An overhead line icing image labeling quality evaluation method, device and equipment

Abstract

The application discloses an overhead line icing image labeling quality evaluation method, device and equipment, and the method comprises the following steps: performing a pretreatment operation on the obtained overhead line image to obtain a set of images to be labeled; performing classification labeling processing on the images in the set of images to be labeled based on a preset labeling rule to obtain an icing image set, wherein the icing image set comprises a type label; after calculating the icing type fluctuation according to the type label, the labeling quality of the icing image set is evaluated through the total number of icing images and the icing type fluctuation to obtain an evaluation result. The application can solve the technical problem that the existing overhead line icing image relies on manual labeling, the labeling quality cannot be ensured, and the model trained by the image lacks reliability.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular to a method, apparatus and equipment for evaluating the quality of image annotation of overhead power line icing. Background Technology

[0002] Low-temperature freezing disasters have caused icing on overhead transmission lines of power grid companies, threatening the safe and stable operation of the power system. Power grid companies have installed numerous monitoring terminals for real-time monitoring of icing on overhead lines, each equipped with a camera to track the icing situation. Over the years, power grid companies have acquired a massive amount of icing images; however, when using deep learning to analyze these images, the annotation of these images has primarily relied on manual work. Due to the large number of icing images, the complexity of actual icing morphology, and the influence of human experience, the annotation quality varies greatly, resulting in a lack of reliability in models trained based on these images. Summary of the Invention

[0003] This application provides a method, apparatus, and equipment for evaluating the quality of annotations on icing images of overhead power lines, which addresses the technical problem that existing methods for annotating icing images of overhead power lines rely on manual annotation, which cannot ensure annotation quality and leads to a lack of reliability in the image training models.

[0004] In view of this, the first aspect of this application provides a method for evaluating the quality of image annotation of overhead power lines covered with ice, including:

[0005] The acquired overhead line images are preprocessed to obtain a set of images to be labeled;

[0006] Based on preset annotation rules, the images in the image set to be annotated are classified and labeled to obtain an icing image set, which includes type labels.

[0007] After calculating the icing type volatility based on the type labels, the quality of the annotation of the icing image set is evaluated by the total number of icing images and the icing type volatility, and the evaluation result is obtained.

[0008] Preferably, the preprocessing operation on the acquired overhead line images to obtain the image set to be labeled includes:

[0009] An initial image set was obtained by acquiring images of overhead lines based on time series data.

[0010] The initial image set is filtered according to the preset spatiotemporal impact range of the cold wave to obtain a filtered image set;

[0011] The filtered image set is subjected to invalid image cleaning operation to obtain the image set to be labeled.

[0012] Preferably, the step of classifying and labeling the images in the image set to be labeled based on preset labeling rules to obtain an icing image set further includes:

[0013] The icing image set is stored as a CSV file in CSV format.

[0014] Preferably, after calculating the icing type variability based on the type label, the quality assessment of the annotation of the icing image set is performed using the total number of icing images and the icing type variability to obtain the assessment result, including:

[0015] Count the number of images of different types of icing based on the type labels;

[0016] Based on the type label, the numerical value corresponding to the icing type fluctuation is calculated using the number of images;

[0017] The quality of the annotation of the icing image set is evaluated based on the total number of icing images, the numerical value corresponding to the icing type fluctuation, and the quantity threshold, and the evaluation result is obtained.

[0018] Preferably, if the evaluation result is:

[0019] If the total number of icing images is less than the number threshold, and the value corresponding to the icing type fluctuation is within a preset strong fluctuation range, then the images corresponding to the icing image set will be reviewed to obtain the review result.

[0020] The second aspect of this application provides a device for evaluating the quality of image annotation of icing on overhead power lines, comprising:

[0021] The preparation unit is used to preprocess the acquired overhead line images to obtain a set of images to be labeled.

[0022] The annotation unit is used to classify and annotate the images in the image set to be annotated based on preset annotation rules to obtain an icing image set, the icing image set including type labels;

[0023] The evaluation unit is used to calculate the icing type volatility based on the type label, and then evaluate the quality of the annotation of the icing image set by the total number of icing images and the icing type volatility, and obtain the evaluation result.

[0024] Preferably, the preparation unit is specifically used for:

[0025] An initial image set was obtained by acquiring images of overhead lines based on time series data.

[0026] The initial image set is filtered according to the preset spatiotemporal impact range of the cold wave to obtain a filtered image set;

[0027] The filtered image set is subjected to invalid image cleaning operation to obtain the image set to be labeled.

[0028] Preferably, it further includes:

[0029] A storage unit is used to store the icing image set as a CSV file in CSV format.

[0030] Preferably, the evaluation unit is specifically used for:

[0031] Count the number of images of different types of icing based on the type labels;

[0032] Based on the type label, the numerical value corresponding to the icing type fluctuation is calculated using the number of images;

[0033] The quality of the annotation of the icing image set is evaluated based on the total number of icing images, the numerical value corresponding to the icing type fluctuation, and the quantity threshold, and the evaluation result is obtained.

[0034] A third aspect of this application provides a quality assessment device for image annotation of icing on overhead power lines, the device including a processor and a memory;

[0035] The memory is used to store program code and transmit the program code to the processor;

[0036] The processor is used to execute the overhead line icing image annotation quality assessment method described in the first aspect according to the instructions in the program code.

[0037] As can be seen from the above technical solutions, the embodiments of this application have the following advantages:

[0038] This application provides a method for evaluating the quality of annotation of overhead line icing images, including: preprocessing the acquired overhead line images to obtain a set of images to be annotated; classifying and annotating the images in the set of images to be annotated based on preset annotation rules to obtain an icing image set, the icing image set including type labels; calculating the icing type fluctuation based on the type labels, and then evaluating the quality of the annotation of the icing image set by the total number of icing images and the icing type fluctuation to obtain the evaluation result.

[0039] The method for evaluating the quality of overhead power line icing image annotations provided in this application performs calculations and analysis on the annotated overhead power line images to obtain the corresponding icing type fluctuations. Then, it conducts targeted quality assessments of the images based on the total number of icing images and the icing type fluctuations. The evaluation is based on quantitative parameters of image distribution characteristics, making the evaluation results more reasonable and reliable. This allows for accurate control of image annotation quality, thereby controlling the quality of images used in training models. Therefore, this application solves the technical problem that existing methods for evaluating overhead power line icing images rely on manual annotation, which cannot guarantee annotation quality and leads to a lack of reliability in image-trained models. Attached Figure Description

[0040] Figure 1 A flowchart illustrating a method for evaluating the quality of image annotation of icing on overhead power lines, provided in an embodiment of this application;

[0041] Figure 2 This is a schematic diagram of the structure of an overhead power line icing image annotation quality assessment device provided in an embodiment of this application. Detailed Implementation

[0042] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0043] For easier understanding, please refer to Figure 1 An embodiment of the image annotation quality assessment method for icing of overhead power lines provided in this application includes:

[0044] Step 101: Perform preprocessing on the acquired overhead line images to obtain a set of images to be labeled.

[0045] Further, step 101 includes:

[0046] An initial image set was obtained by acquiring images of overhead lines based on time series data.

[0047] Based on the preset spatiotemporal impact range of the cold wave, the initial image set is filtered to obtain the filtered image set;

[0048] The selected image set is cleaned of invalid images to obtain the image set to be labeled.

[0049] Overhead line images are collected by multiple different monitoring terminals on the power grid. Storage folders for overhead line images can be built according to the monitoring terminals. Since image acquisition is based on time series, images can also be arranged according to time series during storage to facilitate subsequent analysis and processing.

[0050] The preprocessing operations in this embodiment mainly include image screening and image cleaning. Other preprocessing operations can be added according to the actual image conditions, which are not limited here. The preset spatiotemporal impact range of cold waves refers to the start and end times of cold waves in previous years and the geographical areas affected, forming a quantitative spatiotemporal impact range.

[0051] In this embodiment, image filtering primarily involves preliminary screening of images in the initial image set based on a preset temporal and spatial impact range of a cold wave. Images outside the impact range of that year's cold wave are deleted, including both temporal and spatial ranges, resulting in a filtered image set. Image cleaning mainly involves selecting images that show electrical equipment such as insulators or conductors, and deleting images that do not contain these elements. Then, invalid images caused by black screens, lens icing, etc., are removed, resulting in the image set to be labeled. These preprocessing operations improve image quality and ensure the effectiveness of subsequent labeling and evaluation.

[0052] Step 102: Classify and label the images in the image set to be labeled based on the preset labeling rules to obtain the icing image set, which includes type labels.

[0053] The preset labeling rules in this embodiment define five types of image type labels, including rime, hoarfrost, mixed rime, wet snow, and no ice. An image database can be built based on these five image categories. Since the images all come from specific monitoring terminals, the database can include not only information such as icing type / type label and image file name, but also terminal number.

[0054] An automatic annotation tool can be used to automatically annotate the images in the image set to be annotated in this embodiment. When annotating the next image, the default label is the label of the previous image. It is then determined whether the label is correct. If it is correct, the annotation continues to the next image; otherwise, the current image is re-annotated. Since icing cold waves generally last for several calendar days, adjacent images are very likely to belong to the same type of icing. Therefore, a default judgment method can be used for continuous annotation.

[0055] Furthermore, step 102, followed by:

[0056] The icing image set is stored as a CSV file in CSV format.

[0057] Step 103: After calculating the icing type variability based on the type labels, the quality of the annotation of the icing image set is evaluated by the total number of icing images and the icing type variability, and the evaluation results are obtained.

[0058] Further, step 103 includes:

[0059] Count the number of images of different types of icing based on type labels;

[0060] The numerical value corresponding to the icing type fluctuation is calculated based on the number of images using type labels;

[0061] The quality of the annotation of the icing image set is evaluated using a judgmental approach based on the total number of icing images, the numerical values ​​corresponding to the icing type fluctuations, and the quantity threshold, and the evaluation results are obtained.

[0062] Furthermore, if the evaluation result is:

[0063] If the total number of icing images is less than the number threshold, and the value corresponding to the icing type fluctuation is within the preset strong fluctuation range, then the images corresponding to the icing image set will be re-verified to obtain the verification result.

[0064] The formation mechanisms of the four types of icing on transmission lines—rime, hoarfrost, mixed rime, and wet snow—are different, and they exhibit spatiotemporal distribution patterns. For the same icing monitoring terminal, during the low-temperature freezing disasters of each winter and the following spring, not many icing types are observed; generally, only one or two types are seen, with very few or no images of other icing types. Based on these patterns, the annotation quality of each terminal can be evaluated.

[0065] Based on category labels, the number of images of rime, hoarfrost, mixed rime, and wet snow under each monitoring terminal can be counted, denoted as C, R, M, and S. Then, the total number of icing images can be expressed as:

[0066] X = C + R + M + S

[0067] It can be seen that the total number of icy images is the number of images with ice, excluding images without ice.

[0068] The icing type volatility can be calculated as follows:

[0069]

[0070] in,

[0071]

[0072] Furthermore, the value range of Y is [1,4]. 1≤Y≤2 is defined as the preset weak fluctuation range, and 2≤Y≤4 is defined as the preset strong fluctuation range. It should be noted that the numerical calculation of the corresponding fluctuation of the icing type is based on batches of images. In this embodiment, each terminal is considered as a batch, but different batches can also be defined according to the actual situation.

[0073] The quality assessment of the annotation of an icing image set based on the total number of icing images, the numerical values ​​corresponding to the variability of icing types, and the quantity threshold can be mainly divided into four cases:

[0074] When the total number of icing images X exceeds the threshold, and the icing type fluctuation is 1≤Y≤2, it indicates that the icing time in the images acquired by this type of terminal is relatively long, and the icing type experienced is relatively simple, or the frequency of a certain icing type is much higher than that of other icing types. This situation is consistent with the pattern of icing events on transmission lines and does not require further review.

[0075] When the total number of icing images X is greater than the threshold, and the icing type fluctuation is 2≤Y≤4, it indicates that the icing time in the icing images acquired by this type of terminal is relatively long, the icing types are numerous, and the frequency of occurrence is relatively average. Considering that multiple icing types may switch between each other in actual icing events, the icing type fluctuation is relatively large. Therefore, this situation also conforms to the general pattern of the icing process and does not require verification.

[0076] When the total number of icing images X is less than the threshold, and the icing type fluctuation is 1≤Y≤2, it can be concluded that rime and wet snow have weak adhesion to transmission lines and are easily removed by slight vibrations and wind. If the icing is mild, a very small number of images may appear to show icing. Furthermore, brief instances of rime icing have been observed in actual annotations; considering the hardness and strong adhesion of rime, it is speculated that artificial de-icing was performed. The above discussion demonstrates that brief icing processes are possible, and this data characteristic is consistent with the actual conditions of transmission lines on-site, thus requiring no further verification.

[0077] When the total number of icing images X is less than the threshold, and the icing type fluctuation is 2≤Y≤4, it indicates that the terminal detected multiple brief periods of different icing processes on the line. This does not conform to the principle that multiple icing types should not occur on the same terminal within the same time period. Terminals with a small total number of icing images and large fluctuations in icing types indicate poor annotation quality and require close review.

[0078] It should be noted that the quantity threshold can be set according to the actual situation and is not limited here. This embodiment only provides an example, setting the quantity threshold to 30. When the total number of icing images X is less than 30, images from folders in the corresponding image set whose icing type fluctuation is in the range of 2≤Y≤4 are selected for review. The focus of the review is to check the images with fewer icing types labeled as the above four types. Moreover, images that do not need to be reviewed are labeled as having relatively accurate labels and are qualified images.

[0079] The overhead power line icing image annotation quality assessment method provided in this application analyzes and calculates the annotated overhead power line images to obtain the corresponding icing type fluctuation. Then, it performs a targeted quality assessment based on the total number of icing images and the icing type fluctuation. The assessment is based on quantitative parameters of image distribution characteristics, making the assessment results more reasonable and reliable. This allows for accurate control of image annotation quality, thereby controlling the quality of training model images. Therefore, this application solves the technical problem that existing overhead power line icing image annotation relies on manual annotation, which cannot guarantee annotation quality and leads to a lack of reliability in image-trained models.

[0080] For easier understanding, please refer to Figure 2 This application provides an embodiment of an overhead power line icing image annotation quality assessment device, comprising:

[0081] Preparation unit 201 is used to preprocess the acquired overhead line images to obtain a set of images to be labeled;

[0082] The annotation unit 202 is used to classify and annotate the images in the image set to be annotated based on preset annotation rules to obtain an ice-covered image set, which includes type labels;

[0083] Evaluation unit 203 is used to calculate the icing type variability based on the type label, and then evaluate the quality of the annotation of the icing image set by the total number of icing images and the icing type variability, and obtain the evaluation result.

[0084] Furthermore, preparation unit 201 is specifically used for:

[0085] An initial image set was obtained by acquiring images of overhead lines based on time series data.

[0086] Based on the preset spatiotemporal impact range of the cold wave, the initial image set is filtered to obtain the filtered image set;

[0087] The selected image set is cleaned of invalid images to obtain the image set to be labeled.

[0088] Furthermore, it also includes:

[0089] Storage unit 204 is used to store the icing image set as a CSV file in CSV format.

[0090] Furthermore, the evaluation unit 203 is specifically used for:

[0091] Count the number of images of different types of icing based on type labels;

[0092] The numerical value corresponding to the icing type fluctuation is calculated based on the number of images using type labels;

[0093] The quality of the annotation of the icing image set is evaluated using a judgmental approach based on the total number of icing images, the numerical values ​​corresponding to the icing type fluctuations, and the quantity threshold, and the evaluation results are obtained.

[0094] This application also provides a quality assessment device for image annotation of icing on overhead power lines, the device including a processor and a memory;

[0095] The memory is used to store program code and transfer the program code to the processor;

[0096] The processor is used to execute the overhead line icing image annotation quality assessment method in the above method embodiment according to the instructions in the program code.

[0097] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0098] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0099] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0100] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for executing all or part of the steps of the methods described in the various embodiments of this application through a computer device (which may be a personal computer, server, or network device, etc.). The aforementioned storage medium includes: USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, optical disks, and other media capable of storing program code.

[0101] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for evaluating the quality of image annotation of icing on overhead power lines, characterized in that, include: The acquired overhead line images are preprocessed to obtain an image set to be labeled. The specific process is as follows: An initial image set was obtained by acquiring images of overhead lines based on time series data. The initial image set is filtered according to the preset spatiotemporal impact range of the cold wave to obtain a filtered image set; The filtered image set is subjected to invalid image cleaning operation to obtain the image set to be labeled; Based on preset annotation rules, the images in the image set to be annotated are classified and labeled to obtain an icing image set, which includes type labels. After calculating the icing type variability based on the type labels, the quality of the annotation of the icing image set is evaluated using the total number of icing images and the icing type variability, and the evaluation result is obtained. The specific process is as follows: Count the number of images of different types of icing based on the type labels; Based on the type label, the numerical value corresponding to the icing type volatility is calculated using the number of images. The specific process of volatility-related calculation is expressed as follows: Based on type labels, the number of images of rime, hoarfrost, mixed rime, and wet snow under each monitoring terminal was counted and denoted as follows: C , R , M , S The total number of icing images can then be expressed as: It can be seen that the total number of icing images X This refers to the number of images showing icing. The numerical values ​​corresponding to the icing type fluctuations are calculated as follows: in, and Y The range of values ​​is , Defined as a preset weak fluctuation range, Defined as a preset strong fluctuation range; The quality of the annotation of the icing image set is evaluated based on the total number of icing images, the numerical value corresponding to the icing type fluctuation, and the quantity threshold, and the evaluation result is obtained. If the evaluation result is that the total number of icing images is less than the number threshold, and the value corresponding to the icing type fluctuation is within a preset strong fluctuation range, then the images corresponding to the icing image set will be re-verified to obtain the re-verification result.

2. The method for evaluating the quality of image annotation of overhead power lines icing according to claim 1, characterized in that, The process of classifying and labeling images in the image set to be labeled based on preset labeling rules to obtain an icing image set, further includes: The icing image set is stored as a CSV file in CSV format.

3. A quality assessment device for image annotation of icing on overhead power lines, characterized in that, include: The preparation unit is used to preprocess the acquired overhead line images to obtain a set of images to be labeled. Specifically, the preparation unit is used for: An initial image set was obtained by acquiring images of overhead lines based on time series data. The initial image set is filtered according to the preset spatiotemporal impact range of the cold wave to obtain a filtered image set; The filtered image set is subjected to invalid image cleaning operation to obtain the image set to be labeled; The annotation unit is used to classify and annotate the images in the image set to be annotated based on preset annotation rules to obtain an icing image set, the icing image set including type labels; An evaluation unit is configured to calculate the icing type variability based on the type labels, and then perform a quality assessment of the annotation of the icing image set using the total number of icing images and the icing type variability to obtain an evaluation result. Specifically, the evaluation unit is configured to: Count the number of images of different types of icing based on the type labels; Based on the type label, the numerical value corresponding to the icing type volatility is calculated using the number of images. The specific process of volatility-related calculation is expressed as follows: Based on type labels, the number of images of rime, hoarfrost, mixed rime, and wet snow under each monitoring terminal was counted and denoted as follows: C , R , M , S The total number of icing images can then be expressed as: It can be seen that the total number of icing images X This refers to the number of images showing icing. The numerical values ​​corresponding to the icing type fluctuations are calculated as follows: in, and Y The range of values ​​is , Defined as a preset weak fluctuation range, Defined as a preset strong fluctuation range; The quality of the annotation of the icing image set is evaluated based on the total number of icing images, the numerical value corresponding to the icing type fluctuation, and the quantity threshold, and the evaluation result is obtained. If the evaluation result is that the total number of icing images is less than the number threshold, and the value corresponding to the icing type fluctuation is within the preset strong fluctuation range, then the images corresponding to the icing image set will be re-verified to obtain the re-verification result.

4. The overhead line icing image annotation quality assessment device according to claim 3, characterized in that, Also includes: A storage unit is used to store the icing image set as a CSV file in CSV format.

5. A quality assessment device for image annotation of icing on overhead power lines, characterized in that, The device includes a processor and a memory; The memory is used to store program code and transmit the program code to the processor; The processor is used to execute the overhead line icing image annotation quality assessment method according to any one of claims 1-2 according to the instructions in the program code.

Images