A machine learning-based transmission line image inspection system

Abstract

The present invention relates to the technical field of transmission line monitoring, and in particular to a transmission line image inspection system based on machine learning, including an acquisition layer, a transmission layer, a receiving layer, a scheduling layer, an inspection layer, and an analysis layer. The inspection accuracy levels of each expected site are divided based on the inspection results of the inspection device. When the division of the inspection accuracy levels of the expected sites in each area to be inspected is completed, it is determined whether the operating parameters of the acquisition layer are qualified based on the statistical results of the number of each inspection accuracy level. When it is determined that the operating parameters of the acquisition layer are abnormal, the operating parameters of the acquisition layer are automatically adjusted, improving the inspection efficiency of the transmission line.

Description

Technical Field

[0001] The present invention relates to the technical field of transmission line monitoring, and in particular, to a transmission line image inspection system based on machine learning. Background Art

[0002] With the development of technology, electricity has become an indispensable energy source for people's life and work. The construction of the power system has also become an important part of urban planning. Transmission lines are an important part of the power system. Whether there are defects in the transmission lines greatly affects the normal operation of the power system. Therefore, the requirements for defect analysis of transmission lines are getting higher and higher.

[0003] Poles and towers are important components in transmission lines. Their function is to support overhead line conductors and overhead ground wires. Defect analysis of each part of the poles and towers has also become an important task in the defect analysis of transmission lines. With the development of UAV technology, the power industry has also introduced UAV technology for daily inspection of transmission lines, and the method of inspecting poles and towers by UAVs is becoming more and more widespread.

[0004] The Chinese patent application with the publication number CN108365557A discloses a method and system for fine inspection of transmission lines by UAVs, including the following steps: obtaining the type of the pole and tower to be inspected; obtaining the information of each part to be inspected according to the type of the pole and tower to be inspected; obtaining the image data of each part according to the information of each part to be inspected, and associatively storing the information of each part with the corresponding obtained image data; performing defect analysis on each corresponding part according to the image data of each part, and obtaining a defect analysis report. It can be seen that the above technical solution has the following problems: it does not consider automatically adjusting the operating parameters of the acquisition layer according to the abnormal conditions of the actually detected transmission lines, which affects the inspection efficiency of the transmission lines. Summary of the Invention

[0005] Therefore, the present invention provides a transmission line image inspection system based on machine learning to overcome the problem in the prior art that the inspection efficiency of the transmission line is affected because the operating parameters of the acquisition layer are not automatically adjusted according to the abnormal conditions of the actually detected transmission lines.

[0006] To achieve the above object, the present invention provides a transmission line image inspection system based on machine learning, including:

[0007] An acquisition layer, including a number of UAVs for periodically acquiring detection image information of each area to be inspected of the transmission line;

[0008] A transmission layer, connected to the acquisition layer, for transmitting the detection image information acquired by each UAV;

[0009] A receiving layer, which is connected to the transmission layer, is used to receive detection image information, determine whether there is an abnormality in the corresponding area to be inspected based on the detection image information, and determine the expected site in the case of determining that the detection image information is abnormal;

[0010] A scheduling layer, which is connected to the receiving layer, is used to select a corresponding inspection device for inspection based on the determined expected site;

[0011] An inspection layer, which is connected to the scheduling layer, includes a number of inspection devices. Each inspection device moves along the transmission line. When the inspection device reaches the expected site, the inspection device obtains the abnormal situation of the expected site to identify appearance defects;

[0012] An analysis layer, which is respectively connected to the acquisition layer, the transmission layer and the inspection layer, is used to periodically determine the deviation distance based on the longitude and latitude of the site of the appearance defect and the expected site, divide the inspection accuracy level of the expected site based on the deviation distance, where the inspection accuracy level includes a first-level site, a second-level site and a third-level site. When the analysis layer completes the division of the inspection accuracy level of the expected site of each area to be inspected, it determines whether the operating parameters of the acquisition layer are qualified based on the statistical results of the number of each inspection accuracy level, including:

[0013] Determine that the operating parameters of the acquisition layer are qualified, and control the acquisition layer to continuously use the current operating parameters to detect the transmission line; or, determine that the operating parameters of the acquisition layer are abnormal, and adjust the preset similarity to the corresponding value based on the proportion of the number of third-level sites.

[0014] Further, the receiving layer is used to determine whether there is an abnormality in the corresponding area to be inspected based on the detection image information, including:

[0015] The receiving layer is used to obtain the feature contour in the detection image information, compare the feature contour with each preset contour pre-stored in the database. In the case that the similarity between the feature contour in the detection image information and a single preset contour pre-stored in the database is higher than the preset similarity, determine that there is an abnormality in a single area to be inspected, and frame the corresponding area with a similarity higher than the preset similarity in the feature contour as an abnormal area;

[0016] The receiving layer determines the longitude and latitude of the abnormal area based on the coordinates of the abnormal area in the detection image information and the geographical longitude and latitude of the corresponding area to be inspected, and determines the corresponding longitude and latitude as the expected site;

[0017] The scheduling layer sends an instruction to make the inspection device move along the transmission line to the corresponding expected site for inspection;

[0018] The similarity is to compare the preset contour with the feature contour for a single preset contour, and record the ratio of the area of the overlapping region obtained to the total area of the preset contour as the similarity.

[0019] Further, the way for the inspection device to obtain the abnormal conditions of the expected sites includes:

[0020] The inspection device is used to obtain the detailed image information of the expected sites through an image detector, and analyze the detailed image information through an image processing algorithm to detect the appearance defects of the transmission line. The appearance defects include insulator breakage, conductor strand breakage, and fitting corrosion.

[0021] When the inspection device cannot detect the appearance defects based on the detailed image information, taking the expected site as the center and the transmission line as the extension direction, obtain several pieces of detailed image information on both sides until the appearance defects are recognized, or the moving distance exceeds the corresponding area to be inspected.

[0022] The analysis layer records the distance between the site where the appearance defect is recognized and the expected site as the deviation distance.

[0023] The analysis layer is used to divide the inspection accuracy level of the expected sites based on the deviation distance, including:

[0024] If the deviation distance is less than or equal to the first preset deviation distance, the inspection accuracy level of the expected site is divided into a first-level site.

[0025] If the deviation distance is less than or equal to the second preset deviation distance and greater than the first preset deviation distance, the inspection accuracy level of the expected site is divided into a second-level site.

[0026] If the deviation distance is greater than the second preset deviation distance, the inspection accuracy level of the expected site is divided into a third-level site.

[0027] Further, under the condition that the analysis layer has completed the division of the inspection accuracy levels of the expected sites in each area to be inspected, determine whether the operating parameters of the acquisition layer are qualified based on the statistical results of the number of each inspection accuracy level, including:

[0028] If the number of the first-level sites divided among all the expected sites is the largest, it is determined that the operating parameters of the acquisition layer are qualified, and the acquisition layer is controlled to continuously use the current operating parameters to detect the transmission line.

[0029] If the number of the second-level sites divided among all the expected sites is the largest, determine whether the operating parameters of the acquisition layer are qualified based on the time distribution parameters of each second-level site.

[0030] If the number of the third-level sites divided among all the expected sites is the largest, it is determined that the operating parameters of the acquisition layer are abnormal, and the preset similarity is adjusted to the corresponding value based on the proportion of the number of the third-level sites.

[0031] Further, the analysis layer is used to determine whether the operating parameters of the acquisition layer are qualified based on the time distribution parameters of each secondary site, including:

[0032] It is used to determine the time distribution parameters of each secondary site, determine the time nodes when the acquisition layer obtains the detection image information corresponding to each secondary site, calculate the interval duration between each time node, and solve the average value of each interval duration to obtain the time distribution parameters;

[0033] If the time distribution parameter is less than or equal to the preset time distribution parameter, it is determined that the operating parameters of the acquisition layer are qualified, and the transmission bandwidth of the transmission layer is adjusted to the corresponding value based on the total number of secondary sites;

[0034] If the time distribution parameter is greater than the preset time distribution parameter, the area of the area to be inspected is adjusted to the corresponding value based on the time distribution parameter.

[0035] Further, the analysis layer is used to adjust the area of the area to be inspected to the corresponding value based on the time distribution parameter, where:

[0036] The reduction amplitude of the area of the area to be inspected determined based on the time distribution parameter is inversely proportional to the time distribution parameter.

[0037] Further, the analysis layer is used to adjust the transmission bandwidth of the transmission layer to the corresponding value based on the total number of secondary sites, where:

[0038] The increase amplitude of the transmission bandwidth determined based on the total number of secondary sites is proportional to the total number of secondary sites.

[0039] Further, the analysis layer is used to adjust the preset similarity to the corresponding value based on the proportion of the number of tertiary sites, where:

[0040] The analysis layer is used to record the ratio of the number of tertiary sites to the total number of each expected site as the proportion of the number of tertiary sites;

[0041] The increase amplitude of the preset similarity determined based on the proportion is proportional to the proportion.

[0042] Further, under the condition that the analysis layer completes the adjustment of the preset similarity, it re-divides the inspection accuracy levels of each expected site according to the obtained detection image information; if it is still determined that the number of tertiary sites divided in each expected site is the largest, the operating height of each drone is adjusted to the corresponding value based on the quantity ratio of the proportion obtained again and the proportion obtained in the previous cycle;

[0043] The reduction amplitude of the operating height determined based on the quantity ratio is proportional to the quantity ratio.

[0044] Compared with the prior art, the beneficial effects of the present invention are as follows: the inspection accuracy levels of each expected site are divided based on the inspection results of the inspection device. When the inspection accuracy levels of the expected sites in each area to be inspected are completed, it is determined whether the operating parameters of the acquisition layer are qualified based on the statistical results of the number of each inspection accuracy level. When it is determined that the operating parameters of the acquisition layer are abnormal, the operating parameters of the acquisition layer are automatically adjusted, improving the inspection efficiency of the transmission line.

[0045] Further, during the flight, the unmanned aerial vehicle collects images of the area to be inspected in the transmission line, and compares the feature contour of the obtained detected image information with the pre-stored preset contour. When the similarity between the feature contour in the detected image information and a single preset contour pre-stored in the database is higher than the preset similarity, it is determined that there is an expected abnormal situation in this area. Based on the determination result, the abnormal site is determined, and the abnormal site is determined as the expected site. An instruction is sent to make the inspection device move along the line to the corresponding site for inspection; the inspection device obtains the abnormal situation of the expected site. When the deviation distance is less than or equal to the first preset deviation distance, there is an expected abnormality and the distance between the expected site and the inspection site is small, and the inspection site is the site where the appearance defect is recognized. The image information at the site where the appearance defect is recognized is collected, and local collection is performed, and corresponding emergency repair notifications or maintenance notifications are sent according to the features obtained from the image information. At this time, the expected site in this case is recorded as a first-level site; when the deviation distance is less than or equal to the second preset deviation distance and greater than the first preset deviation distance, there is an expected abnormality at this time and the distance between the expected site and the inspection site is large. In this case, the inspection device collects the detailed image information of the transmission line with the appearance defect, and the analysis layer records the corresponding expected site as a second-level site; when the deviation distance is greater than the second preset deviation distance, there is a situation where no expected abnormality can be recognized in the area to be inspected. The expected site in this case is recorded as a third-level site. The inspection accuracy levels of each determined expected point are divided to determine the accuracy of the determined expected point, further providing a basis for the adjustment of the operating parameters of the acquisition layer, effectively improving the processing efficiency of the collected data while further improving the inspection efficiency of the transmission line.

[0046] Further, after completing the inspection of a single cycle, the number of inspection accuracy levels at each level within the cycle is counted. When there are many first-level sites, it is determined that the acquisition of the acquisition layer is qualified, and continuous monitoring is carried out; when there are many second-level sites, the acquisition layer is further judged based on the time distribution parameter, and the time distribution parameter characterizes the distribution of each second-level site in the time dimension. When the time distribution parameter is less than or equal to the preset time distribution parameter, the time span of the appearance of the second-level site is large and the distribution of each time interval is concentrated within a small range. At this time, the situation of abnormal transmission appears randomly, and each occurrence causes abnormal transmission of multiple data. This is caused by network fluctuations during the data transmission process. At this time, the transmission bandwidth of the transmission layer is adjusted to ensure the accurate transmission of data. When the time distribution parameter is greater than the preset time distribution parameter, the distribution of the second-level sites is scattered, and the acquisition sites of the drone are unqualified. The acquisition parameters for the drone to acquire detection image information are adjusted, and the area of each area to be inspected is corrected based on the time distribution parameter to ensure the accuracy of image acquisition; when the number of third-level sites divided among all expected sites is the largest, it is determined that the operating parameters of the acquisition layer are abnormal, and the preset similarity is adjusted to the corresponding value to improve the accuracy of abnormal recognition, further effectively improving the inspection efficiency of the transmission line. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 FIG. is a block diagram of a transmission line image inspection system based on machine learning according to an embodiment of the present invention;

[0048] Figure 2 FIG. is a logical decision diagram for dividing the inspection accuracy level of expected sites based on the deviation distance in the analysis layer according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0049] In order to make the objectives and advantages of the present invention clearer, the present invention will be further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0050] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present invention and do not limit the protection scope of the present invention.

[0051] It should be noted that in the description of the present invention, the terms indicating the direction or positional relationship such as "upper", "lower", "left", "right", "inner", "outer", etc. are based on the direction or positional relationship shown in the drawings. This is only for convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present invention.

[0052] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and defined, the terms "installation", "connection", and "coupling" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, and it can be the communication inside two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0053] Please refer to Figure 1 and Figure 2 as shown, which are respectively the module block diagram of the transmission line image inspection system based on machine learning in the embodiment of the present invention, and the logical decision diagram for dividing the inspection accuracy level of the expected sites based on the deviation distance in the analysis layer; An transmission line image inspection system based on machine learning in an embodiment of the present invention includes:

[0054] An acquisition layer, including a plurality of unmanned aerial vehicles (UAVs) for periodically acquiring the detection image information of each area to be inspected on the transmission line;

[0055] A transmission layer, which is connected to the acquisition layer and is used for transmitting the detection image information acquired by each UAV;

[0056] A receiving layer, which is connected to the transmission layer, is used for receiving the detection image information, and for determining whether there is an abnormality in the corresponding area to be inspected based on the detection image information, and determining the expected site in the case of determining that the detection image information is abnormal;

[0057] A scheduling layer, which is connected to the receiving layer and is used for selecting the corresponding inspection device for inspection based on the determined expected site;

[0058] An inspection layer, which is connected to the scheduling layer and includes a plurality of inspection devices. Each inspection device moves along the transmission line. When the inspection device reaches the expected site, the inspection device acquires the abnormal situation of the expected site to identify the appearance defect;

[0059] An analysis layer, which is respectively connected to the acquisition layer, the transmission layer and the inspection layer, is used for periodically determining the deviation distance based on the longitude and latitude of the site of the appearance defect and the expected site, dividing the inspection accuracy level of the expected site based on the deviation distance, where the inspection accuracy level includes a first-level site, a second-level site and a third-level site. Under the condition that the analysis layer completes the division of the inspection accuracy level of the expected sites in each area to be inspected, it determines whether the operating parameters of the acquisition layer are qualified based on the statistical results of the number of each inspection accuracy level, including:

[0060] Determine that the operating parameters of the acquisition layer are qualified, and control the acquisition layer to continuously use the current operating parameters to detect the transmission line; or, determine that the operating parameters of the acquisition layer are abnormal, and adjust the preset similarity to the corresponding value based on the proportion of the number of three-level sites.

[0061] Specifically, divide the inspection accuracy levels of each expected site based on the inspection results of the inspection device. When the division of the inspection accuracy levels of the expected sites in each area to be inspected is completed, determine whether the operating parameters of the acquisition layer are qualified based on the statistical results of the number of each inspection accuracy level, and when it is determined that the operating parameters of the acquisition layer are abnormal, automatically adjust the operating parameters of the acquisition layer, improving the inspection efficiency of the transmission line.

[0062] Specifically, the receiving layer is used to determine whether there is an abnormality in the corresponding area to be inspected based on the detection image information, including:

[0063] The receiving layer is used to obtain the feature contour in the detection image information, and compare the feature contour with each preset contour pre-stored in the database. When the similarity between the feature contour in the detection image information and a single preset contour pre-stored in the database is higher than the preset similarity, determine that there is an abnormality in a single area to be inspected, and frame the corresponding area in the feature contour with a similarity higher than the preset similarity as the abnormal area;

[0064] The receiving layer determines the longitude and latitude of the abnormal area based on the coordinates of the abnormal area in the detection image information and the geographical longitude and latitude of the corresponding area to be inspected, and determines the corresponding longitude and latitude as the expected site;

[0065] The scheduling layer sends an instruction to make the inspection device move along the transmission line to the corresponding expected site for inspection;

[0066] The similarity is for a single preset contour. Compare the preset contour with the feature contour, and record the ratio of the area of the overlapping area obtained to the total area of the preset contour as the similarity.

[0067] Specifically, the specific method for obtaining the feature contour in the detection image information is not limited. It can be to obtain the edge information of the transmission line in the detection image information through edge detection algorithms such as Canny edge detection, Sobel operator, Roberts operator, etc. This is the prior art and will not be elaborated.

[0068] Specifically, the specific method for comparing the feature contour with each preset contour pre-stored in the database is not limited. The corresponding preset contour and the feature contour can be translated and compared one by one, or the feature contour can be input into the neural network training model according to the neural network training model, and the suspicious area after calibration is output, and the suspicious area is compared with each preset contour in turn to determine the abnormal area by calculating the similarity. This will not be elaborated.

[0069] Specifically, the ways for the inspection device to obtain the abnormal conditions of the expected sites include:

[0070] The inspection device is used to obtain the detailed image information of the expected sites through an image detector, and analyze the detailed image information through an image processing algorithm to detect the appearance defects of the transmission line. The appearance defects include insulator breakage, conductor strand breakage, and fitting corrosion;

[0071] When the inspection device cannot detect appearance defects based on the detailed image information, taking the expected site as the center and the transmission line as the extension direction, a number of detailed image information are obtained on both sides until appearance defects are recognized, or the moving distance exceeds the corresponding area to be inspected;

[0072] The analysis layer records the distance between the site where appearance defects are recognized and the expected site as the deviation distance;

[0073] The analysis layer is used to divide the inspection accuracy level of the expected site based on the deviation distance, including:

[0074] If the deviation distance is less than or equal to the first preset deviation distance, the inspection accuracy level of the expected site is divided into a first-level site;

[0075] If the deviation distance is less than or equal to the second preset deviation distance and greater than the first preset deviation distance, the inspection accuracy level of the expected site is divided into a second-level site;

[0076] If the deviation distance is greater than the second preset deviation distance, the inspection accuracy level of the expected site is divided into a third-level site.

[0077] Specifically, the first preset deviation distance is selected within the interval [0.0277L0, 0.0312L0], and the second preset deviation distance is selected within the interval [0.05L0, 0.0833L0], where L0 is the perimeter of the area to be inspected.

[0078] Specifically, during the flight of the drone, image acquisition is carried out on the area to be inspected in the transmission line. The characteristic contour of the acquired detection image information is compared with the pre-stored preset contour. When the similarity between the characteristic contour in the detection image information and a single preset contour pre-stored in the database is higher than the preset similarity, it is determined that there is an expected abnormal situation in this area. Based on the determination result, the abnormal site is determined, and the abnormal site is determined as the expected site. An instruction is sent to make the inspection device move along the line to the corresponding site for inspection; the inspection device obtains the abnormal situation of the expected site. When the deviation distance is less than or equal to the first preset deviation distance, there is an expected abnormality and the distance between the expected site and the inspection site is small. The inspection site is the site where the appearance defect is recognized. The image information at the site where the appearance defect is recognized is collected, and local collection is carried out, and corresponding emergency repair notices or maintenance notices are sent according to the characteristics obtained from the image information. At this time, the expected site in this case is recorded as the first-level site; when the deviation distance is less than or equal to the second preset deviation distance and greater than the first preset deviation distance, there is an expected abnormality at this time and the distance between the expected site and the inspection site is large. In this case, the inspection device collects the detailed image information of the transmission line with the appearance defect, and the analysis layer records the corresponding expected site as the second-level site; when the deviation distance is greater than the second preset deviation distance, there is a situation where no expected abnormality can be recognized in the area to be inspected. The expected site in this case is recorded as the third-level site. The inspection accuracy levels of each determined expected point are divided to determine the accuracy of the determined expected points, further providing a basis for adjusting the operation parameters of the acquisition layer, effectively improving the processing efficiency of the collected data, and further improving the inspection efficiency of the transmission line.

[0079] Specifically, under the condition that the analysis layer has completed the division of the inspection accuracy levels of the expected sites in each area to be inspected, it is determined whether the operation parameters of the acquisition layer are qualified based on the statistical results of the quantities of each inspection accuracy level, including:

[0080] If the quantity of the first-level sites divided among all the expected sites is the largest, it is determined that the operation parameters of the acquisition layer are qualified, and the acquisition layer is controlled to continuously use the current operation parameters to detect the transmission line;

[0081] If the quantity of the second-level sites divided among all the expected sites is the largest, it is determined whether the operation parameters of the acquisition layer are qualified based on the time distribution parameters of each second-level site;

[0082] If the quantity of the third-level sites divided among all the expected sites is the largest, it is determined that the operation parameters of the acquisition layer are abnormal, and the preset similarity is adjusted to the corresponding value based on the proportion of the quantity of the third-level sites.

[0083] Specifically, the analysis layer is used to determine whether the operation parameters of the acquisition layer are qualified based on the time distribution parameters of each second-level site, including:

[0084] To determine the time distribution parameters of each secondary site, determine the time nodes for the acquisition layer to obtain the detection image information corresponding to each secondary site, calculate the interval duration between each time node, solve the average value of each interval duration, and obtain the time distribution parameters;

[0085] If the time distribution parameter is less than or equal to the preset time distribution parameter, it is determined that the operating parameters of the acquisition layer are qualified, and the transmission bandwidth of the transmission layer is adjusted to the corresponding value based on the total number of secondary sites;

[0086] If the time distribution parameter is greater than the preset time distribution parameter, the area of the area to be inspected is adjusted to the corresponding value based on the time distribution parameter.

[0087] Specifically, the preset time distribution parameter S0 is taken as 0.33T0, and T0 is the maximum value of each time interval.

[0088] Specifically, the analysis layer is used to adjust the area of the area to be inspected to the corresponding value based on the time distribution parameter, where:

[0089] The reduction amplitude of the area of the area to be inspected determined based on the time distribution parameter is inversely proportional to the time distribution parameter.

[0090] In this embodiment, optionally,

[0091] Compare the time distribution parameter with the first preset time comparison threshold and the second preset time comparison threshold;

[0092] If the time distribution parameter is less than or equal to the first preset time comparison threshold, the area of the area to be inspected is adjusted to 0.81 times the initial area;

[0093] If the time distribution parameter is less than or equal to the second preset time comparison threshold and greater than the first preset time comparison threshold, the area of the area to be inspected is adjusted to 0.87 times the initial area;

[0094] If the time distribution parameter is greater than the second preset time comparison threshold, the area of the area to be inspected is adjusted to 0.93 times the initial area;

[0095] The first preset time comparison threshold is taken as 0.7S0, and the second preset time comparison threshold is taken as 0.63S0.

[0096] Specifically, under the condition that the area of the area to be inspected is adjusted, the adjusted area of the area to be inspected is compared with the preset critical area. If the adjusted area of the area to be inspected is greater than the preset critical area, the adjusted area of the area to be inspected is used as the operating parameter of the acquisition layer; if the adjusted area of the area to be inspected is less than or equal to the preset critical area, the preset critical area is used as the operating parameter of the acquisition layer, and the operating height of each drone is adjusted to 0.91 times the initial operating height.

[0097] Specifically, after completing the inspection of a single cycle, the number of inspection accuracy levels at each level within the cycle is counted. When there are many first-level sites, it is determined that the acquisition of the acquisition layer is qualified and continuous monitoring is carried out; when there are many second-level sites, the acquisition layer is further judged based on the time distribution parameter, and the time distribution parameter characterizes the distribution of each second-level site in the time dimension. When the time distribution parameter is less than or equal to the preset time distribution parameter, the time span of the appearance of the second-level sites is large and the distribution of each time interval is concentrated within a small range. At this time, the situation of abnormal transmission occurs randomly, and each occurrence causes abnormal transmission of multiple data. This is caused by network fluctuations during the data transmission process. At this time, the transmission bandwidth of the transmission layer is adjusted to ensure the accurate transmission of data. When the time distribution parameter is greater than the preset time distribution parameter, the distribution of the second-level sites is scattered, and the acquisition sites of the drones are unqualified. The acquisition parameters when the drones acquire and detect image information are adjusted, and the areas of each area to be inspected are corrected based on the time distribution parameter to ensure the accuracy of image acquisition; when the number of third-level sites divided among each expected site is the largest, it is determined that the operating parameters of the acquisition layer are abnormal, and the preset similarity is adjusted to the corresponding value to improve the accuracy of abnormal recognition, further effectively improving the inspection efficiency of the transmission line.

[0098] Specifically, the analysis layer is used to adjust the transmission bandwidth of the transmission layer to the corresponding value based on the total number of second-level sites, where:

[0099] The increase amplitude of the transmission bandwidth determined based on the total number of second-level sites is proportional to the total number of second-level sites.

[0100] In this embodiment, optionally,

[0101] The total number of second-level sites in each area to be inspected obtained by the analysis layer during the current detection cycle is compared with the first preset number and the second preset number;

[0102] If the total number of second-level sites is less than or equal to the first preset number, the transmission bandwidth of the transmission layer is adjusted to 1.12 times the initial bandwidth;

[0103] If the total number of each secondary site is less than or equal to the second preset number and greater than the first preset number, the transmission bandwidth of the transport layer is adjusted to 1.22 times the initial bandwidth;

[0104] If the total number of each secondary site is greater than the second preset number, the transmission bandwidth of the transport layer is adjusted to 1.28 times the initial bandwidth;

[0105] The first preset number is taken as 0.77N0, the second preset number is taken as 0.92N0, and N0 is the total number of each expected site.

[0106] Specifically, the analysis layer is used to adjust the preset similarity to the corresponding value based on the proportion of the number of tertiary sites, where:

[0107] The analysis layer is used to record the proportion of the number of tertiary sites to the total number of each expected site as the proportion of the number of tertiary sites;

[0108] The increase amplitude of the preset similarity determined based on the proportion is proportional to the proportion.

[0109] In this embodiment, optionally,

[0110] Compare the proportion with the first preset proportion and the second preset proportion;

[0111] If the proportion is less than or equal to the first preset proportion, the preset similarity is adjusted to 1.11 times the initial preset similarity;

[0112] If the proportion is less than or equal to the second preset proportion and greater than the first preset proportion, the preset similarity is adjusted to 1.21 times the initial preset similarity;

[0113] If the proportion is greater than the second preset proportion, the preset similarity is adjusted to 1.29 times the initial preset similarity;

[0114] The first preset proportion is taken as 0.6, and the second preset proportion is taken as 0.8.

[0115] Specifically, under the condition that the analysis layer completes the adjustment of the preset similarity, the inspection accuracy level of each expected site is re-divided according to the obtained information of each detection image; if it is still determined that the number of tertiary sites divided in each expected site is the largest, the operating height of each drone is adjusted to the corresponding value based on the quantity ratio of the proportion obtained again and the proportion obtained in the previous cycle;

[0116] The reduction amplitude of the operating height determined based on the quantity ratio is proportional to the quantity ratio.

[0117] In this embodiment, optionally,

[0118] Compare the quantity ratio with a first preset quantity ratio and a second preset quantity ratio;

[0119] If the quantity ratio is less than or equal to the first preset quantity ratio, adjust the running height to 0.91 times the initial running height;

[0120] If the quantity ratio is less than or equal to the second preset quantity ratio and greater than the first preset quantity ratio, adjust the running height to 0.81 times the initial running height;

[0121] If the quantity ratio is greater than the second preset quantity ratio, adjust the running height to 0.72 times the initial running height;

[0122] The first preset quantity ratio is 0.62, and the second preset quantity ratio is 0.8.

[0123] So far, the technical solution of the present invention has been described in conjunction with the preferred embodiments shown in the drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0124] The above are only the preferred embodiments of the present invention and are not used to limit the present invention; for those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A power transmission line image inspection system based on machine learning, characterized in that: include: The collection layer includes a number of drones for periodically collecting detection image information of each area to be inspected on the transmission line; A transmission layer, which is connected to the acquisition layer and is used to transmit the detection image information collected by each drone; A receiving layer connected to the transmission layer, for receiving the detection image information, and determining whether there is an abnormality in the corresponding area to be inspected based on the detection image information, and determining the expected location when it is determined that there is an abnormality in the detection image information; A scheduling layer, which is connected to the receiving layer and is used to select a corresponding inspection device for inspection based on the determined expected location; An inspection layer, which is connected to the dispatching layer, includes a plurality of inspection devices, each of which moves along the transmission line. When the inspection device reaches the expected location, the inspection device obtains abnormal conditions at the expected location to identify appearance defects; The analysis layer is connected to the acquisition layer, the transmission layer and the inspection layer respectively, and is used to periodically determine the deviation distance based on the longitude and latitude of the location of the appearance defect and the expected location, and divide the inspection accuracy level of the expected location based on the deviation distance, wherein the inspection accuracy level includes the primary location, the secondary location and the tertiary location. Under the condition that the analysis layer is used to complete the division of the inspection accuracy level of the expected location of each area to be inspected, the operation parameters of the acquisition layer are determined based on the quantitative statistical results of each inspection accuracy level, including: Determine that the operating parameters of the collection layer are qualified, and control the collection layer to continue to use the current operating parameters to detect the transmission line; or, determine that the operating parameters of the collection layer are abnormal, and adjust the preset similarity to a corresponding value based on the proportion of the number of third-level sites; The analysis layer is used to classify the inspection accuracy level of the expected location based on the deviation distance, including: If the deviation distance is less than or equal to the first preset deviation distance, the inspection accuracy level of the expected location is classified as a first-level location; If the deviation distance is less than or equal to the second preset deviation distance and greater than the first preset deviation distance, the inspection accuracy level of the expected location is classified as a secondary location; If the deviation distance is greater than the second preset deviation distance, the inspection accuracy level of the expected location is divided into three levels of locations.

2. The power transmission line image inspection system based on machine learning according to claim 1 is characterized in that: The receiving layer is used to determine whether there is an abnormality in the corresponding area to be inspected based on the detected image information, including: The receiving layer is used to obtain the characteristic contour in the detection image information, and compare the characteristic contour with each preset contour pre-stored in the database. When the similarity between the characteristic contour in the detection image information and a single preset contour pre-stored in the database is higher than the preset similarity, it is determined that there is an abnormality in the single area to be inspected, and the corresponding area in the characteristic contour with a similarity higher than the preset similarity is selected as the abnormal area; The receiving layer determines the longitude and latitude of the abnormal area based on the coordinates of the abnormal area in the detection image information and the geographical longitude and latitude of the corresponding area to be inspected, and determines the corresponding longitude and latitude as the expected location; The dispatching layer sends instructions to move the inspection device along the transmission line to the corresponding expected location for inspection; The similarity is for a single preset contour, the preset contour is compared with the characteristic contour, and the ratio of the area of ​​the obtained overlapping region to the total area of ​​the preset contour is recorded as the similarity.

3. The power transmission line image inspection system based on machine learning according to claim 2 is characterized in that: The inspection device acquires abnormal conditions at expected locations in a manner including: The inspection device is used to obtain detailed image information of the expected position through the image detector, and analyze the detailed image information through the image processing algorithm to detect the appearance defects of the transmission line, which include insulator damage, wire breakage, and hardware corrosion; When the inspection device cannot detect the appearance defect based on the detailed image information, it takes the expected location as the center and the transmission line as the extension direction, and acquires a number of detailed image information on both sides until the appearance defect is identified, or the moving distance exceeds the corresponding inspection area; The analysis layer records the distance between the site where the appearance defect is identified and the expected site as the deviation distance.

4. The power transmission line image inspection system based on machine learning according to claim 3 is characterized in that: The analysis layer determines whether the operating parameters of the acquisition layer are qualified based on the quantitative statistics of each inspection accuracy level, after completing the division of the inspection accuracy level of the expected locations in each area to be inspected, including: If the number of first-level sites among the expected sites is the largest, the operating parameters of the acquisition layer are determined to be qualified, and the acquisition layer is controlled to continue to use the current operating parameters to detect the transmission line; If the number of secondary sites divided among the expected sites is the largest, then whether the operating parameters of the acquisition layer are qualified is determined based on the time distribution parameters of each secondary site; If the number of third-level sites among the expected sites is the largest, the operating parameters of the acquisition layer are determined to be abnormal, and the preset similarity is adjusted to the corresponding value based on the proportion of the number of third-level sites.

5. The power transmission line image inspection system based on machine learning according to claim 4 is characterized in that: The analysis layer is used to determine whether the operating parameters of the acquisition layer are qualified based on the time distribution parameters of each secondary site, including: It is used to determine the time distribution parameters of each secondary site, determine the time node for the acquisition layer to obtain the detection image information corresponding to each secondary site, calculate the interval length between each time node, solve the average value of each interval length, and obtain the time distribution parameters; If the time distribution parameter is less than or equal to the preset time distribution parameter, it is determined that the operation parameter of the acquisition layer is qualified, and the transmission bandwidth of the transmission layer is adjusted to a corresponding value based on the total number of secondary sites; If the time distribution parameter is greater than a preset time distribution parameter, the area of ​​the area to be inspected is adjusted to a corresponding value based on the time distribution parameter.

6. The power transmission line image inspection system based on machine learning according to claim 5 is characterized in that: The analysis layer is used to adjust the area of ​​the area to be inspected to a corresponding value based on the time distribution parameter, wherein: The reduction range of the area of ​​the to-be-inspected region determined based on the time distribution parameter is inversely proportional to the time distribution parameter.

7. The power transmission line image inspection system based on machine learning according to claim 6 is characterized in that: The analysis layer is used to adjust the transmission bandwidth of the transmission layer to a corresponding value based on the total number of secondary sites, wherein: The increase in transmission bandwidth determined based on the total number of secondary sites is proportional to the total number of secondary sites.

8. The power transmission line image inspection system based on machine learning according to claim 7 is characterized in that: The analysis layer is used to adjust the preset similarity to a corresponding value based on the proportion of the number of the third-level sites, wherein: The analysis layer is used to record the ratio of the number of tertiary sites to the total number of each expected site as the number ratio of tertiary sites; The increase in the preset similarity determined based on the quantity ratio is proportional to the quantity ratio.

9. The power transmission line image inspection system based on machine learning according to claim 8, characterized in that: The analysis layer, under the condition of completing the adjustment of the preset similarity, re-classifies the inspection accuracy level of each expected location according to the acquired detection image information; if it is still determined that the number of the third-level locations among the expected locations is the largest, the operating height of each drone is adjusted to a corresponding value based on the ratio of the re-acquired number proportion to the number proportion acquired in the previous cycle; The reduction in the operating altitude determined based on the quantity ratio is proportional to the quantity ratio.

Images