A distribution network tower risk assessment method, device, equipment and storage medium

Abstract

The present invention relates to the field of image processing technology, and discloses a distribution network tower risk assessment method, device, equipment and storage medium. The method comprises: using an improved Yolov9‑c recognition model to identify the original image of the distribution network tower, and obtaining a rectangular frame of the tower base area, a rectangular frame of the tower defect area, and a mask of the environmental hazard area; using a preset defect priority strategy to determine the defect state of the distribution network tower; using a preset hazard index calculation rule to determine the type of environmental hazard of the distribution network tower and its degree of influence; according to the defect state of the distribution network tower, the type of environmental hazard and the degree of influence, the distribution network tower is risk assessed to obtain risk prompts and risk warnings of the distribution network tower. The present application uses a deep learning algorithm, a preset defect priority strategy and a preset hazard index calculation rule to comprehensively conduct risk assessment of distribution network towers, thereby improving accuracy and efficiency.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a distribution network tower risk assessment method, device, equipment and storage medium. Background Art

[0002] As a bridge between the power system and power users, distribution network towers play a vital role in ensuring the safety, efficiency and stability of power transmission. Due to the complex and changeable surrounding environment, distribution network towers are often damaged by external forces, environmental corrosion, natural disasters, etc., which leads to frequent power outages. Therefore, in order to ensure the stable operation of the power system, it is very important to conduct risk assessment of distribution network towers in a timely manner and eliminate hidden dangers in the distribution network environment.

[0003] At present, the following two methods are often used to identify distribution network tower faults: (1) traditional image processing algorithms. Due to the changeable environment around distribution network towers, it is difficult to accurately separate the tower area from the complex background using traditional image processing algorithms. (2) deep learning recognition algorithms. Due to the large number of defects in distribution network towers and the complex hidden dangers around them, it is difficult to accurately assess the risks of distribution network towers directly using deep learning algorithms. Summary of the invention

[0004] In view of this, the purpose of the present invention is to overcome the deficiencies in the prior art and to provide a distribution network tower risk assessment method, device, equipment and storage medium.

[0005] The present invention provides the following technical solutions:

[0006] In a first aspect, an embodiment of the present disclosure provides a distribution network tower risk assessment method, the method comprising:

[0007] Acquire an original image of a distribution network tower, and use an improved Yolov9-c recognition model to recognize the original image to obtain a rectangular frame of a tower base area, a rectangular frame of a tower defect area, and a mask of an environmental hazard area of ​​the distribution network tower;

[0008] Determining the defect status of the distribution network tower using a preset defect priority strategy;

[0009] The type of environmental hazard of the distribution network tower and its impact degree are determined by using the preset hazard index calculation rules;

[0010] According to the defect status, environmental hazard type and impact degree of the distribution network tower, a risk assessment is performed on the distribution network tower to obtain risk prompts and risk warnings for the distribution network tower.

[0011] Optionally, the adopting a preset defect priority strategy to determine the defect status of the distribution network tower includes:

[0012] Delete the rectangular frames in the tower base area rectangular frame and the tower defect area rectangular frame whose confidence is less than a first preset confidence threshold and whose overlap rate is greater than a first preset overlap rate threshold, to obtain a filtered tower base area rectangular frame and a filtered tower defect area rectangular frame;

[0013] There are multiple defect status rectangular boxes with different defect hazard levels in the filtered tower defect area rectangular box, and it is judged whether there are corresponding defect status rectangular boxes in the filtered tower defect area rectangular box in descending order of defect hazard levels;

[0014] If it exists, the variation IOU value between the corresponding defect state rectangular box and the other defect state rectangular boxes is calculated according to the preset variation IOU threshold calculation formula, and the defect state of the distribution network tower is determined according to the variation IOU value;

[0015] Among them, the preset variation IOU threshold calculation formula is:

[0016]

[0017] In the formula, is the variant IOU value, is the intersection area of ​​the tower defect area rectangular box ABCD and the tower defect area rectangular box EFGH, is the area of ​​the rectangular box ABCD of the tower defect area, is the area of ​​the rectangular frame EFGH in the tower defect area, is the horizontal coordinate of the upper left point A of the rectangular box ABCD of the tower defect area, is the ordinate of the upper left point A of the rectangular box ABCD of the tower defect area, is the horizontal coordinate of the lower right point C of the rectangular box ABCD of the tower defect area, is the ordinate of the lower right point C of the rectangular box ABCD of the tower defect area, is the horizontal coordinate of the upper left point E of the rectangular box EFGH of the tower defect area, is the ordinate of the upper left point E of the rectangular box EFGH of the tower defect area, is the horizontal coordinate of the lower right point G of the rectangular box EFGH of the tower defect area, It is the ordinate of the lower right point G of the rectangular box EFGH in the tower defect area.

[0018] Optionally, the using of a preset hidden danger index calculation rule to determine the type of environmental hidden danger of the distribution network tower and its impact degree includes:

[0019] Deleting the masks whose confidence is less than a second preset confidence threshold and whose overlap rate is greater than a second preset overlap rate threshold in the environmental hazard area masks, to obtain a filtered environmental hazard area mask;

[0020] Calculate the mask IOU value between the filtered tower base area rectangular frame and the filtered environmental hazard area mask according to a preset mask IOU threshold calculation formula, and determine the maximum mask IOU value;

[0021] Calculate the Euclidean distance between the center point of the filtered rectangular frame of the pole tower base area and the contour point of the filtered environmental hazard area mask according to a preset Euclidean ratio calculation formula, determine the minimum Euclidean distance, calculate the hypotenuse length of the filtered rectangular frame of the pole tower base area, and use the ratio of the minimum Euclidean distance to the hypotenuse length as the minimum Euclidean ratio;

[0022] The type of environmental hazard of the distribution network tower and its impact degree are determined according to the maximum mask IOU value and the minimum Euclidean ratio.

[0023] Optionally, the preset mask IOU threshold calculation formula is:

[0024]

[0025] In the formula, is the mask IOU value, is the mask A image of the environmental hazard area, is the mask B image of the environmental hazard area, is the intersection operation, For sum operation;

[0026] The preset European ratio calculation formula is:

[0027]

[0028] In the formula, is the minimum Euclidean ratio, is the horizontal coordinate of the upper left point M of the rectangular frame of the tower base area after filtering, is the ordinate of the upper left point M of the rectangular frame of the tower base area after filtering, is the horizontal coordinate of the lower right point N of the rectangular frame of the tower base area after filtering, is the ordinate of the lower right point N of the rectangular frame of the tower base area after filtering, is the contour point of the filtered environmental hazard area mask, is the contour point of the filtered environmental hazard area mask The horizontal axis of is the contour point of the filtered environmental hazard area mask The vertical coordinate of is the Euclidean distance from the center point of the filtered rectangular frame of the pole tower base area to the contour point of the i-th filtered environmental hazard area mask, is the minimum Euclidean distance, is the length of the hypotenuse of the rectangular frame of the tower base area after filtering.

[0029] Optionally, the improved Yolov9-c recognition model includes a backbone network, a feature pyramid network and a decoupling head, the feature pyramid network includes a first self-attention convolution module, a second self-attention convolution module, a first convolution module and a second convolution module, the decoupling head includes a target detection module and an instance segmentation module, and the improved Yolov9-c recognition model is used to recognize the original image to obtain a rectangular frame of the tower base area of ​​the distribution network tower, a rectangular frame of the tower defect area, and an environmental hazard area mask, including:

[0030] Inputting the original image into the backbone network to obtain a first-layer feature map, a second-layer feature map, and a third-layer feature map;

[0031] Inputting the third layer feature map and the second layer feature map into the first self-attention convolution module to obtain a first self-attention convolution feature map, and cascading the first self-attention convolution feature map and the second layer feature map to obtain a first fusion map;

[0032] Inputting the first fusion map and the first layer feature map into the second self-attention convolution module to obtain a second self-attention convolution feature map, and cascading the second self-attention convolution feature map with the first layer feature map to obtain a second fusion map;

[0033] Inputting the second fusion map into the first convolution module to obtain a first feature map, and cascading the first feature map with the first fusion map to obtain a third fusion map;

[0034] Inputting the third fusion map into the second convolution module to obtain a second feature map, and cascading the second feature map with the third layer feature map to obtain a fourth fusion map;

[0035] Input the second fusion image, the third fusion image and the fourth fusion image into the target detection module to obtain a rectangular frame of a tower base area and a rectangular frame of a tower defect area of ​​the distribution network tower;

[0036] The second fusion map, the third fusion map and the fourth fusion map are input into the instance segmentation module to obtain the environmental hazard area mask of the distribution network tower.

[0037] Optionally, inputting the third layer feature map and the second layer feature map into the first self-attention convolution module to obtain a first self-attention convolution feature map includes:

[0038] Perform two 1×1 convolutions on the second layer feature map to obtain the first projection matrix and the second projection matrix respectively;

[0039] Performing similarity matching on the first projection matrix and the second projection matrix to obtain a similarity matching matrix, and processing the similarity matching matrix through an activation function to obtain an attention weighted matrix;

[0040] The third layer feature map is sequentially subjected to 1×1 convolution and upsampling to obtain an upsampled feature map, and the upsampled feature map is subjected to 1×1 convolution to obtain a projection matrix;

[0041] Perform a matrix product operation on the projection matrix and the attention weighted matrix to obtain the first self-attention convolution feature map.

[0042] Optionally, the defect state includes a normal state, a damaged state, a tilted state, a broken state and a fallen state, and the impact degree includes a safe degree, a slight degree, a serious degree and a critical degree. The risk assessment of the distribution network tower is performed according to the defect state, the type of environmental hazard and the impact degree of the distribution network tower to obtain the risk prompt and risk warning of the distribution network tower, including:

[0043] If the defect state is normal, the environmental hazard type exists and the impact degree is slight, then the environmental hazard area and the impact degree are recorded;

[0044] If the defect state is normal, the environmental hazard type exists and the impact level is severe or critical, then the risk warning is generated, and the environmental hazard area and the impact level are recorded;

[0045] If the defect state is a damaged state, the environmental hazard type does not exist or the impact level is a safe level, then the defect state and the coordinates of the rectangular frame of the tower defect area are recorded;

[0046] If the defect state is a damaged state, the environmental hazard type exists and the impact degree is slight, then the defect state, the coordinates of the rectangular frame of the tower defect area, the environmental hazard area and the impact degree are recorded;

[0047] If the defect state is a damaged state, the environmental hazard type exists and the impact level is severe or critical, the risk warning is generated, and the defect state, the coordinates of the rectangular frame of the tower defect area, the environmental hazard area and the impact level are recorded;

[0048] If the defect state is a tilted state, a broken state, or a fallen state, and there is no environmental hazard type or the impact level is a safe level, then the risk warning is generated, and the defect state and the coordinates of the rectangular frame of the tower defect area are recorded;

[0049] If the defect state is a tilted state, a broken state, or a fallen state, the environmental hazard type exists and the impact level is mild, severe, or critical, then the risk alarm is generated, and the defect state, the coordinates of the rectangular frame of the tower defect area, the environmental hazard area, and the impact level are recorded.

[0050] In a second aspect, an embodiment of the present disclosure provides a distribution network tower risk assessment device, the device comprising:

[0051] The recognition module is used to obtain the original image of the distribution network tower, and use the improved Yolov9-c recognition model to recognize the original image to obtain the rectangular frame of the tower base area, the rectangular frame of the tower defect area, and the mask of the environmental hazard area of ​​the distribution network tower;

[0052] A first determination module is used to determine the defect status of the distribution network tower using a preset defect priority strategy;

[0053] The second determination module is used to determine the type of environmental hazard of the distribution network tower and its impact degree by using a preset hazard index calculation rule;

[0054] The evaluation module is used to perform risk evaluation on the distribution network tower according to the defect status, environmental hazard type and impact degree of the distribution network tower, and obtain risk prompts and risk warnings of the distribution network tower.

[0055] In a third aspect, a computer device is provided in an embodiment of the present disclosure, the computer device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the steps of the distribution network tower risk assessment method described in the first aspect when executing the computer program.

[0056] In a fourth aspect, a computer-readable storage medium is provided in an embodiment of the present disclosure, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the distribution network tower risk assessment method described in the first aspect are implemented.

[0057] Beneficial effects of this application:

[0058] The distribution network tower risk assessment method provided in the embodiment of the present application obtains the original image of the distribution network tower, and uses the improved Yolov9-c recognition model to identify the original image to obtain the tower base area rectangular frame, tower defect area rectangular frame, and environmental hazard area mask of the distribution network tower; uses a preset defect priority strategy to determine the defect state of the distribution network tower; uses a preset hazard index calculation rule to determine the environmental hazard type and its impact degree of the distribution network tower; according to the defect state, environmental hazard type and impact degree of the distribution network tower, the distribution network tower is risk assessed to obtain the risk prompt and risk warning of the distribution network tower. The present application uses a deep learning algorithm to automatically separate the distribution network tower base area, defect state, and hazard type from a complex background, and uses a preset defect priority strategy to determine the tower defect state, and uses a preset hazard index calculation rule to determine the hazard type and impact degree, and comprehensively conducts risk assessment of the distribution network tower, significantly improving the accuracy and efficiency of the assessment, and providing more scientific and reliable decision support for operation and maintenance personnel.

[0059] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and understandable, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work. In each of the drawings, similar components are numbered similarly.

[0061] Figure 1 A flow chart of a distribution network tower risk assessment method provided by an embodiment of the present application is shown;

[0062] Figure 2 A schematic diagram of the network structure of an improved Yolov9-c recognition model provided in an embodiment of the present application is shown;

[0063] Figure 3 A schematic diagram of the network structure of a self-attention convolution module provided in an embodiment of the present application is shown;

[0064] Figure 4 A schematic structural diagram of a distribution network tower risk assessment device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0065] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0066] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0067] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein in the specification of the template are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0068] Example 1

[0069] like Figure 1 , which is a flow chart of a distribution network tower risk assessment method in an embodiment of the present application. The distribution network tower risk assessment method provided in an embodiment of the present application includes the following steps:

[0070] Step S110, obtaining the original image of the distribution network tower, using the improved Yolov9-c recognition model to recognize the original image, and obtaining the rectangular frame of the tower base area, the rectangular frame of the tower defect area, and the mask of the environmental hazard area of ​​the distribution network tower.

[0071] In this embodiment, after the original image of the distribution network tower is obtained, in order to accurately and quickly perform risk assessment of the distribution network tower, the improved Yolov9-c recognition model is used to identify the original image. Generally speaking, the identification of tower defects and environmental hazards adopts a two-stage strategy, namely: the first stage is to identify tower defects, and the second stage is to identify environmental hazards. In order to quickly obtain tower defects and environmental hazards, the present application embeds the two-stage strategy into the network structure of the Yolov9-c recognition model, and introduces a self-attention convolution module (SACM), which can quickly and accurately extract the tower base area rectangular frame, tower defect area rectangular frame, and environmental hazard area mask of the distribution network tower.

[0072] Specifically, Figure 2As shown in the figure, the improved Yolov9-c recognition model includes a backbone network, a feature pyramid network and a decoupling head. The feature pyramid network includes a first self-attention convolution module, a second self-attention convolution module, a first convolution module and a second convolution module. The decoupling head includes a target detection module and an instance segmentation module. First, the original image passes through the backbone network (Conv3~Conv5) constructed by the RepNSCPELAN4 module to obtain the first layer feature map F3, the second layer feature map F4 and the third layer feature map F5. Then F3~F5 are input into the improved feature pyramid network PA-SACM-FPN, where the third layer feature map F5 and the second layer feature map F4 pass through the first self-attention convolution module to generate the first self-attention convolution feature map SACF4, and then the first self-attention convolution feature map SACF4 is cascaded with the second layer feature map F4 to generate the first fusion map ~P4. Then the first fusion map ~P4 and the first layer feature map F3 pass through the second self-attention convolution module to generate the second self-attention convolution feature map SACF3, and then the second self-attention convolution feature map SACF3 is cascaded with the first layer feature map F3 to generate the second fusion map P3. Then the second fusion map P3 passes through the first convolution module to generate the first feature map PAF4 with the same resolution as the first fusion map ~P4. The first feature map PAF4 is further cascaded with the first fusion map ~P4 to generate the third fusion map P4. Then the third fusion map P4 passes through the second convolution module to generate the second feature map PAF5 with the same resolution as the third layer feature map F5. The second feature map PAF5 is further cascaded with the third layer feature map F5 to generate the fourth fusion map P5, thus obtaining a cross-level fusion map (P3~P5) from top to bottom and bottom-up.

[0073] Finally, the fusion image (P3~P5) passes through the decoupling head, among which the second fusion image P3, the third fusion image P4 and the fourth fusion image P5 are subjected to target detection (including classification and regression) through the target detection module to obtain the tower base type, the tower base area rectangular frame and the tower defect state, and the tower defect area rectangular frame of the distribution network tower. The second fusion image P3, the third fusion image P4 and the fourth fusion image P5 are subjected to instance segmentation (including classification, regression and segmentation) through the instance segmentation module to obtain the environmental hazard type, the environmental hazard area rectangular frame and the environmental hazard area mask of the distribution network tower.

[0074] It should be noted that in Figure 2 In the example, concat refers to the operation of splicing multiple images along a certain dimension to facilitate subsequent calculations.

[0075] It should be noted that if Figure 3As shown in the figure, the working principle of the self-attention convolution module in the above process is as follows: Taking the third-layer feature map F5 and the second-layer feature map F4 as inputs to the first self-attention convolution module to obtain the first self-attention convolution feature map SACF4 as an example, firstly, the low-level high-resolution feature map (the second-layer feature map F4) is subjected to two different 1×1 convolutions to obtain the first projection matrix Q and the second projection matrix K respectively. Then, the first projection matrix Q and the second projection matrix K are similarly matched to obtain a similarity matching matrix, which is then processed through the activation function to obtain the attention weighting matrix W. Then, the high-level low-resolution feature map (the third-layer feature map F5) is sequentially subjected to 1×1 convolution and upsampling to obtain an upsampled feature map with the same resolution as the low-level high-resolution feature map. At this time, the upsampled feature map is subjected to 1×1 convolution to obtain the projection matrix V. Finally, the projection matrix V is matrix-producted with the attention weighting matrix W to obtain the first self-attention convolution feature map SACF4. The specific formula is as follows:

[0076]

[0077] In the formula, is the first self-attention convolutional feature map SACF4, is the vector dimension of similarity matching in the second projection matrix K, It is the flexible maximum transfer function operation.

[0078] It should be noted that, in the process of inputting the first fusion image ~P4 and the first layer feature map F3 into the second self-attention convolution module to obtain the second self-attention convolution feature map SACF3, the first fusion image ~P4 is a high-level low-resolution feature map, and the first layer feature map F3 is a low-level high-resolution feature map. The principle is consistent with the principle of the above-mentioned first self-attention convolution module, and this embodiment will not be repeated here.

[0079] In order to quickly obtain the rectangular frame of the tower base area, the rectangular frame of the tower defect area and the mask of the environmental hazard area, the above method embeds the two-stage recognition strategy into the Yolov9-c algorithm network structure, which can simultaneously complete the target detection of the tower base and tower defects and the instance segmentation of environmental hazards. At the same time, the self-attention convolution module in the PA-SACM-FPN network is used to help the model focus on the key areas in the whole image, giving the high-level feature map the attention to the details of the low-level feature map, and also integrating the detail information of the low-level feature map, thereby improving the overall feature representation ability, and then effectively improving the accuracy of distribution network tower target detection and instance segmentation.

[0080] Step S120: Determine the defect status of the distribution network tower using a preset defect priority strategy.

[0081] Understandably, this application is divided into tower damage, tower tilt, tower fracture, and tower collapse according to the hazard level of distribution network tower defects, among which the hazard level of tower collapse is higher than that of tower fracture, the hazard level of tower fracture is higher than that of tower tilt, and the hazard level of tower tilt is higher than that of tower damage. Based on the hazard level of distribution network towers, this application formulates a preset defect priority strategy to further determine the defect status of distribution network towers.

[0082] Specifically, first, according to the recognition result in step S110, the rectangular frame of the pole tower base area and the rectangular frame of the pole tower defect area (including pole tower damage, pole tower tilt, pole tower fracture, and pole tower collapse) are screened out. Then, for the rectangular frame of the pole tower base area and each rectangular frame of the pole tower defect area, a first preset confidence threshold (e.g., 0.30) and a first preset overlap rate threshold (e.g., 0.70) are set, and the rectangular frames with confidence lower than the first preset confidence threshold are filtered out respectively. Then, the rectangular frames with overlap rate greater than the first preset overlap rate threshold are filtered out respectively, and the filtered rectangular frame of the pole tower base area and the filtered rectangular frame of the pole tower defect area are obtained. There are multiple defect state rectangular frames with different defect hazard levels in the filtered rectangular frame of the pole tower defect area.

[0083] Furthermore, according to the defect priority strategy, that is, in order from high to low defect hazard levels, it is determined whether there is a corresponding defect state rectangular box in the filtered tower defect area rectangular box. If so, the variation IOU value between the corresponding defect state rectangular box and the remaining defect state rectangular boxes is calculated according to the preset variation IOU threshold calculation formula, and the defect state of the distribution network tower is determined according to the variation IOU value.

[0084] In this embodiment, it is first determined whether there is a tower fall defect type with the highest hazard level in the filtered tower defect area rectangular frame. If not, the tower fracture defect type with the second highest hazard level is determined. If so, the variation IOU values ​​between all tower fall rectangular frames and other tower defect area rectangular frames are calculated based on the preset variation IOU threshold calculation formula, and all other tower defect area rectangular frames whose variation IOU values ​​are greater than the first preset variation IOU threshold (for example, 0.80) are deleted. In this way, a tower fall rectangular frame that does not contain other tower defect area rectangular frames is obtained, and it is determined that the distribution network tower in the rectangular frame is in a fallen state.

[0085] Furthermore, based on the remaining tower defect area rectangular boxes, it is then determined whether there is a tower fracture defect type with the second highest hazard level. If not, a tower tilt defect type with the second lowest hazard level is determined. If so, the variation IOU values ​​between all tower fracture rectangular boxes and other tower defect area rectangular boxes are calculated based on the preset variation IOU threshold calculation formula, and all other tower defect area rectangular boxes whose variation IOU values ​​are greater than the second preset variation IOU threshold (for example, 0.80) are deleted. In this way, a tower fracture rectangular box that does not include other tower defect area rectangular boxes is obtained, and the distribution network tower in the rectangular box is determined to be in a fractured state.

[0086] Furthermore, based on the remaining tower defect area rectangular boxes, it is determined whether there is a tower tilt defect type with the second lowest hazard level. If not, the tower damage defect type with the lowest hazard level is determined. If so, the variation IOU values ​​between all tower tilt rectangular boxes and other tower defect area rectangular boxes are calculated based on the preset variation IOU threshold calculation formula, and all other tower defect area rectangular boxes whose variation IOU values ​​are greater than the third preset variation IOU threshold (for example, 0.80) are deleted. In this way, a tower tilt rectangular box that does not contain other tower defect area rectangular boxes is obtained, and the distribution network tower in the rectangular box is determined to be in a tilted state.

[0087] Furthermore, based on the remaining defective rectangular boxes, it is finally determined whether there is a tower damage defect type with the lowest hazard level. If not, the distribution network tower is determined to be in a normal state. If so, the distribution network tower in the corresponding rectangular box is determined to be in a damaged state. The preset variation IOU threshold calculation formula is:

[0088]

[0089] In the formula, is the variation IOU value, is the intersection area of ​​the tower defect area rectangular box ABCD and the tower defect area rectangular box EFGH, is the area of ​​the rectangular box ABCD of the tower defect area, is the area of ​​the rectangular frame EFGH in the tower defect area, is the horizontal coordinate of the upper left point A of the rectangular box ABCD of the tower defect area, is the ordinate of the upper left point A of the rectangular box ABCD of the tower defect area, is the horizontal coordinate of the lower right point C of the rectangular box ABCD of the tower defect area, is the ordinate of the lower right point C of the rectangular box ABCD of the tower defect area, is the horizontal coordinate of the upper left point E of the rectangular box EFGH of the tower defect area, is the ordinate of the upper left point E of the rectangular box EFGH of the tower defect area, is the horizontal coordinate of the lower right point G of the rectangular box EFGH of the tower defect area, It is the ordinate of the lower right point G of the rectangular box EFGH in the tower defect area.

[0090] It is understandable that if there is no rectangular frame of the tower defect area, there is no need to calculate the variation IOU value between the rectangular frames of the tower defect area, which proves that the distribution network tower is in a normal state at this time.

[0091] It should be noted that the first preset variation IOU threshold, the second preset variation IOU threshold and the third preset variation IOU threshold may be the same value or different values, which is not limited in the embodiment of the present application.

[0092] The above method adopts the tower state determination strategy of defect priority and calculates the variation IOU value between rectangular boxes, which can not only effectively remove the overlapping boxes inside and outside the defective rectangular box, but also give priority to determining the defect state of the tower with a high hazard level.

[0093] Step S130, using a preset hidden danger index calculation rule to determine the type of environmental hidden danger of the distribution network tower and its impact degree.

[0094] Understandably, according to the actual environment of the distribution network tower, this application selects common environmental hazard types such as landslides, water accumulation at the pole base, and soil excavation at the pole base as hazard influencing factors. Considering the diverse changes in hazard forms, the use of traditional rectangular frame anchoring targets cannot accurately estimate the impact of environmental hazards on the tower. This application uses a rectangular frame to select the pole base area and a polygon to circle the hazard target, and then determines the type of environmental hazard of the distribution network tower and its impact according to the preset hazard index calculation rules.

[0095] Specifically, first, according to the recognition result in step S110, the environmental hazard area masks are screened out. Then, for each environmental hazard area mask, a second preset confidence threshold (e.g., 0.30) and a second preset overlap rate threshold (e.g., 0.70) are set, and masks with confidences lower than the second preset confidence threshold are filtered out respectively, and then masks with overlap rates greater than the second preset overlap rate threshold are filtered out respectively, to obtain filtered environmental hazard area masks.

[0096] Furthermore, in order to accurately measure the relationship between the pole tower and the hidden danger, the present invention sets a preset hidden danger index calculation rule: first, the mask IOU value between the filtered pole tower base area rectangular frame and multiple filtered environmental hidden danger area masks is calculated according to the preset mask IOU threshold calculation formula, and the maximum mask IOU value is determined; then, the Euclidean distance between the center point of the filtered pole tower base area rectangular frame and the contour point of the filtered environmental hidden danger area mask is calculated according to the preset Euclidean ratio calculation formula, and the minimum Euclidean distance is determined; then, the hypotenuse length of the filtered pole tower base area rectangular frame is calculated, and the ratio of the minimum Euclidean distance to the hypotenuse length is used as the minimum Euclidean ratio.

[0097] Finally, the types of environmental hazards of distribution network towers and their impact levels are determined based on the maximum mask IOU value and the minimum Euclidean ratio. Exemplarily, if the maximum mask IOU value is greater than 0, it means that the pole base is located in the environmental hazard area, and the environmental hazard type is determined to be a pole tower hazard type, that is, there is an environmental hazard, and the degree of impact of the environmental hazard on the distribution network pole tower is critical; if the maximum mask IOU value is 0, it means that the pole base is located outside the environmental hazard area, but if the minimum Euclidean ratio is less than or equal to 1, it means that the environmental hazard is very close to the distribution network pole tower, and the environmental hazard type is determined to be a pole tower hazard type, that is, there is an environmental hazard, and the degree of impact of the environmental hazard on the distribution network pole tower is severe; if the maximum mask IOU value is 0 and the minimum Euclidean ratio is greater than 1 but less than 2, the environmental hazard type is determined to be a pole tower hazard type, that is, there is an environmental hazard, and the degree of impact of the environmental hazard on the distribution network pole tower is slight; if the maximum mask IOU value is 0 and the minimum Euclidean ratio is greater than 2, the environmental hazard type is determined to be a pole tower hazard type, that is, there is an environmental hazard, and the degree of impact of the environmental hazard on the distribution network pole tower is safe. In addition, it should be noted that if there is no environmental hazard area mask, there is no need to calculate the mask IOU value and Euclidean ratio between the pole base area rectangular frame and the environmental hazard area mask. At this time, there is no environmental hazard on the distribution network tower.

[0098] It should be noted that regarding the process of obtaining the mask IOU value, it is assumed that the two mask images are , , the size of the image remains the same. It should be noted that the recognition result of environmental hazards in this application is a mask image. , we only need to use the image flood filling algorithm to fill the rectangular area of ​​the pole base to get the mask image Among them, the preset mask IOU threshold calculation formula is:

[0099]

[0100] In the formula, is the mask IOU value, is the mask A image of the environmental hazard area, is the mask B image of the environmental hazard area, represents the intersection operation, Represents a sum operation.

[0101] Regarding the process of obtaining the Euclidean ratio, it is necessary to first extract the coordinates of the contour points of the mask image. The recognition result of the environmental hazards in this application is a mask image, and the coordinates of the contour points of the mask image are found through the contour retrieval algorithm. Assume that the upper left point of the rectangular box of the tower base area is , the lower right point is , the mask contour points of the environmental hazard area are Among them, the preset European ratio calculation formula is:

[0102]

[0103] In the formula, is the minimum European ratio, is the horizontal coordinate of the upper left point M of the rectangular box of the tower base area after filtering, is the ordinate of the upper left point M of the rectangular box of the filtered tower base area, is the horizontal coordinate of the lower right point N of the rectangular box of the tower base area after filtering, is the ordinate of the lower right point N of the rectangular box of the filtered tower base area, is the contour point of the mask of the environmental hazard area after filtering, Contour points of the filtered environmental hazard area mask The horizontal axis of Contour points of the filtered environmental hazard area mask The vertical coordinate of is the Euclidean distance from the center point of the rectangular frame of the filtered tower base area to the contour point of the i-th filtered environmental hazard area mask, is the minimum Euclidean distance, It is the length of the hypotenuse of the rectangular box of the filtered tower base area.

[0104] It should be noted that the first preset confidence threshold and the second preset confidence threshold can be the same value or different values, and the first preset overlap rate threshold and the second preset overlap rate threshold can be the same value or different values, and the embodiment of the present application does not limit this.

[0105] The above method adopts the tower hidden danger determination strategy based on the hidden danger index calculation rule, and calculates the mask IOU value and Euclidean ratio between the tower base and the hidden danger, which can not only effectively determine the type of tower hidden danger, but also accurately determine the degree of impact.

[0106] Step S140, performing risk assessment on the distribution network towers according to the defect status, environmental hazard type and impact degree of the distribution network towers, and obtaining risk prompts and risk warnings for the distribution network towers.

[0107] Finally, based on the defect status, environmental hazard type and impact degree of the distribution network tower obtained in the above steps, the distribution network tower is risk assessed to obtain risk prompts and risk warnings of the distribution network tower. Among them, the defect status includes normal state, damaged state, tilted state, broken state and fallen state, the environmental hazard type includes landslide, water accumulation at the pole base and soil excavation at the pole base, and the impact degree includes safety degree, slight degree, serious degree and critical degree.

[0108] Specifically, if the defect status of the distribution network tower is normal, but there is no environmental hazard type or the impact level of the environmental hazard is safe, no risk warning will be generated, nor will a risk alarm be generated; if the defect status of the distribution network tower is normal, there is an environmental hazard type and the impact level is slight, no risk warning will be generated, but the environmental hazard type and impact level need to be recorded; if the defect status of the distribution network tower is normal, there is an environmental hazard type and the impact level is severe and critical, then a risk warning will be generated, and the environmental hazard type and impact level will be recorded.

[0109] Furthermore, if the defect state of the distribution network tower is a damaged state, but there is no environmental hazard type or the impact level is a safe level, no risk warning is generated, and no risk alarm is generated, and the defect state and the coordinates of the rectangular box of the tower defect area are recorded; if the defect state of the distribution network tower is a damaged state, there is an environmental hazard type and the impact level is a slight level, no risk warning is generated, and no risk alarm is generated, and the defect state, the coordinates of the rectangular box of the tower defect area, the environmental hazard type and the impact level are recorded; if the defect state of the distribution network tower is a damaged state, there is an environmental hazard type and the impact level is a serious level and a critical level, a risk warning is generated, and the defect state, the coordinates of the rectangular box of the tower defect area, the environmental hazard type and the impact level are recorded.

[0110] Furthermore, if the defect state of the distribution network tower is a tilted state, a broken state, or a fallen state, there is no environmental hazard type or the impact level is safe, a risk alarm is generated, and the defect state and the coordinates of the rectangular box of the tower defect area are recorded; if the defect state of the distribution network tower is a tilted state, a broken state, or a fallen state, there is an environmental hazard type and the impact level is mild or severe and critical, a risk alarm is generated, and the defect state, the coordinates of the rectangular box of the tower defect area, the environmental hazard type, and the impact level are recorded.

[0111] The above method conducts risk assessment on distribution network towers according to the defect status, environmental hidden danger type and impact degree of the distribution network towers, and then generates risk prompts and risk warnings, which helps to timely discover and deal with potential safety hazards, prevent accidents, ensure the safe and stable operation of the distribution network, and significantly reduce the frequency and intensity of manual inspections, thereby improving operation and maintenance efficiency. At the same time, since potential risks can be discovered and dealt with in a timely manner, the overall safety of the distribution network is also improved.

[0112] The distribution network tower risk assessment method provided in the embodiment of the present application automatically separates the distribution network tower base area, defect status, and hidden danger type from the complex background by using a deep learning algorithm, and determines the tower defect status using a preset defect priority strategy, and determines the hidden danger type and impact degree using a preset hidden danger indicator calculation rule, and comprehensively conducts risk assessment of distribution network towers, significantly improving the accuracy and efficiency of the assessment, and providing more scientific and reliable decision support for operation and maintenance personnel.

[0113] Example 2

[0114] like Figure 4 , which is a schematic diagram of the structure of a distribution network tower risk assessment device 400 in an embodiment of the present application, and the device includes:

[0115] The recognition module 410 is used to obtain the original image of the distribution network tower, and use the improved Yolov9-c recognition model to recognize the original image to obtain the rectangular frame of the tower base area, the rectangular frame of the tower defect area, and the mask of the environmental hazard area of ​​the distribution network tower;

[0116] A first determination module 420 is used to determine the defect status of the distribution network tower using a preset defect priority strategy;

[0117] The second determination module 430 is used to determine the type of environmental hazard of the distribution network tower and its impact degree by using a preset hazard index calculation rule;

[0118] The evaluation module 440 is used to perform risk evaluation on the distribution network towers according to the defect status, environmental hazard type and impact degree of the distribution network towers, and obtain risk prompts and risk warnings of the distribution network towers.

[0119] The distribution network tower risk assessment device provided in the embodiment of the present application automatically separates the distribution network tower base area, defect status, and hidden danger type from the complex background by using a deep learning algorithm, and determines the tower defect status using a preset defect priority strategy, and determines the hidden danger type and impact degree using a preset hidden danger indicator calculation rule, and comprehensively conducts risk assessment of the distribution network tower, significantly improving the accuracy and efficiency of the assessment, and providing more scientific and reliable decision support for operation and maintenance personnel.

[0120] A computer device is also provided in an embodiment of the present disclosure. The computer device includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, the steps of the distribution network tower risk assessment method in Example 1 are implemented.

[0121] A computer-readable storage medium is also provided in an embodiment of the present disclosure. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the steps of the distribution network tower risk assessment method in Example 1 are implemented.

[0122] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and structure diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in an alternative implementation, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the structure diagram and / or the flow diagram, and the combination of boxes in the structure diagram and / or the flow diagram, can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0123] In addition, the functional modules or units in the various embodiments of the present invention may be integrated together to form an independent part, or each module may exist independently, or two or more modules may be integrated to form an independent part.

[0124] If the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a smart phone, a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium can be a non-volatile storage medium or a volatile storage medium. For example, the storage medium can be: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a disk or an optical disk, and other media that can store program codes.

[0125] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A distribution network tower risk assessment method, characterized in that: The method comprises: Acquire an original image of a distribution network tower, and use an improved Yolov9-c recognition model to recognize the original image to obtain a rectangular frame of a tower base area, a rectangular frame of a tower defect area, and a mask of an environmental hazard area of ​​the distribution network tower; Determining the defect status of the distribution network tower using a preset defect priority strategy; The type of environmental hazard of the distribution network tower and its impact degree are determined by using the preset hazard index calculation rules; According to the defect status, environmental hazard type and impact degree of the distribution network tower, risk assessment is performed on the distribution network tower to obtain risk prompts and risk warnings for the distribution network tower; Among them, the improved Yolov9-c recognition model includes a backbone network, a feature pyramid network and a decoupling head, the feature pyramid network includes a first self-attention convolution module, a second self-attention convolution module, a first convolution module and a second convolution module, the decoupling head includes a target detection module and an instance segmentation module, and the improved Yolov9-c recognition model is used to recognize the original image to obtain a rectangular frame of the tower base area of ​​the distribution network tower, a rectangular frame of the tower defect area, and an environmental hazard area mask, including: Inputting the original image into the backbone network to obtain a first-layer feature map, a second-layer feature map, and a third-layer feature map; Inputting the third layer feature map and the second layer feature map into the first self-attention convolution module to obtain a first self-attention convolution feature map, and cascading the first self-attention convolution feature map and the second layer feature map to obtain a first fusion map; Inputting the first fusion map and the first layer feature map into the second self-attention convolution module to obtain a second self-attention convolution feature map, and cascading the second self-attention convolution feature map with the first layer feature map to obtain a second fusion map; Inputting the second fusion map into the first convolution module to obtain a first feature map, and cascading the first feature map with the first fusion map to obtain a third fusion map; Inputting the third fusion map into the second convolution module to obtain a second feature map, and cascading the second feature map with the third layer feature map to obtain a fourth fusion map; Input the second fusion image, the third fusion image and the fourth fusion image into the target detection module to obtain a rectangular frame of a tower base area and a rectangular frame of a tower defect area of ​​the distribution network tower; Inputting the second fusion map, the third fusion map and the fourth fusion map into the instance segmentation module to obtain an environmental hazard area mask of the distribution network tower; The step of inputting the third layer feature map and the second layer feature map into the first self-attention convolution module to obtain a first self-attention convolution feature map includes: Perform two 1×1 convolutions on the second layer feature map to obtain the first projection matrix and the second projection matrix respectively; Performing similarity matching on the first projection matrix and the second projection matrix to obtain a similarity matching matrix, and processing the similarity matching matrix through an activation function to obtain an attention weighted matrix; The third layer feature map is sequentially subjected to 1×1 convolution and upsampling to obtain an upsampled feature map, and the upsampled feature map is subjected to 1×1 convolution to obtain a projection matrix; Perform a matrix product operation on the projection matrix and the attention weighted matrix to obtain the first self-attention convolution feature map.

2. The distribution network tower risk assessment method according to claim 1, characterized in that: The method of using a preset defect priority strategy to determine the defect status of the distribution network tower includes: Delete the rectangular frames in the tower base area rectangular frame and the tower defect area rectangular frame whose confidence is less than a first preset confidence threshold and whose overlap rate is greater than a first preset overlap rate threshold, to obtain a filtered tower base area rectangular frame and a filtered tower defect area rectangular frame; There are multiple defect status rectangular boxes with different defect hazard levels in the filtered tower defect area rectangular box, and it is judged whether there are corresponding defect status rectangular boxes in the filtered tower defect area rectangular box in descending order of defect hazard levels; If it exists, the variation IOU value between the corresponding defect state rectangular box and the other defect state rectangular boxes is calculated according to the preset variation IOU threshold calculation formula, and the defect state of the distribution network tower is determined according to the variation IOU value; Among them, the preset variation IOU threshold calculation formula is: In the formula, is the variant IOU value, is the intersection area of ​​the tower defect area rectangular box ABCD and the tower defect area rectangular box EFGH, is the area of ​​the rectangular box ABCD of the tower defect area, is the area of ​​the rectangular frame EFGH in the tower defect area, is the horizontal coordinate of the upper left point A of the rectangular box ABCD of the tower defect area, is the ordinate of the upper left point A of the rectangular box ABCD of the tower defect area, is the horizontal coordinate of the lower right point C of the rectangular box ABCD of the tower defect area, is the ordinate of the lower right point C of the rectangular box ABCD of the tower defect area, is the horizontal coordinate of the upper left point E of the rectangular box EFGH of the tower defect area, is the ordinate of the upper left point E of the rectangular box EFGH of the tower defect area, is the horizontal coordinate of the lower right point G of the rectangular box EFGH of the tower defect area, It is the ordinate of the lower right point G of the rectangular box EFGH in the tower defect area.

3. The distribution network tower risk assessment method according to claim 2, characterized in that: The method of using a preset hidden danger index calculation rule to determine the type of environmental hidden danger of the distribution network tower and its impact degree includes: Deleting the masks whose confidence is less than a second preset confidence threshold and whose overlap rate is greater than a second preset overlap rate threshold in the environmental hazard area masks, to obtain a filtered environmental hazard area mask; Calculate the mask IOU value between the filtered tower base area rectangular frame and the filtered environmental hazard area mask according to a preset mask IOU threshold calculation formula, and determine the maximum mask IOU value; Calculate the Euclidean distance between the center point of the filtered rectangular frame of the pole tower base area and the contour point of the filtered environmental hazard area mask according to a preset Euclidean ratio calculation formula, determine the minimum Euclidean distance, calculate the hypotenuse length of the filtered rectangular frame of the pole tower base area, and use the ratio of the minimum Euclidean distance to the hypotenuse length as the minimum Euclidean ratio; The type of environmental hazard of the distribution network tower and its impact degree are determined according to the maximum mask IOU value and the minimum Euclidean ratio.

4. The distribution network tower risk assessment method according to claim 3 is characterized in that: The preset mask IOU threshold calculation formula is: In the formula, is the mask IOU value, is the mask A image of the environmental hazard area, is the mask B image of the environmental hazard area, is the intersection operation, For sum operation; The preset European ratio calculation formula is: In the formula, is the minimum Euclidean ratio, is the horizontal coordinate of the upper left point M of the rectangular frame of the tower base area after filtering, is the ordinate of the upper left point M of the rectangular frame of the tower base area after filtering, is the horizontal coordinate of the lower right point N of the rectangular frame of the tower base area after filtering, is the ordinate of the lower right point N of the rectangular frame of the tower base area after filtering, is the contour point of the filtered environmental hazard area mask, is the contour point of the filtered environmental hazard area mask The horizontal axis of is the contour point of the filtered environmental hazard area mask The vertical coordinate of is the Euclidean distance from the center point of the filtered rectangular frame of the pole tower base area to the contour point of the i-th filtered environmental hazard area mask, is the minimum Euclidean distance, is the length of the hypotenuse of the rectangular frame of the tower base area after filtering.

5. The distribution network tower risk assessment method according to claim 1, characterized in that: The defect state includes normal state, damaged state, tilted state, broken state and fallen state, and the impact degree includes safety degree, slight degree, serious degree and critical degree. According to the defect state of the distribution network tower, the type of environmental hazard and the impact degree, the risk assessment of the distribution network tower is performed to obtain the risk prompt and risk warning of the distribution network tower, including: If the defect state is normal, the environmental hazard type exists and the impact degree is slight, then the environmental hazard type and the impact degree are recorded; If the defect state is normal, the environmental hazard type exists and the impact level is severe or critical, then the risk warning is generated and the environmental hazard type and the impact level are recorded; If the defect state is a damaged state, the environmental hazard type does not exist or the impact level is a safe level, then the defect state and the coordinates of the rectangular frame of the tower defect area are recorded; If the defect state is a damaged state, the environmental hazard type exists and the impact degree is slight, then the defect state, the coordinates of the rectangular frame of the tower defect area, the environmental hazard type and the impact degree are recorded; If the defect state is a damaged state, the environmental hazard type exists and the impact level is severe or critical, the risk warning is generated, and the defect state, the coordinates of the rectangular frame of the tower defect area, the environmental hazard type and the impact level are recorded; If the defect state is a tilted state, a broken state, or a fallen state, and there is no environmental hazard type or the impact level is a safe level, then the risk warning is generated, and the defect state and the coordinates of the rectangular frame of the tower defect area are recorded; If the defect state is a tilted state, a broken state, or a fallen state, the environmental hazard type exists and the impact level is mild, severe, or critical, then the risk alarm is generated, and the defect state, the coordinates of the rectangular box of the tower defect area, the environmental hazard type, and the impact level are recorded.

6. A distribution network tower risk assessment device, characterized in that: The device comprises: The recognition module is used to obtain the original image of the distribution network tower, and use the improved Yolov9-c recognition model to recognize the original image to obtain the rectangular frame of the tower base area, the rectangular frame of the tower defect area, and the mask of the environmental hazard area of ​​the distribution network tower; A first determination module is used to determine the defect status of the distribution network tower using a preset defect priority strategy; The second determination module is used to determine the type of environmental hazard of the distribution network tower and its impact degree by using a preset hazard index calculation rule; An evaluation module is used to perform risk evaluation on the distribution network tower according to the defect status, environmental hazard type and impact degree of the distribution network tower, and obtain risk prompts and risk warnings for the distribution network tower; Among them, the improved Yolov9-c recognition model includes a backbone network, a feature pyramid network and a decoupling head, the feature pyramid network includes a first self-attention convolution module, a second self-attention convolution module, a first convolution module and a second convolution module, the decoupling head includes a target detection module and an instance segmentation module, and the improved Yolov9-c recognition model is used to recognize the original image to obtain a rectangular frame of the tower base area of ​​the distribution network tower, a rectangular frame of the tower defect area, and an environmental hazard area mask, including: Inputting the original image into the backbone network to obtain a first-layer feature map, a second-layer feature map, and a third-layer feature map; Inputting the third layer feature map and the second layer feature map into the first self-attention convolution module to obtain a first self-attention convolution feature map, and cascading the first self-attention convolution feature map and the second layer feature map to obtain a first fusion map; Inputting the first fusion map and the first layer feature map into the second self-attention convolution module to obtain a second self-attention convolution feature map, and cascading the second self-attention convolution feature map with the first layer feature map to obtain a second fusion map; Inputting the second fusion map into the first convolution module to obtain a first feature map, and cascading the first feature map with the first fusion map to obtain a third fusion map; Inputting the third fusion map into the second convolution module to obtain a second feature map, and cascading the second feature map with the third layer feature map to obtain a fourth fusion map; Input the second fusion image, the third fusion image and the fourth fusion image into the target detection module to obtain a rectangular frame of a tower base area and a rectangular frame of a tower defect area of ​​the distribution network tower; Inputting the second fusion map, the third fusion map and the fourth fusion map into the instance segmentation module to obtain an environmental hazard area mask of the distribution network tower; The step of inputting the third layer feature map and the second layer feature map into the first self-attention convolution module to obtain a first self-attention convolution feature map includes: Perform two 1×1 convolutions on the second layer feature map to obtain the first projection matrix and the second projection matrix respectively; Performing similarity matching on the first projection matrix and the second projection matrix to obtain a similarity matching matrix, and processing the similarity matching matrix through an activation function to obtain an attention weighted matrix; The third layer feature map is sequentially subjected to 1×1 convolution and upsampling to obtain an upsampled feature map, and the upsampled feature map is subjected to 1×1 convolution to obtain a projection matrix; Perform a matrix product operation on the projection matrix and the attention weighted matrix to obtain the first self-attention convolution feature map.

7. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the distribution network tower risk assessment method according to any one of claims 1 to 5 when executing the computer program.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the distribution network tower risk assessment method according to any one of claims 1 to 5 are implemented.

Images