Transmission line single-image distance measurement method and system based on three-dimensional point cloud

Abstract

The present invention discloses a single-image distance measurement method and system for power transmission lines based on three-dimensional point clouds, and relates to the field of single-image distance measurement technology. The method includes: obtaining a target scanning video; reading an image preprocessing plan to obtain a first two-dimensional target image; calling an intelligent annotation model for annotation analysis to obtain first annotation information; constructing a three-dimensional channel model to obtain a visible graph of the power transmission channel; obtaining any hazard source in the visible graph of the power transmission channel, and obtaining any peak point of the any hazard source; and obtaining the target distance in combination with the any peak point. The present invention solves the technical problem in the prior art that the distance measurement between the conductor and the hazard source in the power transmission line lacks accuracy, resulting in safety risks in the power transmission line, and achieves the technical effect of improving the accuracy and reliability of the distance measurement between the conductor and the hazard source in the power transmission line and improving the safety of the power transmission line.

Description

Technical Field

[0001] The present invention relates to the technical field of single-image distance measurement, and in particular to a single-image distance measurement method and system for power transmission lines based on three-dimensional point clouds. Background Art

[0002] In today's society, electricity is an important energy support, and the safe and stable operation of transmission lines is of vital importance. With the continuous expansion of the scale of the power grid, transmission lines often cross complex geographical environments and face various potential dangers. Traditional transmission line distance measurement methods are insufficient in accuracy, and it is difficult to accurately determine the distance between the conductor and the source of danger. For example, some methods that rely on manual inspections are inefficient and prone to errors, while some methods based on simple image analysis are difficult to fully consider the actual situation in three-dimensional space. This makes the distance measurement between the conductor and the source of danger in the transmission line lack accuracy in practical applications, and it is impossible to timely and effectively discover potential safety risks, which brings huge challenges to the safe operation of the transmission line.

[0003] The existing technology has a technical problem of lack of accuracy in measuring the distance between the conductors and the dangerous sources in the transmission line, resulting in safety risks in the transmission line. Summary of the invention

[0004] The present application provides a single-image distance measurement method and system for power transmission lines based on three-dimensional point clouds, which is used to solve the technical problem in the prior art of lack of accuracy in distance measurement between conductors and dangerous sources in power transmission lines, resulting in safety risks in power transmission lines.

[0005] In view of the above problems, the present application provides a single-image distance measurement method and system for power transmission lines based on three-dimensional point clouds.

[0006] In a first aspect of the present application, a single-image distance measurement method for a power transmission line based on a three-dimensional point cloud is provided, the method comprising:

[0007] Acquire a target scanning video, wherein the target scanning video refers to a video obtained by dynamically scanning a target power transmission channel through a target camera device; read an image preprocessing plan, and process a first two-dimensional scanning image in the target scanning video according to the image preprocessing plan to obtain a first two-dimensional target image; call up an intelligent annotation model to perform annotation analysis on the first two-dimensional target image to obtain first annotation information, wherein the first annotation information includes a first wire area and a first hazard source; construct a three-dimensional channel model according to the channel feature information of the target power transmission channel collected by a laser radar device; map the first two-dimensional target image annotated with the first wire area and the first hazard source to the three-dimensional channel model to obtain a visible graph of the power transmission channel; obtain any hazard source in the visible graph of the power transmission channel, and obtain any peak point of the any hazard source; obtain a target distance in combination with the any peak point, wherein the target distance refers to the spatial distance between the wire in the target power transmission channel and the any hazard source.

[0008] The second aspect of the present application provides a transmission line single-image ranging system based on three-dimensional point cloud, the system comprising:

[0009] a target scanning video acquisition module, the target scanning video acquisition module is used to acquire a target scanning video, the target scanning video refers to a video obtained by dynamically scanning a target power transmission channel through a target camera device; a first two-dimensional target image acquisition module, the first two-dimensional target image acquisition module is used to read an image preprocessing plan, and process the first two-dimensional scanning image in the target scanning video according to the image preprocessing plan to obtain a first two-dimensional target image; a first annotation information acquisition module, the first annotation information acquisition module is used to call an intelligent annotation model to perform annotation analysis on the first two-dimensional target image to obtain first annotation information, the first annotation information includes a first conductor area and a first hazard source; a three-dimensional channel model construction module, the three A three-dimensional channel model construction module is used to construct a three-dimensional channel model according to the channel feature information of the target transmission channel collected by the laser radar equipment; a transmission channel visible map acquisition module, the transmission channel visible map acquisition module is used to map the first two-dimensional target image marked with the first wire area and the first hazard source to the three-dimensional channel model to obtain a transmission channel visible map; an arbitrary peak point acquisition module, the arbitrary peak point acquisition module is used to obtain any hazard source in the transmission channel visible map, and obtain any peak point of the arbitrary hazard source; a target distance acquisition module, the target distance acquisition module is used to obtain the target distance in combination with the arbitrary peak point, and the target distance refers to the spatial distance between the wire in the target transmission channel and the arbitrary hazard source.

[0010] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0011] Obtain target scanning video; read image preprocessing plan, and process the first two-dimensional scanning image in the target scanning video according to the image preprocessing plan to obtain the first two-dimensional target image; call the intelligent annotation model to annotate and analyze the first two-dimensional target image to obtain the first annotation information; construct a three-dimensional channel model according to the channel characteristic information of the target transmission channel collected by the laser radar equipment; map the first two-dimensional target image annotated with the first wire area and the first hazard source to the three-dimensional channel model to obtain a visible diagram of the transmission channel; obtain any hazard source in the visible diagram of the transmission channel, and obtain any peak point of the any hazard source; obtain the target distance in combination with the any peak point. The technical effect of improving the accuracy and reliability of the distance measurement between the wire and the hazard source in the transmission line and improving the safety of the transmission line is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0013] Figure 1 A schematic diagram of the flow of a single-image distance measurement method for a power transmission line based on three-dimensional point cloud provided in an embodiment of the present application;

[0014] Figure 2 A schematic diagram of the structure of a single-image distance measurement system for power transmission lines based on three-dimensional point clouds provided in an embodiment of the present application.

[0015] Explanation of the reference numerals: target scanning video acquisition module 10, first two-dimensional target image acquisition module 20, first annotation information acquisition module 30, three-dimensional channel model construction module 40, transmission channel visual map acquisition module 50, arbitrary peak point acquisition module 60, target distance acquisition module 70. DETAILED DESCRIPTION

[0016] The present application provides a single-image distance measurement method and system for power transmission lines based on three-dimensional point clouds, which is used to solve the technical problem in the prior art of lack of accuracy in measuring the distance between conductors and hazardous sources in power transmission lines, resulting in safety risks in power transmission lines.

[0017] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0018] Embodiment 1, as Figure 1 As shown, the present application provides a single-image distance measurement method for power transmission lines based on three-dimensional point clouds, the method comprising:

[0019] Step S100: Acquire a target scanning video, where the target scanning video refers to a video obtained by dynamically scanning a target power transmission channel through a target camera device.

[0020] Specifically, the target transmission channel is dynamically scanned by a specific target camera device. During this process, the camera device continuously records the image information of the target transmission channel, forming a video over time, namely the target scanning video. The video contains various states and characteristics of the target transmission channel at different time points, providing a basic data source for subsequent analysis and processing.

[0021] Step S200: reading an image preprocessing plan, and processing the first two-dimensional scanning image in the target scanning video according to the image preprocessing plan to obtain a first two-dimensional target image.

[0022] Specifically, the pre-set image preprocessing plan is read, and the specific content of the image preprocessing plan is to perform pixel rasterization processing on the first two-dimensional scanned image. Then, the first two-dimensional scanned image is extracted from the target scanning video, and the first two-dimensional scanned image is processed according to the image preprocessing plan. The processing process includes geometric correction to correct the deformation and distortion in the image; de-distortion processing to eliminate image distortion caused by the optical characteristics of the camera equipment and other reasons; and also includes adjusting the brightness, contrast, color balance and other operations of the image to improve the quality and readability of the image. Through these processing steps, the first two-dimensional scanned image is converted into a first two-dimensional target image that is more suitable for subsequent analysis and annotation. This target image has more accurate geometry, clearer details and more suitable visual effects, which lays a good foundation for subsequent intelligent annotation and three-dimensional model construction.

[0023] Step S300: calling the intelligent labeling model to perform labeling analysis on the first two-dimensional target image to obtain first labeling information, where the first labeling information includes a first wire area and a first hazard source.

[0024] Specifically, the intelligent annotation model is called to annotate and analyze the first two-dimensional target image. This intelligent annotation model is a recognition model built based on a convolutional neural network and has powerful image recognition and annotation capabilities. First, the model will comprehensively analyze the input first two-dimensional target image. It uses the multi-layer structure of the convolutional neural network to automatically extract various features in the image. For the transmission wire in the image, the model will accurately mark the first wire area by identifying the specific shape, texture, and color of the wire. For example, the wire usually has a slender shape and a specific color, and the model will identify and annotate according to these features. At the same time, the model will also focus on detecting the first hazard source in the image, which includes cranes, crane booms, excavators, bulldozers, and cement pump trucks. The operation of these devices near the transmission channel will pose a potential threat to the safety of the transmission line. The model will determine their location in the image by identifying the unique features of these devices, such as shape, color, and logos, and mark them as the first hazard source. Through such an annotation analysis process, the model finally outputs the first annotation information, which clearly includes the first wire area and the first hazard source. These annotation information provides key data support for subsequent steps such as 3D model construction and distance measurement, which helps to better assess the safety status of the transmission channel, timely detect potential dangers and take corresponding measures.

[0025] Step S400: constructing a three-dimensional channel model according to the channel characteristic information of the target power transmission channel collected by the laser radar equipment.

[0026] Specifically, firstly, laser radar equipment is used to collect data on the target transmission channel and its surrounding environment. By emitting laser beams and receiving reflected signals, the laser radar equipment can accurately obtain the three-dimensional point cloud data of the target area. These point cloud data contain rich channel feature information of the transmission channel and its surrounding environment. After the point cloud data is collected, it needs to be denoised. Due to the interference of various environmental factors during the collection process, there will be some noise points in the point cloud data. Through a specific noise reduction algorithm, these noise points are removed to improve the quality and accuracy of the data. The point cloud data after noise reduction is purer and can better reflect the true form of the target transmission channel. Then, the processed high-precision point cloud data is used to construct a three-dimensional channel model. According to the spatial distribution and characteristics of the point cloud, the advanced three-dimensional modeling technology is used to gradually construct the three-dimensional structure of the target transmission channel. In the construction process, the details of the transmission channel will be fully considered, such as the support structure of the conductor, the shape of the tower, and the influence of the surrounding terrain. By continuously integrating and optimizing the point cloud data, a high-precision three-dimensional channel model is finally generated. This model can intuitively present the actual spatial form of the target transmission channel, providing a reliable basis for subsequent analysis and application, and helping to more accurately perform distance measurement, safety assessment, planning and design, etc.

[0027] Step S500: Mapping the first two-dimensional target image marked with the first wire area and the first hazard source to the three-dimensional channel model to obtain a visible diagram of the power transmission channel.

[0028] Specifically, based on the feature matching algorithm and the iterative closest point (ICP) algorithm, the first two-dimensional target image marked with the first conductor area and the first hazard source is registered with the three-dimensional channel model constructed by the laser radar equipment. The ICP algorithm gradually achieves accurate registration by continuously searching for corresponding feature points in the two-dimensional image and the three-dimensional point cloud and minimizing the distance error between them. In this process, the algorithm analyzes the geometric shape, texture and other features in the image and point cloud to determine the best matching relationship. Based on the registration results, the transmission lines and hazard sources marked in the two-dimensional target image are accurately mapped to the three-dimensional point cloud data. This mapping process needs to consider factors such as coordinate conversion, scaling and perspective differences between the two-dimensional image and the three-dimensional point cloud. Through precise calculation and adjustment, it is ensured that the annotation information can be correctly positioned in the three-dimensional space. Finally, the transmission channel visible map obtained combines the annotation information of the two-dimensional image and the spatial structure advantages of the three-dimensional point cloud. In this visible map, the specific position and form of the first conductor area and the first hazard source marked in the three-dimensional space can be clearly seen. This provides an intuitive and comprehensive visualization tool for subsequent safety assessment and distance measurement of transmission channels, enabling staff to better understand and analyze the status of transmission channels, promptly identify potential safety hazards, and take corresponding maintenance and management measures.

[0029] Step S600: obtaining any hazard source in the visible graph of the power transmission channel, and obtaining any peak point of the hazard source.

[0030] Specifically, first, the hazard source is selected from the generated transmission channel visual map. The transmission channel visual map integrates the annotation information of the two-dimensional image and the spatial structure of the three-dimensional point cloud model, and can clearly show the specific position and form of the annotated first conductor area and the first hazard source in the three-dimensional space. In this visual map, the staff can intuitively observe the distribution of each hazard source. Then, an arbitrary hazard source is randomly selected from the numerous hazard sources. This hazard source may be the area where trees, buildings or other objects that may pose a threat to power transmission safety are located near the transmission line. After determining the arbitrary hazard source, further obtain any peak point in the area. The peak point here refers to the point closest to the wire, which is crucial for calculating the distance between the wire and the hazard source. By accurately finding this peak point, the potential threat of the hazard source to the transmission line can be more accurately assessed. After finding the peak point, the next step is to calculate the distance between it and the wire. The calculation of this distance is usually based on the principle of spatial geometry, considering the coordinate position of the peak point and the wire in three-dimensional space. Through accurate calculation, a specific value can be obtained, representing the spatial distance between the wire and the hazard source. Such operations help to promptly identify potential safety hazards and provide a basis for taking appropriate measures to ensure the safe operation of the transmission channel.

[0031] Step S700: obtaining a target distance in combination with the arbitrary peak point, wherein the target distance refers to the spatial distance between the conductor in the target power transmission channel and the arbitrary hazard source.

[0032] Specifically, the determination of the target distance is based on the analysis of any peak point. Any peak point is a specific point of any hazardous source obtained from the visible image of the transmission channel. By analyzing the position and characteristics of these arbitrary peak points, the spatial distance between the conductor and any hazardous source in the target transmission channel, that is, the target distance, can be calculated. First, it is necessary to determine the position and direction of the conductor in the three-dimensional space, and obtain relevant information through the previous annotation analysis of the first two-dimensional target image and the construction of the three-dimensional channel model. Then, for any peak point of any hazardous source, the straight-line distance or the shortest spatial path between them is calculated by using the spatial geometry calculation method, considering the coordinates of the peak point and the position information of the conductor. The calculation of this target distance is very important because it is directly related to the safe operation of the transmission line. If the spatial distance between the conductor and the hazardous source is too close, there will be safety hazards, such as short circuits, discharges and other accidents. Therefore, accurately calculating the target distance can help relevant personnel take timely measures, such as adjusting the position of the hazardous source, strengthening protective measures, etc., to ensure the safe and stable operation of the transmission line.

[0033] In one possible implementation, step 200 further includes:

[0034] Step 210: compress the target scanned video using a dynamic picture expert group to obtain a target compressed video, wherein the target compressed video includes a plurality of compression encoding units.

[0035] Step 220: Build a target candidate scanning image set based on the first target frame image extracted from the first compression coding unit, wherein the first compression coding unit is any one of the multiple compression coding units.

[0036] Step 230: randomly extract any frame of image from the target candidate scan image set and record it as the first two-dimensional scan image.

[0037] Specifically, the target scanning video is compressed using the technology of the Moving Picture Experts Group (MPEG). The Moving Picture Experts Group has developed a series of video compression standards that can significantly reduce the data volume of video files while maintaining video quality. For the target scanning video, the video is obtained by dynamically scanning the target power transmission channel through the target camera device. Since the video data is usually very large, direct processing will take up a lot of storage space and computing resources. Therefore, compression using MPEG technology is an effective solution. During the compression process, the video is divided into multiple parts, each of which is encoded into a compression coding unit. These compression coding units contain key information of the video, such as image color, brightness, texture and other features, as well as their changes in time and space. In this way, the target scanning video is compressed into a target compressed video, which contains multiple compression coding units. These compression coding units can be extracted and processed separately in subsequent processing steps to improve processing efficiency and reduce computational burden.

[0038] Determine one as the first compression coding unit from multiple compression coding units. In the process of video compression coding, there may be virtual frames. These virtual frames are generated to improve compression efficiency or achieve specific coding effects, but are not applicable to the subsequent accurate analysis of the target power transmission channel. Therefore, in this step, only the real scanned images in the first compression coding unit are extracted, that is, those images obtained by directly dynamically scanning the target power transmission channel through the target camera device. The purpose of this is to ensure that the subsequent three-dimensional model mapping is true and effective. If virtual frames are used, the mapping results will be inaccurate, affecting the judgment of the power transmission channel status. By extracting these real first target frame images, a target candidate scan image set can be formed. This image set will serve as an important resource for subsequent processing steps. It contains multiple real scanned images screened from the first compression coding unit, which provides a basis for subsequent image analysis, annotation, and three-dimensional model construction. By selecting only real scanned images, the accuracy of the entire processing process can be improved, ensuring that the analysis and evaluation of the target power transmission channel are more accurate and reliable.

[0039] The target candidate scan image set has been formed. This set contains the first target frame image of the real scan extracted from the first compression coding unit. In order to further carry out subsequent processing, it is necessary to randomly extract a frame of image from this target candidate scan image set. This random extraction process adds a certain degree of uncertainty and diversity. Since the images in the set are all obtained from real scans, each has its own unique perspective and information. Through random selection, the deviation that may be caused by the fixed selection of a specific image can be avoided, making the entire processing process more objective and comprehensive. When a frame of image is randomly extracted, it is recorded as the first two-dimensional scan image. This image will become the starting point for subsequent processing steps, such as image preprocessing, annotation analysis and other operations. It represents the state of the target transmission channel at a specific time and angle, and provides a specific image data basis for the subsequent in-depth analysis and evaluation of the transmission channel.

[0040] In one possible implementation, step 200 further includes:

[0041] Step 240: The image preprocessing plan refers to performing pixel rasterization processing on the first two-dimensional scanned image.

[0042] Specifically, the specific content of the image preprocessing plan is to perform pixel rasterization processing on the first two-dimensional scanned image. After the first two-dimensional scanned image is randomly extracted from the target candidate scanned image set, it needs to be preprocessed to better adapt to the subsequent analysis and processing steps. Pixel rasterization is an important preprocessing method. In this process, the first two-dimensional scanned image is divided into tiny pixel units, each of which has specific color, brightness and other attributes. Through pixel rasterization, the image can be converted into a matrix form composed of many pixels, which is convenient for subsequent digital processing and analysis. This processing method helps to analyze the characteristics of the image more accurately. For example, each pixel unit can be processed and analyzed separately to extract its color characteristics, brightness characteristics and other information. At the same time, pixel rasterization also provides a more fine-grained image data basis for subsequent operations such as annotation analysis, so that the intelligent annotation model can more accurately identify the wire area and hazard source in the image. By performing pixel rasterization processing on the first two-dimensional scanned image, the accuracy and reliability of the single-image ranging method of the entire transmission line can be improved.

[0043] In one possible implementation, step 300 further includes:

[0044] Step 310: The first two-dimensional target image includes a plurality of pixel grids.

[0045] Step 320: extracting a first pixel grid from the plurality of pixel grids, and calling a grid feature extraction plan to perform feature extraction analysis on the first pixel grid to obtain first grid feature information.

[0046] Step 330: In combination with the first grid feature information, a center-surrounding operator is used to perform local contrast analysis on the first two-dimensional target image to obtain a first center-surrounding difference map.

[0047] Step 340: Perform significant pixel grid screening on the first two-dimensional target image according to the first center-surrounding difference map through the screening layer in the intelligent annotation model to obtain a candidate pixel grid set.

[0048] Step 350: labeling the candidate pixel grid set through the labeling layer in the intelligent labeling model to obtain the first labeling information.

[0049] Specifically, the first two-dimensional target image is a preprocessed image, which is composed of a large number of pixel grids. Each pixel grid is actually a small area containing multiple pixels. These pixel grids are like tiny puzzle pieces, which together constitute the complete first two-dimensional target image. The multiple pixels in each pixel grid each have specific color values, brightness values ​​and other attributes. These attributes combined together determine the presentation effect of the pixel grid in the image. By dividing the image into multiple pixel grids, it is more convenient to perform local analysis and processing on the image.

[0050] A first pixel grid is randomly selected from multiple pixel grids contained in the first two-dimensional target image. The first pixel grid represents a local area in the image. The pre-set grid feature extraction plan is called. This plan contains a series of methods and algorithms for extracting pixel grid features. By using these methods, an in-depth feature extraction analysis is performed on the first pixel grid. In the analysis process, multiple features are considered. For example, in terms of color features, the average color value of the first pixel grid, the statistical information of color distribution, etc. are extracted; in terms of texture features, the grayscale change law between pixels, the direction and periodicity of the texture, etc. are analyzed. After these analysis steps, the first grid feature information is finally obtained. This information describes the unique properties and characteristics of the first pixel grid in detail, providing an important basis for subsequent image analysis and processing.

[0051] The first grid feature information has been obtained. This information describes the features of a specific pixel grid and provides a basis for subsequent analysis. The center-surround operator is used to perform local contrast analysis on the first two-dimensional target image. The basic principle of the center-surround operator is to compare the difference between each pixel or pixel grid in the image and the pixels or pixel grids within a certain range around it. Integrating the first grid feature information into this process can make the contrast analysis more targeted. For each pixel grid in the first two-dimensional target image, it is taken as the center, and then the pixel grids within a certain range around it are considered. By calculating the differences in color, texture, etc. between the central pixel grid and the surrounding pixel grids, the local feature changes in the image can be highlighted. After point-by-point analysis of the entire image, the first center-surround difference map is obtained. This difference map can clearly show the local contrast and feature changes of each area in the image. In the difference map, those areas with large differences usually correspond to important features or target areas in the image, such as wire areas and the edges of dangerous sources. The first center-surround difference map provides an important basis for the subsequent screening and annotation of significant pixel grids, which helps to more accurately identify and annotate key areas in the image.

[0052] The screening layer of the intelligent annotation model uses a specific algorithm to screen the significant pixel grids of the first two-dimensional target image. First, a difference threshold is determined. This threshold can be determined by analyzing and experimenting with a large number of sample images, or set according to specific application requirements and experience. For each pixel grid in the first two-dimensional target image, its difference value in the first center-surrounding difference map is calculated. The difference value is a comprehensive measure of multiple features such as color difference and texture difference. Then, the difference value of each pixel grid is compared with the difference threshold. If the difference value is greater than the threshold, the pixel grid is marked as a significant pixel grid. The entire image is traversed, and all pixel grids marked as significant pixel grids are collected to form a candidate pixel grid set. The following specific calculation method is used to determine the difference value. For color features, the sum of the absolute values ​​of the differences between the average color values ​​of the central pixel grid and the surrounding pixel grids is calculated. For texture features, the gray level co-occurrence matrix and other methods are used to calculate the similarity or difference measure of the texture. The difference values ​​of these different features are weighted summed to obtain a comprehensive difference value. Through such an algorithm, the screening layer can effectively screen out significant pixel grids according to the first center-surrounding difference map, and provide an accurate set of candidate pixel grids for subsequent annotation processing, thereby improving the accuracy and efficiency of the intelligent annotation model.

[0053] The annotation layer in the intelligent annotation model begins to annotate the candidate pixel grid set, which is a set of pixel grids that are considered to be related to the wire area or hazard source after being screened by the screening layer. The annotation layer performs annotation based on pre-trained models and rules. First, the annotation layer analyzes the features of each candidate pixel grid, including color, texture, shape, etc. Then, these features are matched with the feature patterns of known wire areas and hazard sources. If the features of a candidate pixel grid match the feature pattern of the wire area, the annotation layer will annotate it as part of the first wire area. Similarly, if its features match the feature pattern of the hazard source, it will be annotated as part of the first hazard source. In the annotation process, convolutional neural networks in deep learning are used to automatically learn the feature patterns in the image, and accurately classify and annotate. After annotating each pixel grid in the candidate pixel grid set, the first annotation information is finally obtained. This annotation information clearly points out which pixel grids in the first two-dimensional target image belong to the first conductor area and which belong to the first hazard source. The first annotation information will provide key data support for subsequent operations such as building a three-dimensional channel model and performing distance measurement, helping to accurately assess the safety status of the transmission channel.

[0054] In one possible implementation, step 320 further includes:

[0055] Step 321: The grid feature extraction plan includes a brightness feature extraction plan and a color feature extraction plan.

[0056] Step 322: Acquire first brightness feature information of the first pixel grid according to the brightness feature extraction plan.

[0057] Step 323: According to the color feature extraction plan, first red channel feature information, first green channel feature information and first blue channel feature information of the first pixel grid are obtained in sequence.

[0058] Step 324: the first brightness characteristic information, the first red channel characteristic information, the first green channel characteristic information and the first blue channel characteristic information constitute the first grid characteristic information.

[0059] Specifically, in this process, the grid feature extraction plan is clearly divided into the brightness feature extraction plan and the color feature extraction plan. The brightness feature extraction plan aims to extract feature information related to brightness from the pixel grid. By analyzing the brightness of the pixel grid, different objects and areas in the image can be better identified. For example, the brighter wire part and the relatively darker background area can be distinguished. The color feature extraction plan focuses on extracting the color feature information of the pixel grid, including the analysis of the three channels of red, green and blue to determine the specific color attributes of the pixel grid. The two plans work together to comprehensively extract the important features of the first pixel grid, provide strong support for subsequent image analysis and processing, and help to accurately identify and mark the wire area and hazard sources in the first two-dimensional target image, as well as conduct more in-depth image understanding and analysis.

[0060] According to the brightness feature extraction plan, the first brightness feature information of the first pixel grid is obtained. The brightness feature is crucial for image analysis, and it reflects the brightness of the pixel grid. By calculating the average brightness value and the variance of the brightness of the pixel grid, the information that can describe the brightness characteristics of the first pixel grid is extracted. These brightness feature information can help distinguish different areas and objects in the image, such as identifying the wire area or the danger source with different brightness from the surrounding environment based on the brightness difference. The first brightness feature information will provide an important basis for the subsequent image processing steps, and help to more accurately perform local contrast analysis, significant pixel grid screening, and annotation processing.

[0061] According to the color feature extraction plan, the first pixel grid is processed to obtain its color channel feature information. Color images are usually composed of three channels: red, green, and blue, and each channel carries specific color information. First, the first red channel feature information of the first pixel grid is obtained, which includes the average color value of the red channel, the range of color distribution, etc. Then, the first green channel feature information and the first blue channel feature information are obtained, and their respective properties are analyzed in the same way. By extracting the features of these three channels separately, the performance of the first pixel grid in different color dimensions can be fully understood.

[0062] The first brightness feature information and the first red channel feature information, the first green channel feature information and the first blue channel feature information are combined to form the first grid feature information. The brightness feature reflects the brightness of the pixel grid, while the feature information of the red, green and blue channels fully displays the color attributes of the pixel grid. By integrating these different features, a more complete and detailed description of the first pixel grid can be obtained. This first grid feature information will provide a rich data foundation for subsequent image processing and analysis. When performing local contrast analysis, significant pixel grid screening and annotation processing, comprehensive consideration of multiple features such as brightness and color can more accurately identify important areas in the image, such as wire areas and dangerous sources, thereby providing a key basis for operations such as building a three-dimensional channel model and performing distance measurement.

[0063] In one possible implementation, step 350 further includes:

[0064] Step 351: Extract the first candidate grid from the candidate pixel grid set.

[0065] Step 352: taking first candidate features of the first candidate grid as input information of the annotation layer to obtain a first output result, wherein the first candidate features include first candidate texture features and first candidate color features.

[0066] Step 353: labeling the first candidate grid according to the first output result, and forming the first labeling information.

[0067] Specifically, a specific grid is selected as the first candidate grid from the candidate pixel grid set, which is a set of pixel grids that are considered to be related to the wire area or hazard source after the previous screening step. By extracting the first candidate grid, it is prepared for a more in-depth feature analysis and annotation processing. This first candidate grid will become the focus of further analysis, and its features will be input into the annotation layer to determine whether it belongs to the first wire area or the first hazard source.

[0068] The first candidate features of the first candidate grid are input into the annotation layer. The first candidate features here include the first candidate texture features and the first candidate color features. The texture features can reflect the pattern regularity, roughness and other characteristics in the first candidate grid, such as whether it has a specific line direction or granularity. The color features reflect the color attributes of the grid, including the main color components and the distribution of colors. The annotation layer will process and analyze these input feature information, and output a result, i.e., the first output result, through the pre-trained model. This result will help determine the specific category of the first candidate grid in the image, such as whether it belongs to the first wire area or the first hazard source.

[0069] The first candidate grid is labeled according to the first output result of the labeling layer. If the first output result shows that the grid matches the characteristics of the first wire area, it will be labeled as part of the first wire area; if the result shows that it matches the characteristics of the first hazard source, the first candidate grid will be labeled as part of the first hazard source. By accurately labeling the first candidate grid in this way, its specific category attributes in the image can be clarified. When each grid in the candidate pixel grid set is processed similarly, the first labeling information can be formed. This labeling information comprehensively points out which pixel grids in the first two-dimensional target image belong to the first wire area and which belong to the first hazard source, which provides a crucial basis for the subsequent construction of a three-dimensional channel model and distance measurement operations, and helps to more accurately assess the safety status of the transmission channel.

[0070] In one possible implementation, step 350 further includes:

[0071] Step 354: The annotation layer is a classification model obtained by performing supervised training on the data in the historical power transmission line database based on the support vector machine principle.

[0072] Specifically, the annotation layer is a classification model obtained by supervised training of data in the historical power transmission line database based on the principle of support vector machine. The historical power transmission line database contains a large amount of annotated power transmission line image data, which covers different scenes, weather conditions and line status. Support vector machine is a powerful machine learning algorithm. Its basic idea is to find a hyperplane in the feature space that can separate different categories of data as much as possible. During the training process, the features of the image data in the historical database, such as texture features, color features, etc., are used as input, and the corresponding annotation information (i.e., whether it belongs to the wire area or the hazard source, etc.) is used as the output target. By continuously adjusting the parameters of the model, the model can predict the correct annotation results as accurately as possible based on the input features. After supervised training, the annotation layer model can learn the relationship pattern between different features and the wire area and the hazard source. When the features of the new first candidate grid are input into the annotation layer, it can classify and judge according to the previously learned pattern, and output the annotation result of which area the grid belongs to. The use of this annotation layer model provides a reliable method for accurately annotating the wire area and the hazard source in the first two-dimensional target image, and lays the foundation for the subsequent construction of the three-dimensional channel model and distance measurement.

[0073] Embodiment 2 is based on the same inventive concept as the single-image distance measurement method for power transmission lines based on three-dimensional point clouds in the above-mentioned embodiment. Figure 2As shown, the present application provides a transmission line single-image distance measurement system based on three-dimensional point cloud, and the system and method embodiments in the present application are based on the same inventive concept. The system includes:

[0074] The target scanning video acquisition module 10 is used to acquire a target scanning video. The target scanning video refers to a video obtained by dynamically scanning a target power transmission channel through a target camera device.

[0075] The first two-dimensional target image acquisition module 20 is used to read the image preprocessing plan, and process the first two-dimensional scanning image in the target scanning video according to the image preprocessing plan to obtain the first two-dimensional target image.

[0076] The first annotation information acquisition module 30 is used to call the intelligent annotation model to perform annotation analysis on the first two-dimensional target image to obtain first annotation information, wherein the first annotation information includes a first wire area and a first hazard source.

[0077] The three-dimensional channel model construction module 40 is used to construct a three-dimensional channel model according to the channel characteristic information of the target power transmission channel collected by the laser radar equipment.

[0078] The transmission channel visible graph acquisition module 50 is used to map the first two-dimensional target image marked with the first wire area and the first hazard source to the three-dimensional channel model to obtain a transmission channel visible graph.

[0079] An arbitrary peak point acquisition module 60 is used to acquire any hazard source in the visible graph of the power transmission channel and acquire any peak point of the arbitrary hazard source.

[0080] The target distance acquisition module 70 is used to obtain the target distance in combination with the arbitrary peak point. The target distance refers to the spatial distance between the conductor in the target power transmission channel and the arbitrary hazard source.

[0081] Furthermore, the first two-dimensional target image acquisition module 20 also includes:

[0082] The target compressed video acquisition unit is used to compress the target scanning video using a dynamic picture expert group to obtain a target compressed video, and the target compressed video includes a plurality of compression encoding units.

[0083] A target candidate scanning image set building unit, wherein the target candidate scanning image set building unit builds a target candidate scanning image set based on the first target frame image extracted from the first compression coding unit, wherein the first compression coding unit is any one of the multiple compression coding units.

[0084] The first two-dimensional scanning image acquisition unit is used to randomly extract any frame of image from the target candidate scanning image set, which is recorded as the first two-dimensional scanning image.

[0085] Furthermore, the first two-dimensional target image acquisition module 20 also includes:

[0086] A pixel rasterization processing unit, wherein the pixel rasterization processing unit is used for the image preprocessing scheme to perform pixel rasterization processing on the first two-dimensional scanned image.

[0087] Furthermore, the first annotation information acquisition module 30 further includes:

[0088] The first two-dimensional object image includes a plurality of pixel grids.

[0089] The first grid feature information acquisition unit is used to extract a first pixel grid from the multiple pixel grids, and call a grid feature extraction plan to perform feature extraction analysis on the first pixel grid to obtain first grid feature information.

[0090] The first center-surrounding difference map acquisition unit is used to combine the first grid feature information and use the center-surrounding operator to perform local contrast analysis on the first two-dimensional target image to obtain a first center-surrounding difference map.

[0091] A candidate pixel grid set acquisition unit is used to perform significant pixel grid screening on the first two-dimensional target image according to the first center-surrounding difference map through a screening layer in the intelligent annotation model to obtain a candidate pixel grid set.

[0092] A pixel grid set annotation processing unit is used to perform annotation processing on the candidate pixel grid set through the annotation layer in the intelligent annotation model to obtain the first annotation information.

[0093] Furthermore, the first grid feature information acquisition unit further includes:

[0094] The grid feature extraction plan includes a brightness feature extraction plan and a color feature extraction plan.

[0095] A first brightness feature information acquisition unit is used to acquire first brightness feature information of the first pixel grid according to the brightness feature extraction plan.

[0096] A feature information acquisition unit is used to sequentially acquire first red channel feature information, first green channel feature information and first blue channel feature information of the first pixel grid according to the color feature extraction plan.

[0097] The first grid feature information composition unit is used for the first brightness feature information, the first red channel feature information, the first green channel feature information and the first blue channel feature information to compose the first grid feature information.

[0098] Furthermore, the pixel grid set annotation processing unit further includes:

[0099] The first candidate grid extraction unit is used to extract the first candidate grid in the candidate pixel grid set.

[0100] The first output result acquisition unit is used to use the first candidate feature of the first candidate grid as input information of the annotation layer to obtain a first output result, wherein the first candidate feature includes a first candidate texture feature and a first candidate color feature.

[0101] The first candidate grid labeling unit is used to label the first candidate grid according to the first output result and form the first labeling information.

[0102] Furthermore, the pixel grid set annotation processing unit further includes:

[0103] The annotation layer is a classification model obtained by performing supervised training on the data in the historical power transmission line database based on the support vector machine principle.

[0104] Furthermore, the target distance acquisition module 70 further includes:

[0105] An arbitrary wire parallel line acquisition unit is used to determine a plan according to the target distance to acquire an arbitrary wire parallel line of the arbitrary peak point, wherein the arbitrary wire parallel line is parallel to the first target wire in the first wire area.

[0106] A target arbitrary point acquisition unit is used to translate the arbitrary conductive line parallel line to obtain the target arbitrary point of the arbitrary hazard source.

[0107] A target distance determination unit is used to record the distance from any point of the target to the first target wire as the target distance.

[0108] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0109] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

[0110] This specification and the drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A single-image distance measurement method for power transmission lines based on three-dimensional point cloud, characterized in that: include: Obtaining a target scanning video, wherein the target scanning video refers to a video obtained by dynamically scanning a target power transmission channel through a target camera device; Reading an image preprocessing plan, and processing the first two-dimensional scanned image in the target scanning video according to the image preprocessing plan to obtain a first two-dimensional target image; Retrieving an intelligent annotation model to perform annotation analysis on the first two-dimensional target image to obtain first annotation information, where the first annotation information includes a first wire area and a first hazard source; Constructing a three-dimensional channel model according to channel characteristic information of the target power transmission channel collected by the laser radar device; Mapping the first two-dimensional target image marked with the first wire area and the first hazard source to the three-dimensional channel model to obtain a visible diagram of the power transmission channel; Obtain any hazard source in the visible graph of the power transmission channel, and obtain any peak point of the hazard source; The target distance is obtained in combination with the arbitrary peak point, and the target distance refers to the spatial distance between the conductor in the target power transmission channel and the arbitrary hazard source.

2. The transmission line single image distance measurement method based on three-dimensional point cloud according to claim 1 is characterized in that: include: Using a dynamic picture expert group to compress the target scanned video to obtain a target compressed video, wherein the target compressed video includes a plurality of compression encoding units; Building a target candidate scanning image set based on the first target frame image extracted from the first compression coding unit, wherein the first compression coding unit is any one of the multiple compression coding units; An arbitrary frame of image in the target candidate scan image set is randomly extracted and recorded as the first two-dimensional scan image.

3. The transmission line single image distance measurement method based on three-dimensional point cloud according to claim 1 is characterized in that: The image preprocessing plan refers to performing pixel rasterization processing on the first two-dimensional scanned image.

4. The transmission line single image distance measurement method based on three-dimensional point cloud according to claim 1 is characterized in that: include: The first two-dimensional target image includes a plurality of pixel grids; Extracting a first pixel grid from the plurality of pixel grids, and calling a grid feature extraction plan to perform feature extraction analysis on the first pixel grid to obtain first grid feature information; In combination with the first grid feature information, a center-surrounding operator is used to perform local contrast analysis on the first two-dimensional target image to obtain a first center-surrounding difference map; Performing significant pixel grid screening on the first two-dimensional target image according to the first center-surrounding difference map through the screening layer in the intelligent annotation model to obtain a candidate pixel grid set; The candidate pixel grid set is labeled by using the labeling layer in the intelligent labeling model to obtain the first labeling information.

5. The transmission line single image distance measurement method based on three-dimensional point cloud according to claim 4 is characterized in that: include: The grid feature extraction plan includes a brightness feature extraction plan and a color feature extraction plan; Acquire first brightness feature information of the first pixel grid according to the brightness feature extraction plan; According to the color feature extraction plan, first red channel feature information, first green channel feature information and first blue channel feature information of the first pixel grid are sequentially acquired; The first brightness characteristic information, the first red channel characteristic information, the first green channel characteristic information and the first blue channel characteristic information constitute the first grid characteristic information.

6. The transmission line single image distance measurement method based on three-dimensional point cloud according to claim 4 is characterized in that: include: Extracting a first candidate grid from the candidate pixel grid set; Using first candidate features of the first candidate grid as input information of the annotation layer to obtain a first output result, wherein the first candidate features include first candidate texture features and first candidate color features; The first candidate grid is labeled according to the first output result, and the first labeling information is formed.

7. The transmission line single-image distance measurement method based on three-dimensional point cloud according to claim 6 is characterized in that: The annotation layer is a classification model obtained by performing supervised training on the data in the historical power transmission line database based on the support vector machine principle.

8. A single-image distance measurement system for power transmission lines based on three-dimensional point cloud, characterized in that: The system comprises: A target scanning video acquisition module, wherein the target scanning video acquisition module is used to acquire a target scanning video, wherein the target scanning video refers to a video obtained by dynamically scanning a target power transmission channel through a target camera device; A first two-dimensional target image acquisition module, wherein the first two-dimensional target image acquisition module is used to read an image preprocessing plan and process the first two-dimensional scanned image in the target scanning video according to the image preprocessing plan to obtain a first two-dimensional target image; A first annotation information acquisition module, the first annotation information acquisition module is used to call an intelligent annotation model to perform annotation analysis on the first two-dimensional target image to obtain first annotation information, wherein the first annotation information includes a first wire area and a first hazard source; A three-dimensional channel model building module, wherein the three-dimensional channel model building module is used to build a three-dimensional channel model according to channel feature information of the target power transmission channel collected by a laser radar device; A transmission channel visible graph acquisition module, the transmission channel visible graph acquisition module is used to map the first two-dimensional target image marked with the first wire area and the first hazard source to the three-dimensional channel model to obtain a transmission channel visible graph; An arbitrary peak point acquisition module, the arbitrary peak point acquisition module is used to acquire any hazard source in the visible graph of the power transmission channel, and acquire any peak point of the arbitrary hazard source; A target distance acquisition module is used to obtain a target distance in combination with the arbitrary peak point, and the target distance refers to the spatial distance between the conductor in the target power transmission channel and the arbitrary hazard source.

Images