accurately classify the newly input scene images at the pixel level in actual applications and identify the power transmission channels and their nearby engineering equipment.

[0092] It should be noted that the semantic segmentation model preset in this embodiment is preliminarily trained with synthetic data before fine-tuning using real historical scene images. Synthetic data is usually generated by simulating real scenes and can be automatically completed by the server. The synthetic data carries pixel-level category information and can cover various possible scenes and objects, so that the semantic segmentation model can obtain sufficiently extensive training in the initial stage, thereby enhancing its generalization ability. Similarly, the depth estimation model can also be trained based on similar synthetic data, and the depth estimation model and the classification model can be jointly trained. The training data is synthetic data that carries both pixel-level category information and pixel-level depth information. In this way, the depth estimation model and the classification model can not only share computing resources and improve computing efficiency, but also use category information to enhance depth estimation during training, especially in scenes with clear object categories (engineering equipment and power transmission channels in this embodiment). This method can significantly reduce model errors.

[0093] In this embodiment, through preliminary training with synthetic data, the semantic segmentation model can maintain high stability when facing a wide range of input data, and then fine-tune the parameters in combination with real scene images. The classification model finally trained has both good generalization ability and high accuracy and practicality, which improves the performance of the classification model in complex scenarios, thereby improving the accuracy of external damage detection of transmission channels.

[0094] In order to make a clearer description of the method for detecting hidden dangers of power transmission channel external damage provided by the present application, the following is a specific embodiment and the attached Figure 5 To illustrate, this specific embodiment includes the following steps:

[0095] S501, obtaining a scene image, the scene image including the power transmission channel to be tested, taking the scene image as input, calling a classification model and a depth estimation model respectively, and obtaining a category feature vector output by the trained classification model and a depth feature vector output by the trained depth estimation model.

[0096] S502, merging the category feature vector and the depth feature vector to obtain a merged target feature vector, calling a decoder to perform a decoding operation on the target feature vector to obtain a decoding result.

[0097] S503: Perform an up-sampling operation on the decoding result to obtain category information and depth information of each pixel in the scene image.

[0098] S504: Determine the distance between the power transmission channel to be measured and the engineering equipment in the scene image based on the category information and depth information of the pixel points.

[0099] S505: When the distance is less than or equal to a preset distance threshold, it is determined that there is a risk of external damage in the power transmission channel to be tested.

[0100] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0101] Based on the same inventive concept, the embodiment of the present application also provides a power transmission channel external damage hidden danger detection device for implementing the power transmission channel external damage hidden danger detection method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more embodiments of the power transmission channel external damage hidden danger detection device provided below can refer to the limitations of the power transmission channel external damage hidden danger detection method above, and will not be repeated here.

[0102] In an exemplary embodiment, Figure 6 As shown, a transmission channel external damage hidden danger detection device 600 is provided, comprising: a data acquisition module 610, a data processing module 620 and a hidden danger detection module 630, wherein:

[0103] The data acquisition module 610 is used to acquire a scene image, where the scene image includes a power transmission channel to be tested.

[0104] The data processing module 620 is used to perform target detection and depth estimation on the scene image to obtain category information and depth information of each pixel in the scene image.

[0105] The hidden danger detection module 630 is used to detect whether there is an external damage hidden danger in the power transmission channel to be tested based on the category information and depth information of the pixel points.

[0106] In an exemplary embodiment, the scene image includes engineering equipment, and the hidden danger detection module 630 is also used to determine the distance between the power transmission channel to be tested and the engineering equipment in the scene image based on the category information and depth information of the pixel points, and when the distance is less than or equal to a preset distance threshold, it is determined that there is a hidden danger of external damage in the power transmission channel to be tested.

[0107] In an exemplary embodiment, the hidden danger detection module 630 is also used to divide the scene image into a first pixel area including the power transmission channel to be tested and a second pixel area including the engineering equipment based on the category information, and determine the distance between the power transmission channel to be tested and the engineering equipment based on the depth information of the pixel points in the first pixel area and the depth information of the pixel points in the second pixel area.

[0108] In an exemplary embodiment, the data processing module 620 is also used to call a trained external damage hazard detection model with a scene image as input to obtain category information and depth information of each pixel in the scene image, wherein the trained external damage hazard detection model is trained based on historical scene images carrying category information and synthetic data carrying category information and depth information, and the synthetic data is automatically generated by the server.

[0109] In an exemplary embodiment, the external damage hidden danger detection model includes a classification model and a depth estimation model, and the classification model and the depth estimation model share a decoder. The data processing module 620 is also used to take the scene image as input, call the classification model and the depth estimation model respectively, obtain the category feature vector output by the trained classification model, and the depth feature vector output by the trained depth estimation model, merge the category feature vector and the depth feature vector to obtain a merged target feature vector, call the decoder to perform a decoding operation on the target feature vector to obtain a decoding result, and upsample the decoding result to obtain the category information and depth information of each pixel in the scene image.

[0110] In an exemplary embodiment, the transmission channel external damage hidden danger detection device 600 is also used to obtain a historical scene image set, the historical scene images in the historical scene image set include transmission channels and engineering equipment, and the transmission channels and engineering equipment in each historical scene image are categorized to obtain multiple historical scene images carrying category information, and based on the multiple historical scene images carrying category information, the model parameters of a preset semantic segmentation model are adjusted to obtain a trained classification model, and the preset semantic segmentation model is trained based on synthetic data carrying category information.

[0111] In an exemplary embodiment, the transmission channel external damage hidden danger detection device 600 is also used to use a polygon annotation method to perform category annotation on the historical scene image to obtain the historical scene image carrying category information.

[0112] Each module in the above-mentioned power transmission channel external damage hidden danger detection device can be fully or partially implemented by software, hardware and their combination. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the corresponding operations of each of the above modules.

[0113] In an exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be shown in FIG7. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data such as scene images. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for detecting hidden dangers of external damage to a power transmission channel is implemented.

[0114] Those skilled in the art will understand that the structure shown in FIG. 7 is merely a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0115] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the steps in any one of the above-mentioned embodiments of the method for detecting hidden dangers of external damage to a power transmission channel are implemented.

[0116] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in any one of the above-mentioned power transmission channel external damage hazard detection methods are implemented.

[0117] In one embodiment, a computer program product is provided, including a computer program, which, when executed by a processor, implements the steps in any one of the above-mentioned power transmission channel external damage hidden danger detection embodiments.

[0118] 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, stored data, displayed data, 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 relevant data must comply with relevant regulations.

[0119] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile memory 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. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0120] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, 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 application.

[0121] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for detecting hidden dangers of external damage in a power transmission channel, characterized in that: The method comprises: Acquire a scene image, wherein the scene image includes a power transmission channel to be tested; Performing target detection and depth estimation on the scene image to obtain category information and depth information of each pixel in the scene image; According to the category information and depth information of the pixel points, it is detected whether the power transmission channel to be tested has a hidden danger of external damage.

2. The method according to claim 1, characterized in that The scene image includes engineering equipment, and detecting whether there is a hidden danger of external damage in the power transmission channel to be tested in the scene image according to the category information and depth information of the pixel points includes: Determine the distance between the power transmission channel to be tested and the engineering equipment in the scene image based on the category information and depth information of the pixel points; When the distance is less than or equal to a preset distance threshold, it is determined that there is a hidden danger of external damage in the power transmission channel to be tested.

3. The method according to claim 2, characterized in that The determining, based on the category information and depth information of the pixel points, the distance between the power transmission channel to be measured and the engineering equipment in the scene image includes: Based on the category information, the scene image is divided into a first pixel area including the power transmission channel to be tested and a second pixel area including the engineering equipment; Based on the depth information of the pixel points in the first pixel area and the depth information of the pixel points in the second pixel area, the distance between the power transmission channel to be measured and the engineering equipment is determined.

4. The method according to any one of claims 1 to 3, characterized in that: The performing target detection and depth estimation on the scene image to obtain category information and depth information of each pixel in the scene image includes: Taking the scene image as input, calling the trained external damage hidden danger detection model to obtain category information and depth information of each pixel in the scene image; The trained external damage hazard detection model is trained based on historical scene images carrying category information and synthetic data carrying category information and depth information, and the synthetic data is automatically generated by a server.

5. The method according to claim 4, characterized in that The external damage hidden danger detection model includes a classification model and a depth estimation model, and the classification model and the depth estimation model share a decoder; When the external damage hidden danger detection model is called, the following steps are performed: Taking the scene image as input, respectively calling the classification model and the depth estimation model to obtain a category feature vector output by the trained classification model and a depth feature vector output by the trained depth estimation model; Merging the category feature vector and the depth feature vector to obtain a merged target feature vector; Calling the decoder to perform a decoding operation on the target feature vector to obtain a decoding result; An upsampling operation is performed on the decoding result to obtain category information and depth information of each pixel in the scene image.

6. The method according to claim 5, characterized in that The classification model is trained by the following steps: Acquire a historical scene image set, wherein the historical scene images in the historical scene image set include power transmission channels and engineering equipment; Marking the power transmission channel and the engineering equipment in each of the historical scene images by category, to obtain a plurality of historical scene images carrying category information; Based on multiple historical scene images carrying category information, model parameters of a preset semantic segmentation model are adjusted to obtain a trained classification model, where the preset semantic segmentation model is trained based on synthetic data carrying category information.

7. The method according to claim 6, characterized in that The step of labeling the historical scene images in the historical scene image set by categories to obtain a plurality of historical scene images carrying category information includes: The historical scene image is labeled with categories by using a polygon labeling method to obtain a historical scene image carrying category information.

8. A device for detecting hidden dangers of external damage in a power transmission channel, characterized in that: The device comprises: A data acquisition module, used to acquire a scene image, wherein the scene image includes a power transmission channel to be tested; A data processing module, used to perform target detection and depth estimation on the scene image to obtain category information and depth information of each pixel in the scene image; The hidden danger detection module is used to detect whether there is an external damage hidden danger in the power transmission channel to be tested according to the category information and depth information of the pixel points.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images