Power transmission tower footing hidden danger identification method and device and computer equipment

Abstract

The invention relates to a power transmission tower footing hidden danger identification method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring an original power transmission tower footing image, preprocessing the original power transmission tower footing image to obtain a target power transmission tower footing image, taking the target power transmission tower footing image as input, calling a preset hidden danger identification model, and performing target detection and mask prediction on the power transmission tower footing image to identify whether the power transmission tower footing is wound by a plant or not, the preset hidden danger recognition model is obtained by training the hybrid task cascade model based on the historical power transmission tower footing image carrying the plant winding labeling information. By adopting the scheme, the accuracy of hidden danger identification can be improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method, device, computer equipment, computer-readable storage medium and computer program product for identifying hidden dangers in transmission tower foundations. Background Art

[0002] With the rapid development of power systems, the safe operation of transmission lines has become a key link in ensuring the reliability and stability of power supply. Among them, transmission towers are important basic supporting components of transmission lines. In humid climates, vines are prone to grow near the base of transmission towers, causing the soil to loosen, leading to safety hazards in the tower base. Therefore, how to quickly and accurately identify hidden dangers in the base of transmission towers has become one of the problems to be solved.

[0003] In traditional solutions, manual inspections are usually carried out on transmission towers, or shooting equipment is installed near the transmission towers to regularly or irregularly shoot images of the tower bases of the transmission towers, and the images are sent back to analyze whether there are any safety hazards in the tower bases of the transmission towers based on the image data.

[0004] However, the above-mentioned manual inspection method requires large human resources investment and is prone to missing hidden dangers. Hidden danger identification is performed by remotely acquiring image data. The hidden danger identification results depend on the quality of the image data and are easily affected by the complex background and light interference of the transmission tower, resulting in low accuracy of hidden danger identification. 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 that can improve the accuracy of identifying hidden dangers in transmission tower foundations in order to solve the above-mentioned technical problems.

[0006] In a first aspect, the present application provides a method for identifying hidden dangers of a transmission tower foundation, comprising:

[0007] Obtain the original transmission tower base image;

[0008] Preprocessing the original transmission tower base image to obtain a target transmission tower base image;

[0009] Taking the target transmission tower base image as input, calling a preset hidden danger identification model, performing target detection and mask prediction on the transmission tower base image, so as to identify whether the transmission tower base is entangled by plants, and obtaining a hidden danger identification result;

[0010] Among them, the preset hidden danger identification model is obtained by training the hybrid task cascade model based on historical transmission tower base images carrying plant entanglement annotation information.

[0011] In one embodiment, the hidden danger identification model includes a full convolution branch, and when the hidden danger identification model is called, the following steps are performed:

[0012] Extracting image features of the target transmission tower base image;

[0013] Performing a pooling operation on the image features to obtain a pooling result of the transmission tower base image;

[0014] Based on the pooling result, performing target detection on the transmission tower base image to obtain the target detection result;

[0015] Using the target detection result as prior knowledge, calling the full convolution branch, performing mask prediction on the target transmission tower base image, and obtaining a binary mask map;

[0016] Using the target detection result and the binary mask map as prior knowledge, a multi-stage hidden danger identification task is performed to determine a target binary mask map;

[0017] Based on the target binary mask image, it is identified whether the transmission tower base is entangled by plants to obtain a hidden danger identification result.

[0018] In one embodiment, the hidden danger identification model is called to perform the following steps when performing the hidden danger identification task at each stage:

[0019] Using the target detection result obtained in the previous stage and the binary mask image obtained in the previous stage as prior knowledge, based on the pooling result, target detection is performed on the transmission tower base image, and the target detection result is updated;

[0020] The updated target detection result and the binary mask map generated in the previous stage are used as prior knowledge, the full convolution branch is called, the mask prediction is performed on the target transmission tower base image, and the binary mask map is updated.

[0021] In one embodiment, after obtaining the hidden danger identification result, the method further includes:

[0022] In the case where the hidden danger identification result indicates that the transmission tower base is entangled by plants, determining a coverage area of ​​the transmission tower base covered by plants;

[0023] Determining the hidden danger level of the transmission tower foundation based on the coverage area of ​​the plants;

[0024] Based on the coverage area and the hidden danger level, a preset hidden danger report template is called to generate a hidden danger report of the transmission tower foundation.

[0025] In one embodiment, the hidden danger identification model is trained based on the following steps:

[0026] Acquire a historical transmission tower base image, wherein the historical transmission tower base image includes a transmission tower base entangled by plants;

[0027] Performing instance segmentation on the historical transmission tower base image, marking the category label of each pixel in the historical transmission tower base image, and obtaining a marked historical transmission tower base image;

[0028] The annotated historical transmission tower base images are used as training data to train the mixed task cascade model to obtain a hidden danger identification model.

[0029] In one embodiment, the preprocessing of the original transmission tower base image to obtain the target transmission tower base image includes at least one of the following:

[0030] The first item: performing brightness enhancement on the original transmission tower base image to obtain the target transmission tower base image;

[0031] The second item is to perform contrast enhancement on the original transmission tower base image to obtain the target transmission tower base image;

[0032] The third item: performing a denoising operation on the original transmission tower base image to obtain the target transmission tower base image.

[0033] In a second aspect, the present application also provides a transmission tower foundation hidden danger identification device, comprising:

[0034] A data acquisition module, used to acquire the original transmission tower base image;

[0035] A data preprocessing module, used for preprocessing the original transmission tower base image to obtain a target transmission tower base image;

[0036] A hidden danger identification module is used to use the target transmission tower base image as input, call a preset hidden danger identification model, perform target detection and mask prediction on the transmission tower base image, so as to identify whether the transmission tower base is entangled by plants, and obtain a hidden danger identification result;

[0037] Among them, the preset hidden danger identification model is obtained by training the hybrid task cascade model based on historical transmission tower base images carrying plant entanglement annotation information.

[0038] 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 identifying hidden dangers in a transmission tower foundation are implemented.

[0039] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps in any one of the above-mentioned transmission tower foundation hazard identification method embodiments are implemented.

[0040] 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 one of the above-mentioned transmission tower foundation hazard identification method embodiments.

[0041] The above-mentioned hidden danger identification method, device, computer equipment, computer-readable storage medium and computer program product can improve the quality of the image subsequently input into the hidden danger identification model by preprocessing the original transmission tower base image, and provide a basis for subsequent accurate hidden danger identification. Then, based on the trained hidden danger identification model, target detection and mask prediction are performed on the target transmission tower base image, which can not only roughly and quickly determine the target object in the scene image by target detection, such as marking the transmission tower base, plants, etc. by a bounding box, but also perform pixel-level target recognition on the target object in the scene image by mask prediction. In addition, the hidden danger identification model is obtained by training the hybrid task cascade model based on the historical transmission tower base image carrying the plant entanglement annotation information, and can integrate the complementary features in the target detection result and the mask prediction result, so as to accurately locate the transmission tower base and plants in the target transmission tower base image, and then accurately identify whether the transmission tower base is entangled by plants, thereby improving the accuracy of hidden danger identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] 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.

[0043] FIG1 is a diagram showing an application environment of a method for identifying hidden dangers of a transmission tower foundation according to an embodiment;

[0044] Figure 2 A schematic diagram of a flow chart of a method for identifying hidden dangers of a transmission tower foundation in one embodiment;

[0045] Figure 3 A schematic diagram of a process for determining a hidden danger identification result in one embodiment;

[0046] Figure 4 A framework diagram of a hidden danger identification model in one embodiment;

[0047] Figure 5A schematic diagram of a process for generating a hidden danger report in one embodiment;

[0048] Figure 6 A schematic diagram of a flow chart of a method for identifying hidden dangers of a transmission tower foundation in a detailed embodiment;

[0049] Figure 7 A structural block diagram of a transmission tower foundation hidden danger identification device in one embodiment;

[0050] Figure 8 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0051] 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.

[0052] The transmission tower foundation hidden danger identification method provided in the embodiment of the present application can be applied to the application environment shown in FIG1. ​​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.

[0053] Specifically, the inspection personnel may upload the original transmission tower base image to the server 104 through the terminal 102, or the server 104 may directly obtain the original transmission tower base image from the data storage system. Then, the server 104 pre-processes the original transmission tower base image to obtain the target transmission tower base image, takes the target transmission tower base image as input, calls the preset hidden danger identification model, performs target detection and mask prediction on the transmission tower base image, and identifies whether the transmission tower base is entangled by plants, thereby obtaining the hidden danger identification result. Among them, the preset hidden danger identification model is obtained by training the hybrid task cascade model based on the historical transmission tower base image carrying the plant entanglement annotation information.

[0054] 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.

[0055] In an exemplary embodiment, Figure 2 As shown, a method for identifying hidden dangers of a transmission tower foundation is provided, and the method is applied to the server 104 in FIG. 1 as an example for explanation, including the following steps:

[0056] S100, obtaining an original transmission tower base image.

[0057] The transmission tower base is the basic part that supports the transmission line. When encountering a humid and rainy climate environment, the continuous growth of vines will cause the soil near the transmission tower base to loosen, and the vines will entangle the transmission tower base, resulting in a decrease in the stability of the transmission tower base. The original transmission tower base image can be an infrared image or a visible light image. For example, an infrared camera or a visible light camera is used to shoot the transmission tower base area to obtain the original transmission tower base image. For example, in a certain power inspection task, the original transmission tower base image of the transmission tower base can be obtained by shooting by a drone.

[0058] S200, preprocessing the original transmission tower base image to obtain a target transmission tower base image.

[0059] Among them, the preprocessing operation includes but is not limited to operations such as image enhancement, noise removal and image standardization. Exemplarily, the original transmission tower base image is subjected to brightness and contrast enhancement processing by an adaptive histogram equalization algorithm. According to the histogram distribution characteristics of the original transmission tower base image, the brightness distribution of the image is adaptively adjusted to improve the overall contrast of the image, and the transmission tower base image with enhanced brightness and contrast is obtained. Then, the transmission tower base image with enhanced brightness and contrast is subjected to noise removal operation, for example, a Gaussian filtering algorithm is used, and a suitable filter size and standard deviation are selected, so that the edge information of the transmission tower base image can be retained as much as possible while removing noise. Finally, an image standardization operation is performed on the transmission tower base image, and the size of the transmission tower base image is adjusted to the same size to facilitate the processing of the subsequent hidden danger identification model.

[0060] S300, taking the target transmission tower base image as input, calling the preset hidden danger identification model, performing target detection and mask prediction on the transmission tower base image, so as to identify whether the transmission tower base is entangled by plants, and obtaining the hidden danger identification result.

[0061] Among them, the preset hidden danger identification model is obtained by training the hybrid task cascade model based on the historical transmission tower base image carrying plant entanglement annotation information. In this embodiment, the hybrid task cascade model is a multi-stage pipeline processing flow that combines multiple tasks (target detection tasks and mask prediction tasks) together, and in the task execution stage, it can utilize the relationship between the target detection task and the mask prediction task, and continuously improve the accuracy of the model output results through multi-stage gradual refinement and task interweaving.

[0062] In this embodiment, target detection refers to identifying specific target objects in an image, such as a transmission tower base and plants (such as vines) near the transmission tower base. A bounding box can be used to mark the above-mentioned specific target objects to obtain a target detection result. Mask prediction refers to generating a pixel-level mask for each target object in the target transmission tower base image, thereby marking the precise outline of the target object and generating a binary mask map. Finally, combining the target detection result and the binary mask map, a comprehensive analysis is performed in the target transmission tower base image to determine whether the transmission tower base is entangled by plants, as well as the location, area, and shape of the entanglement by plants, to obtain a hidden danger identification result. Furthermore, if the hidden danger identification result indicates that the transmission tower base is entangled by plants, a hidden danger warning message is generated to notify relevant personnel to handle the situation on site.

[0063] The above-mentioned hidden danger identification method, by preprocessing the original transmission tower base image to obtain the target transmission tower base image, can improve the quality of the subsequent input hidden danger identification model image, and provide a basis for subsequent accurate hidden danger identification. Then, based on the trained hidden danger identification model, target detection and mask prediction are performed on the target transmission tower base image, which can not only roughly and quickly determine the target object in the scene image by target detection, such as marking the transmission tower base, plants, etc. by a bounding box, but also perform pixel-level target recognition on the target object in the scene image by mask prediction. In addition, the hidden danger identification model is based on the historical transmission tower base image carrying plant entanglement annotation information, and is obtained by training the hybrid task cascade model. It can integrate the complementary features in the target detection results and the mask prediction results, so that the transmission tower base and plants in the target transmission tower base image can be accurately located, and then accurately identify whether the transmission tower base is entangled by plants, thereby improving the accuracy of hidden danger identification.

[0064] Following the above embodiment, the hidden danger identification process needs to perform target detection tasks and mask prediction tasks. These two tasks complement each other. Target detection can identify the location of the transmission tower base and plants, while mask prediction can further refine the target detection results and accurately calibrate the specific area where the vines are entangled. In an exemplary embodiment, Figure 3 As shown, the hidden danger identification model includes a full convolution branch. When the hidden danger identification model is called, the following steps are performed:

[0065] S310, extracting image features of the target transmission tower base image.

[0066] S320, performing a pooling operation on the image features to obtain a pooling result of the transmission tower base image.

[0067] S330: Based on the pooling result, target detection is performed on the transmission tower base image to obtain a target detection result.

[0068] S340, using the target detection result as prior knowledge, calling the full convolution branch, performing mask prediction on the target transmission tower base image, and obtaining a binary mask map.

[0069] S350, using the target detection result and the binary mask map as prior knowledge, performing a multi-stage hidden danger identification task and determining the target binary mask map.

[0070] S360, based on the target binary mask image, identify whether the transmission tower base is entangled by plants to obtain a hidden danger identification result.

[0071] Among them, the model architecture of the hidden danger identification model can be as follows: Figure 4 shown. Figure 4 In the hidden danger identification model, there are multiple stages of target detection task branches and mask prediction task branches. Each stage will perform target detection tasks and mask prediction tasks, and the results of the previous stage will be embedded in the next stage, thereby improving the accuracy of the final hidden danger identification results. When the mask prediction task is executed, the target transmission tower base can be first segmented at the pixel level through a fully convolutional network, and the probability of each pixel belonging to the plant entanglement area can be predicted to obtain the mask prediction result. Then, the mask prediction result is converted into a binary mask map through threshold processing. The binary mask map can characterize the position and shape of the plant entanglement area in the target transmission tower base image. In this embodiment, the plants entangled with the transmission tower base are explained as vines.

[0072] Exemplarily, the input target transmission tower base image is first subjected to feature extraction and pooling operations through a feature extraction network and a pooling layer, and then the target detection task is performed to identify the transmission tower base in the image and generate an initial bounding box for the transmission tower base. Afterwards, the target transmission tower base image carrying the initial bounding box is subjected to a pooling operation to obtain a pooling result. Based on the pooling result, the hidden danger identification model performs a mask prediction task on the target transmission tower base image to generate an initial binary mask map. At this time, the initial binary mask map may not be very accurate, especially in a complex background, where the vine-entangled area may be confused with other parts of the transmission tower base. However, in the subsequent stage of performing the hidden danger identification task, the hidden danger identification model will perform the target detection task again based on the initial binary mask map generated in the previous stage. At this time, since the initial binary mask map can help the hidden danger identification model better understand the shape and size of the target object (transmission tower base and vines), a more refined bounding box can be generated. For example, the bounding box generated in this stage is smaller in area than the initial bounding box, reducing the coverage of the bounding box to the range outside the target object. In addition, the bounding box can still provide a basis for the mask prediction task in the current stage. The binary mask map obtained by mask prediction is also more accurate than the initial binary mask map. Moreover, when performing the mask prediction task, since the mask prediction task branches of different stages are directly connected, the initial binary mask map of the previous stage can also be referred to. The initial binary mask map is used as prior information to generate a more accurate binary mask map.

[0073] Furthermore, in the subsequent stage of executing the hidden danger identification task, the hidden danger identification model will repeatedly perform the above-mentioned target detection task and mask prediction task. With the iteration of each stage, the generated binary mask map will become more and more accurate, and will be continuously passed to the target detection task branch of the next stage and the mask prediction task branch of the next stage, further improving the hidden danger identification model's ability to recognize vine-entangled areas.

[0074] Finally, based on the binary mask map output by the hidden danger identification model, it is possible to identify whether the transmission tower base is entangled by vines and obtain hidden danger identification results. Specifically, based on the binary mask map, the server can extract each independent vine object in the binary mask map through the connected domain analysis algorithm, and number and mark each vine object. For each vine object, the geometric parameters such as the area, length, width, and proportion of the vine object entangled with the transmission tower base are calculated. Combined with the spatial distribution position of the vine on the surface of the transmission tower base, it is determined whether each vine object constitutes a hidden danger and the hidden danger level to obtain the hidden danger identification result.

[0075] In this embodiment, through the multi-stage hidden danger identification task and its interactive embedding mechanism, the hidden danger identification model can gradually refine the bounding box and binary mask map, improve the target detection and mask prediction accuracy, thereby improving the recognition accuracy of the transmission tower base and the vine-entangled area, and improving the accuracy of the hidden danger identification results.

[0076] Following the above-mentioned embodiment, the hidden danger identification model continuously optimizes the hidden danger identification results of the transmission tower base through iterative operations at each stage. At each stage, new target detection and mask prediction are performed based on the target detection results and binary mask maps obtained in the previous stage as prior knowledge. In an exemplary embodiment, when the hidden danger identification model is called to perform the hidden danger identification task of each stage, the following steps are performed: using the target detection results obtained in the previous stage and the binary mask map obtained in the previous stage as prior knowledge, based on the pooling results, target detection is performed on the transmission tower base image, the target detection results are updated, and the updated target detection results and the binary mask map generated in the previous stage are used as prior knowledge to call the full convolution branch, perform mask prediction on the target transmission tower base image, and update the binary mask map.

[0077] Among them, in this embodiment, the prior knowledge refers to the target detection results and binary mask images output in the previous stage. These prior knowledge provide basic information for the current stage to help the hidden danger identification model perform target detection and mask prediction more efficiently and accurately.

[0078] Specifically, the target detection results generated in the previous stage usually include a bounding box that marks the location of the transmission tower base or plant in the target transmission tower base image in the current stage. This information can help the subsequent stages narrow the search range and reduce the computational complexity, so that the hidden danger identification model can focus more on the areas in the target transmission tower base image where it is known that the target may exist. The binary mask map generated by the mask prediction task in the previous stage clearly indicates which pixels in the target transmission tower base image belong to the target (such as plants or transmission tower base) and which belong to the background. It can be used as prior knowledge to provide information on the target shape and target boundary for the next stage. By using the above prior knowledge, the hidden danger identification model can continuously learn from the target detection and mask prediction in the previous stage, and gradually improve its understanding and recognition capabilities of the target.

[0079] Exemplarily, at each stage, the hidden danger identification model will perform target detection on the target transmission tower base image, which may be done by bounding box regression. In this process, the hidden danger identification model will fine-tune and correct the target detection results of the previous stage. This gradual optimization process can reduce the possibility of misjudgment or missed detection by the hidden danger identification model, especially in complex backgrounds or multi-scale targets. For example, plant vines may have different shapes and sizes, which are difficult to detect accurately at one time. After the target detection task, the hidden danger identification model will perform mask prediction on the target area selected by the bounding box based on the target detection results, and generate a binary mask map to complete it. In the binary mask map, the pixel points with a pixel value of "1" belong to the target area, and the pixel points with a pixel value of "0" belong to the background area. In the target area, each pixel point is distinguished as a different target, such as a transmission tower base and a plant.

[0080] It should be noted that the full convolution branch is called in the mask prediction process at each stage. Unlike traditional convolution operations, the full convolution can capture the global context information of the target transmission tower base image, thereby distinguishing the target from the background in a complex background. It is especially suitable for target transmission tower base images with complex backgrounds. Through the processing of the full convolution branch, the hidden danger identification model can more accurately generate binary mask maps, thereby improving the accuracy of hidden danger detection results.

[0081] In this embodiment, the hidden danger identification model is called to execute each stage of the hidden danger identification task, and the target detection results and binary mask images are gradually updated and optimized through the above series of steps. Through multiple iterations, each stage will have higher accuracy than the previous stage, thereby improving the accuracy of the hidden danger detection results.

[0082] In an exemplary embodiment, S200 includes at least one of the following: first, performing brightness enhancement on the original transmission tower base image to obtain a target transmission tower base image; second, performing contrast enhancement on the original transmission tower base image to obtain a target transmission tower base image; third, performing denoising on the original transmission tower base image to obtain a target transmission tower base image.

[0083] The original transmission tower base image is preprocessed, including but not limited to using a histogram equalization algorithm to enhance image contrast and brightness, and using a median filter to remove salt and pepper noise to obtain a clear target transmission tower base image.

[0084] Specifically, the original transmission tower base image can be enhanced in brightness and contrast by using an adaptive histogram equalization algorithm. According to the histogram distribution characteristics of the original transmission tower base image, the brightness distribution of the original transmission tower base image is adaptively adjusted to improve the overall contrast of the original transmission tower base image, thereby obtaining a transmission tower base image with enhanced brightness and contrast. Then, the transmission tower base image with enhanced brightness and contrast is denoised by using a median filtering algorithm. The median filtering algorithm replaces the gray value of the current pixel with the median value according to the median gray value of the pixel in the local area of ​​the image, thereby effectively removing salt and pepper noise and Gaussian noise in the image, thereby obtaining a denoised target transmission tower base image.

[0085] In this embodiment, by performing brightness enhancement, contrast enhancement, denoising and other operations on the original transmission tower base image, the image quality of the target transmission tower base image can be improved, providing a clearer and higher-quality identification basis for the subsequent hidden danger identification model, thereby improving the accuracy of hidden danger identification.

[0086] In an exemplary embodiment, Figure 5 As shown, after S300, the method further includes:

[0087] S410, when the hidden danger identification result indicates that the transmission tower base is entangled by plants, determining the coverage area of ​​the transmission tower base covered by the plants.

[0088] S420, determining the hidden danger level of the transmission tower foundation based on the plant coverage area.

[0089] S430, based on the coverage area and the hidden danger level, calling a preset hidden danger report template to generate a hidden danger report of the transmission tower foundation.

[0090] Exemplarily, based on the target binary mask map, it can be determined whether the transmission tower base is entangled by plants. For example, the plant mask in the target binary mask map represents the area where the transmission tower base is entangled by plants. By analyzing the position and shape of the plant mask, the coverage area of ​​the plants can be determined. When the coverage area of ​​the plants is greater than a preset area threshold, it is determined that there is a hidden danger in the transmission tower base.

[0091] Furthermore, by using the preset hidden danger level assessment rules and combining the coverage area of ​​the plants, the hidden danger level of the transmission tower base can be assessed, for example, the hidden danger level can be divided into mild hidden danger, moderate hidden danger and severe hidden danger. In addition, the hidden danger position of the transmission tower base can be analyzed according to the position and shape of the plant mask, and the hidden danger report template can be matched from the preset hidden danger report template library in combination with the coverage area of ​​the plants, the hidden danger level of the transmission tower base, and the hidden danger position. According to the hidden danger report template, the coverage area of ​​the plants, the hidden danger level of the transmission tower base, and the hidden danger position, a hidden danger description text is generated to obtain a hidden danger report of the transmission tower base.

[0092] In addition, if the hidden danger level of the transmission tower foundation is a severe hidden danger, the early warning mechanism will be triggered, the hidden danger report will be automatically pushed to the relevant responsible person, and a hidden danger handling work order will be generated. It is also possible to extract the key information from the hidden danger report of the transmission tower foundation through data visualization technology, generate an intuitive hidden danger distribution heat map and hidden danger trend map, so that the relevant responsible person can grasp the overall situation of the hidden danger of the transmission tower foundation. Furthermore, natural language processing technology can also be used to perform semantic analysis on the hidden danger description text in the hidden danger report, extract keywords, and then build a hidden danger knowledge graph based on the keywords, so that the staff can perform intelligent retrieval of hidden danger information through the hidden danger knowledge graph. It is also possible to associate the hidden danger report of the transmission tower foundation with the geographical location information of the transmission tower foundation, the power equipment information on the transmission line where the transmission tower foundation is located, etc., to form a complete transmission line hidden danger data set, which provides data support for subsequent hidden danger prediction, risk assessment and maintenance strategy optimization.

[0093] For example, if according to the target binary mask map, it is analyzed that the plant entanglement is located in the middle of the transmission tower base, and the mask surface area of ​​the plant accounts for 60% of the total coverage area of ​​the binary mask map, which exceeds the preset area threshold, the preset hidden danger level assessment rule is used to assess the hidden danger level of the transmission tower base as a moderate hidden danger, and then the plant entanglement position is mapped to the middle of the southeast side of the transmission tower base at a height of 1.5 meters through the geographic information system to determine the hidden danger position of the transmission tower base, and further generate a hidden danger description text "The middle of the southeast side of the transmission tower base is entangled by plants at a height of 1.5 meters, the degree of entanglement is 60%, and there is a moderate hidden danger risk", which is recorded in the preset hidden danger report template to generate a hidden danger report for the transmission tower base.

[0094] In this embodiment, combined with the binary mask image, it is possible to analyze whether there is a hidden danger of plant entanglement at the transmission tower base, as well as information such as the location of the hidden danger, the level of the hidden danger, etc., and to generate a detailed hidden danger report, providing a basis for subsequent hidden danger management.

[0095] Following the above embodiment, the training process of the hidden danger identification model involves a series of steps, including obtaining historical transmission tower base images, performing instance segmentation, and using these data to train the hybrid task cascade model. In an exemplary embodiment, the hidden danger identification model is trained based on the following steps: obtaining historical transmission tower base images, the historical transmission tower base images include transmission tower bases entangled by plants, performing instance segmentation on the historical transmission tower base images, marking the category label of each pixel in the historical transmission tower base images, obtaining the marked historical transmission tower base images, using the marked historical transmission tower base images as training data, training the hybrid task cascade model, and obtaining the hidden danger identification model.

[0096] Among them, historical transmission tower base images refer to transmission tower base images that have been collected and stored in the past, including transmission tower base images entangled by plants taken in various different situations, such as historical transmission tower base images entangled by plants taken in different climatic environments (such as rainy days, foggy weather, snowy days, etc.), and there can be many kinds of plants entangled in transmission tower bases, such as vines, branches, etc. Instance segmentation is different from traditional object detection. It not only detects objects in the image (such as transmission tower bases or entanglements), but also can accurately annotate the contours of each object to obtain pixel-level annotation information.

[0097] Then, according to the specific location and degree of the transmission tower base being entangled by plants, the historical transmission tower base image is instance segmented. Through instance segmentation, each pixel in the historical transmission tower base image will be assigned a category label. For example, the pixel points belonging to the plant part will be labeled as "plant", the pixel points belonging to the transmission tower base part will be labeled as "transmission tower base", and the pixel points belonging to the background part will be labeled as "background". In addition, the entire historical transmission tower base image can also carry annotation information such as the degree of hidden dangers, the location of hidden dangers, etc., to obtain the labeled historical transmission tower base image.

[0098] Finally, the annotated historical transmission tower base images are randomly divided into a training set and a validation set, where the training set is used to train the hidden danger identification model, and the validation set is used to verify the performance of the trained hidden danger identification model. Specifically, a deep learning algorithm is used to train the hybrid task cascade model based on the training set, and an end-to-end training method is used to optimize the model parameters through continuous iteration to obtain the trained hybrid task cascade model. The trained hybrid task cascade model is verified using the validation set, and the performance indicators of the model on the validation set are calculated to determine whether the trained hybrid task cascade model meets the expected requirements. If the performance of the trained hybrid task cascade model does not meet the requirements, the hybrid task cascade model continues to be trained and optimized until the performance of the trained hybrid task cascade model meets the requirements, and the final hidden danger identification model is obtained.

[0099] In this embodiment, the hybrid task cascade model is trained by annotating historical transmission tower base images, which can help the hybrid task cascade model learn the characteristics of whether there are hidden dangers in the transmission tower base, the location of the hidden dangers, and the level of hidden dangers, providing a basis for subsequent hidden danger identification.

[0100] In order to make a clearer description of the transmission tower foundation hidden danger identification method provided by the present application, the following is a specific embodiment and the attached Figure 6 To illustrate, this specific embodiment includes the following steps:

[0101] S601, obtaining an original transmission tower base image, performing brightness enhancement processing, contrast enhancement processing, and denoising operations on the original transmission tower base image, and obtaining a target transmission tower base image.

[0102] S602, extracting image features of the target transmission tower base image, performing a pooling operation on the image features, and obtaining a pooling result of the transmission tower base image.

[0103] S603, based on the pooling result, target detection is performed on the transmission tower base image to obtain a target detection result, and the target detection result is used as prior knowledge to call the full convolution branch to perform mask prediction on the target transmission tower base image to obtain a binary mask map.

[0104] S604, using the target detection result and the binary mask map as prior knowledge, performing a multi-stage hidden danger identification task, determining the target binary mask map, and based on the target binary mask map, identifying whether the transmission tower base is entangled by plants to obtain a hidden danger identification result.

[0105] S605: When the hidden danger identification result indicates that the transmission tower base is entangled by plants, determine the coverage area of ​​the transmission tower base covered by the plants.

[0106] S606, based on the coverage area, determine the hidden danger level of the transmission tower foundation, and based on the coverage area and the hidden danger level, call a preset hidden danger report template to generate a hidden danger report of the transmission tower foundation.

[0107] 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.

[0108] Based on the same inventive concept, the embodiment of the present application also provides a transmission tower foundation hidden danger device for implementing the above-mentioned transmission tower foundation hidden danger method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in one or more embodiments of the transmission tower foundation hidden danger device provided below can refer to the limitations of the transmission tower foundation hidden danger method above, and will not be repeated here.

[0109] In an exemplary embodiment, Figure 7As shown, a transmission tower foundation hidden danger identification device 700 is provided, comprising: a data acquisition module 710, a data preprocessing module 720 and a hidden danger identification module 730, wherein:

[0110] The data acquisition module 710 is used to acquire the original transmission tower base image.

[0111] The data preprocessing module 720 is used to preprocess the original transmission tower base image to obtain the target transmission tower base image.

[0112] The hidden danger identification module 730 is used to take the target transmission tower base image as input, call the preset hidden danger identification model, perform target detection and mask prediction on the transmission tower base image, so as to identify whether the transmission tower base is entangled by plants and obtain the hidden danger identification result.

[0113] Among them, the preset hidden danger identification model is obtained by training the hybrid task cascade model based on historical transmission tower base images carrying plant entanglement annotation information.

[0114] In one embodiment, the hidden danger identification model includes a full convolution branch, and the hidden danger identification module 730 is also used to extract image features of the target transmission tower base image, perform pooling operations on the image features, and obtain the pooling results of the transmission tower base image; based on the pooling results, target detection is performed on the transmission tower base image to obtain the target detection result; using the target detection result as prior knowledge, the full convolution branch is called to perform mask prediction on the target transmission tower base image to obtain a binary mask map; using the target detection result and the binary mask map as prior knowledge, a multi-stage hidden danger identification task is performed to determine the target binary mask map; based on the target binary mask map, it is identified whether the transmission tower base is entangled by plants to obtain a hidden danger identification result.

[0115] In one embodiment, the hidden danger identification module 730 is specifically used to use the target detection results obtained in the previous stage and the binary mask map obtained in the previous stage as prior knowledge, perform target detection on the transmission tower base image based on the pooling results, update the target detection results, use the updated target detection results and the binary mask map generated in the previous stage as prior knowledge, call the full convolution branch, perform mask prediction on the target transmission tower base image, and update the binary mask map.

[0116] In one embodiment, the transmission tower base hidden danger identification device 700 is also used to determine the coverage area of ​​the transmission tower base covered by plants when the hidden danger identification result indicates that the transmission tower base is entangled by plants, determine the hidden danger level of the transmission tower base based on the coverage area of ​​the plants, and call a preset hidden danger report template based on the coverage area and the hidden danger level to generate a hidden danger report on the transmission tower base.

[0117] In one embodiment, the transmission tower base hidden danger identification device 700 is also used to obtain historical transmission tower base images, which include transmission tower bases entangled by plants, perform instance segmentation on the historical transmission tower base images, mark the category label of each pixel in the historical transmission tower base images, and obtain the marked historical transmission tower base images. The marked historical transmission tower base images are used as training data to train the hybrid task cascade model to obtain a hidden danger identification model.

[0118] In one embodiment, the data preprocessing module 720 is specifically used to perform at least one of the following: first, perform brightness enhancement on the original transmission tower base image to obtain a target transmission tower base image; second, perform contrast enhancement on the original transmission tower base image to obtain a target transmission tower base image; third, perform denoising on the original transmission tower base image to obtain a target transmission tower base image.

[0119] Each module in the above-mentioned transmission tower foundation hidden danger identification device can be implemented in whole or in part 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 the above modules.

[0120] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. Figure 8 As shown. 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 the target transmission tower base image. 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 identifying hidden dangers in the transmission tower base is implemented.

[0121] Those skilled in the art will understand that Figure 8The structure shown in the figure is only a block diagram of a part of the 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 those shown in the figure, or combine certain components, or have a different arrangement of components.

[0122] 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 transmission tower foundation hazard identification methods are implemented.

[0123] 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 transmission tower foundation hazard identification methods are implemented.

[0124] 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 transmission tower foundation hidden danger identification methods.

[0125] 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.

[0126] 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.

[0127] 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.

[0128] 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 identifying hidden dangers of transmission tower foundation, characterized in that: The method comprises: Obtain the original transmission tower base image; Preprocessing the original transmission tower base image to obtain a target transmission tower base image; Taking the target transmission tower base image as input, calling a preset hidden danger identification model, performing target detection and mask prediction on the transmission tower base image, so as to identify whether the transmission tower base is entangled by plants, and obtaining a hidden danger identification result; Among them, the preset hidden danger identification model is obtained by training the hybrid task cascade model based on historical transmission tower base images carrying plant entanglement annotation information.

2. The method according to claim 1, characterized in that The hidden danger identification model includes a full convolution branch. When the hidden danger identification model is called, the following steps are performed: Extracting image features of the target transmission tower base image; Performing a pooling operation on the image features to obtain a pooling result of the transmission tower base image; Based on the pooling result, performing target detection on the transmission tower base image to obtain the target detection result; Using the target detection result as prior knowledge, calling the full convolution branch, performing mask prediction on the target transmission tower base image, and obtaining a binary mask map; Using the target detection result and the binary mask map as prior knowledge, a multi-stage hidden danger identification task is performed to determine a target binary mask map; Based on the target binary mask image, it is identified whether the transmission tower base is entangled by plants to obtain a hidden danger identification result.

3. The method according to claim 2, characterized in that The hidden danger identification model is called to perform the hidden danger identification task at each stage, and the following steps are performed: Using the target detection result obtained in the previous stage and the binary mask image obtained in the previous stage as prior knowledge, based on the pooling result, target detection is performed on the transmission tower base image, and the target detection result is updated; The updated target detection result and the binary mask map generated in the previous stage are used as prior knowledge, the full convolution branch is called, the mask prediction is performed on the target transmission tower base image, and the binary mask map is updated.

4. The method according to any one of claims 1 to 3, characterized in that: After obtaining the hidden danger identification result, the method further includes: In the case where the hidden danger identification result indicates that the transmission tower base is entangled by plants, determining a coverage area of ​​the transmission tower base covered by plants; Determining the hidden danger level of the transmission tower foundation based on the coverage area of ​​the plants; Based on the coverage area and the hidden danger level, a preset hidden danger report template is called to generate a hidden danger report of the transmission tower foundation.

5. The method according to any one of claims 1 to 3, characterized in that: The hidden danger identification model is trained based on the following steps: Acquire a historical transmission tower base image, wherein the historical transmission tower base image includes a transmission tower base entangled by plants; Performing instance segmentation on the historical transmission tower base image, marking the category label of each pixel in the historical transmission tower base image, and obtaining a marked historical transmission tower base image; The annotated historical transmission tower base images are used as training data to train the mixed task cascade model to obtain a hidden danger identification model.

6. The method according to any one of claims 1 to 3, characterized in that: The preprocessing of the original transmission tower base image to obtain the target transmission tower base image includes at least one of the following: The first item: performing brightness enhancement on the original transmission tower base image to obtain the target transmission tower base image; The second item is to perform contrast enhancement on the original transmission tower base image to obtain the target transmission tower base image; The third item: performing a denoising operation on the original transmission tower base image to obtain the target transmission tower base image.

7. A transmission tower foundation hidden danger identification device, characterized in that: The device comprises: A data acquisition module, used to acquire the original transmission tower base image; A data preprocessing module, used for preprocessing the original transmission tower base image to obtain a target transmission tower base image; A hidden danger identification module is used to use the target transmission tower base image as input, call a preset hidden danger identification model, perform target detection and mask prediction on the transmission tower base image, so as to identify whether the transmission tower base is entangled by plants, and obtain a hidden danger identification result; Among them, the preset hidden danger identification model is obtained by training the hybrid task cascade model based on historical transmission tower base images carrying plant entanglement annotation information.

8. 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 6 are implemented.

9. 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 6 are implemented.

10. A computer program product, comprising a computer program, 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 6 are implemented.

Images