CN115880231A - Power transmission line hidden danger detection method and system based on deep learning - Google Patents
Power transmission line hidden danger detection method and system based on deep learning Download PDFInfo
- Publication number
- CN115880231A CN115880231A CN202211461867.8A CN202211461867A CN115880231A CN 115880231 A CN115880231 A CN 115880231A CN 202211461867 A CN202211461867 A CN 202211461867A CN 115880231 A CN115880231 A CN 115880231A
- Authority
- CN
- China
- Prior art keywords
- hidden danger
- picture
- model
- transmission line
- power transmission
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Abstract
The invention relates to the technical field of power transmission lines, in particular to a method and a system for detecting hidden dangers of a power transmission line based on deep learning, wherein the detection method comprises the following steps: s1, converting a hidden danger picture into a model input picture, training the model input picture by using a deep learning yolov5 neural network, and deploying a yolov5 model at a detection terminal; s2, acquiring images around the power transmission line at regular time, inputting model input pictures into a yolov5 model, detecting whether hidden danger targets exist in the model input pictures or not through the yolov5 model, and classifying, grading and marking if hidden danger targets exist in the model input pictures; and S3, uploading the picture and the data of the category, score and coordinate of the hidden danger target in the picture to a detection background, rechecking the detection result if the hidden danger occurs, and informing personnel to handle the problem of the potential safety hazard if the detection result is true.
Description
Technical Field
The invention relates to the technical field of power transmission lines, in particular to a power transmission line hidden danger detection method and system based on deep learning.
Background
Along with the development of national economy, the demand of electric power is more and more big, and electric power is sent by the power station, and send to the place that needs electric power all over the country through the transmission line, because transmission line span is big, the circuit is long, and peripheral objects such as mountain fire, smog, tower crane, hoist, excavator have very big threat to transmission line's safety, so to the detection and the processing of these hidden troubles be very important to transmission line's protection. The image detection is widely applied to the field of target detection due to low cost, convenient installation and strong applicability, so the image detection is an important method for detecting hidden dangers of the power transmission line.
The accuracy, detection type and detection speed of image detection are key indexes of the quality detection. The image matching method and the monitoring system for the power transmission channel external damage protection, which are provided by the institute of electrical power science of electric power companies in Shandong province of China network, adopt the image matching method based on the OPENCV Haar characteristic density map to detect hidden dangers, have good detection effect in a specific scene, but have poor effect after the scene is changed. The method is based on a fast-RCNN two-stage object detection algorithm, and is low in detection speed and few in detection types. The 'transmission line peripheral smoke and fire detection method based on improved YOLOv 4' proposed by southern China university of science and engineering uses YOLOv4 to detect the smoke and fire around the transmission line, but the detection precision is low, the detection types are few, and the three methods have low detection precision, few detection types and slow speed.
The detection efficiency of the system is also one of the important factors for detecting the quality. "an unmanned aerial vehicle electric power inspection system and detection method based on intelligent vision" that western's ann oil university provided adopts unmanned aerial vehicle to shoot the discernment to the transmission line, and this kind of method needs regularly to control unmanned aerial vehicle and shoots the transmission line, and to more remote position, manual operation unmanned aerial vehicle has very big inconvenience, inefficiency moreover.
The cost and maintenance of the detection system are key factors in determining whether the system can be popularized. The picture science and technology limited company provides a self-learning identification system and a method based on the external hidden danger of the power transmission line, an edge computing platform is adopted to collect images of the surrounding environment of the power transmission line, the images and results are identified and returned to a cloud computer, the cloud computer synchronously carries out detection and training, and an updated model is returned to the edge computing platform.
The method detects hidden danger targets by using an image recognition technology, but some methods have low detection precision, few detection types and low speed, some methods need to operate equipment in a specific place, the efficiency is low, and some methods have high requirements on hardware configuration and high cost.
Disclosure of Invention
One of the purposes of the invention is to provide a power transmission line hidden danger detection method based on deep learning, which can quickly and accurately judge hidden danger targets of a power transmission line and can timely process the hidden danger targets, so that the defects in the prior art are avoided.
The invention also aims to provide a power transmission line hidden danger detection and early warning system.
In order to achieve one of the above purposes, the invention provides the following technical scheme:
the method for detecting the hidden danger of the power transmission line based on deep learning comprises the following steps:
s1, shooting a hidden danger picture of the power transmission line, converting the hidden danger picture into a model input picture, training the model input picture by using a deep learning model yolov5 neural network to obtain a yolov5 model for detecting a hidden danger target of the power transmission line, and deploying the yolov5 model at a detection terminal;
s2, acquiring images around the power transmission line at regular time, converting the images into model input pictures, inputting the model input pictures into a yolov5 model, detecting whether the hidden danger targets exist in the model input pictures through the yolov5 model, and classifying, grading and marking the hidden danger targets if the hidden danger targets exist in the model input pictures to obtain the pictures and the types, the scores and the coordinates of the hidden danger targets in the pictures;
and S3, uploading the picture and the data of the category, the score and the coordinate of the hidden danger target in the picture to a detection background, outputting a detection result by the detection background if the hidden danger occurs, rechecking the detection result, and informing nearby personnel of a hidden danger place to handle the potential safety hazard problem if the situation is true.
In some embodiments, the step of converting the picture or image into a model input picture comprises:
and under the condition that the size scale of the original image is kept unchanged and the long edge is zoomed to 640, zooming the original image, pasting the zoomed image to the middle of a 640 multiplied by 640 preset image, and obtaining a model input image.
In some embodiments, the step of training the deep learning model yolov5 neural network to model input pictures comprises:
marking the hidden danger target of each picture to generate a corresponding xml file, wherein the marked information comprises the width, the height and the name of the picture, and the coordinates of the upper left corner and the lower right corner of a target circumscribed rectangle, the coordinates of the upper left corner are (xmin, ymin), and the coordinates of the lower right corner are (xmax, ymax);
building a deep learning model yolov5 neural network model based on tensoflow-gpu by using python, setting the size of an input image of the yolov5 neural network model to be 640 multiplied by 640, and setting an output image to be three feature layers with different sizes;
and expanding an original data set of the model input picture by a data enhancement method, then inputting the picture and the labeling information in the xml file into a yolov5 network, and training to obtain the yolov5 model.
In some embodiments, the data enhancement method comprises scaling, flipping, stitching, cropping, or color space transformation.
In some embodiments, the trained yolov5 model is deployed to the detection terminal using C + + language.
In some embodiments, in step S2, the step of detecting whether a hidden danger target exists through the yolov5 model and classifying, scoring and marking the hidden danger target includes:
s31, inputting the model input picture into a yolov5 model to obtain 3 4-dimensional output characteristic layers:
output0：(1，20，20，42)
output1：(1，40，40，42)
output2：(1，80，80，42)，
in the array, the first dimension is the number 1 of pictures in the group, and the second dimension and the third dimension are preset frames anchors of each group; the fourth dimension is the information predicted by the anchor:
(t x ，t y ，t w ，t h ，conf，class_prob)，
wherein:
t x ，t y is the offset of the center coordinate of the prediction box from the current anchor
t w ，t h Is the offset of the predicted frame width height relative to the current anchor width height
conf is the probability that the current anchor is responsible for prediction
class _ prob is the score for each class of detection target;
each anchor predicts a 14-dimensional vector, each anchor has 3 anchors, and each anchor outputs a 42-dimensional vector;
s32, setting a confidence coefficient threshold conf _ threshold, and screening out detection results of which all thresholds conf are greater than the conf _ threshold;
s33, converting the coordinate offset of the prediction frame into the relative coordinate of the feature layer of the picture:
in the formula (I), the compound is shown in the specification,
t x ，t y ，t w ，t h predicted box center coordinates and width-to-height offsets, c, for each anchor prediction x ，c y ，p w ，p h For the coordinates of the grid of the characteristic layer where each anchor is located and the preset width and height of the anchor, b x ，b y ，b w ，b h The central coordinate and the width and the height of the prediction frame relative to the current characteristic layer are obtained;
s34, converting the relative coordinates of the height and the width of the center of the characteristic layer of the prediction frame into the relative coordinates of the height and the width of the center of the original picture
In the formula (I), the compound is shown in the specification,
nw, nh is the width and height, grid of the original image after data enhancement w ，grid h Is the number of meshes of the current feature layer, r x ，r y ，r w ，r h Relative coordinates of the prediction frame relative to the center width and the height of the original picture are obtained;
s35, converting the relative coordinates of the center width and the height of the original picture of the prediction frame into real coordinates of the upper left corner and the lower right corner
In the formula (I), the compound is shown in the specification,
w, h is the width and height of the original picture, T xmin ，T ymin ，T xmax ，T ymax True coordinates of the upper left corner and the lower right corner of the prediction frame;
s36, finding out a category maximum value for the score class _ prob of each prediction frame, representing that the probability of the prediction frame belonging to which category is the maximum, classifying the prediction frame into the category, then sorting the scores of all the prediction frames of each category, wherein the higher the score in the front is, the higher the probability of belonging to the category is, finally setting an nms _ threshold non-maximum value inhibition threshold, and removing the prediction frames with higher coincidence degree in the same category from all the prediction frames sorted in the same category according to the condition that the intersection ratio iou is smaller than the nms _ threshold;
and S37, coordinates of the upper left corner and the lower right corner of the prediction frame, the category of the prediction frame and the score of the prediction frame are obtained, and then the information is marked on the original image to obtain the picture and the category, the score and the coordinates of the hidden danger target in the picture.
In some embodiments, in step S3, an alarm is issued according to whether the manual review condition is true or not and after the manual review condition is true.
The power transmission line hidden danger detection method based on deep learning has the beneficial effects that:
(1) According to the method for detecting the hidden danger of the power transmission line based on the deep learning, the yolov5 model capable of identifying the hidden danger target is trained, the hidden danger is identified at the terminal through the rapid single-step target yolov5 model, compared with the traditional hidden danger identification target, the yolov5 deep learning model is multiple in identification types, high in positioning precision, accurate in classification, high in score, low in error identification rate and high in detection speed, and under the condition that the changes of complex scenes, weather, illumination, installation geographic positions and the like are large, the good effect can be achieved, the identification efficiency of the hidden danger target is improved, the problem of scene influence is solved, and the intelligence degree is high.
(2) According to the method for detecting the hidden danger of the power transmission line based on the deep learning, if the hidden danger target is identified, the alarm information is automatically reported, and after the alarm information is manually rechecked and confirmed, nearby workers are informed to handle the hidden danger, so that the safety of the power transmission line is effectively protected, and the workload of operation and maintenance personnel is reduced.
In order to achieve the second purpose, the invention provides the following technical scheme:
the power transmission line hidden danger detection early warning system comprises a storage medium, wherein program codes are stored in the storage medium, and when the program codes are executed by a processor, the power transmission line hidden danger detection method based on deep learning is realized.
Drawings
Fig. 1 is a schematic flow chart of a power transmission line hidden danger detection method based on deep learning according to an embodiment.
Detailed Description
Preferred embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. While the preferred embodiments of the present invention are shown in the drawings, it should be understood that the present invention may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in this disclosure and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.
It is to be understood that, although the terms first, second, third, etc. may be used herein to describe various information, such information should not be limited by these terms. These terms are only used to distinguish one type of information from another. For example, first information may also be referred to as second information, and similarly, second information may also be referred to as first information, without departing from the scope of the present invention. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the present invention, "a plurality" means two or more unless specifically defined otherwise.
Example 1
The hidden danger targets are detected by using an image recognition technology at present, but some detection methods are low in detection precision, few in detection types and low in speed, some detection methods need to operate equipment at a specific place, the efficiency is low, and some detection methods have high requirements on hardware configuration and high cost.
In view of the above problems, the method for detecting hidden danger of power transmission line based on deep learning disclosed in this embodiment, as shown in fig. 1, includes the following steps:
s1, shooting a hidden danger picture of the power transmission line, wherein a camera is also installed on a detection terminal for shooting, the hidden danger picture is converted into a model input picture, the picture is input by using a deep learning model yolov5 neural network training model to obtain a yolov5 model for detecting a hidden danger target of the power transmission line, and the yolov5 model is deployed on the detection terminal;
s2, acquiring images around the power transmission line at regular time, converting the images into model input pictures, inputting the model input pictures into a yolov5 model, detecting whether the hidden danger targets exist in the model input pictures through the yolov5 model, and classifying, grading and marking the hidden danger targets if the hidden danger targets exist in the model input pictures to obtain the pictures and the types, the scores and the coordinates of the hidden danger targets in the pictures;
and S3, uploading the picture and the data of the type, the score and the coordinates of the hidden danger target in the picture to a detection background, outputting a detection result by the detection background if the hidden danger exists, rechecking the detection result, and informing nearby personnel of a hidden danger site to handle the potential safety hazard problem if the situation is true.
According to the method for detecting the hidden danger of the power transmission line based on deep learning, the yolov5 model capable of identifying the hidden danger target is trained firstly, the hidden danger is identified at the terminal through the quick single-step target yolov5 model, compared with the traditional hidden danger target identification method, the yolov5 deep learning model is multiple in identification types, high in positioning accuracy, accurate in classification, high in score, low in false identification rate and high in detection speed, and under the condition that the changes of complex scenes, weather, illumination, installation geographic positions and the like are large, the good effect can be achieved, the identification efficiency of the hidden danger target is improved, the problem of scene influence is solved, and the intelligent degree is high. If the hidden danger target is identified, the alarm information is automatically reported, and after the alarm information is confirmed through manual rechecking, nearby workers are informed to handle the hidden danger, so that the safety of the power transmission line is effectively protected, and the workload of operation and maintenance personnel is reduced.
In this embodiment, the step of converting the picture or the image into the model input picture includes:
and under the condition that the size scale of the original image is kept unchanged and the long edge is zoomed to 640, zooming the original image, pasting the zoomed image to the middle of a 640 multiplied by 640 preset image, and obtaining a model input image.
In this embodiment, the step of training the model input picture by the deep learning model yolov5 neural network includes:
marking the hidden danger target of each picture to generate a corresponding xml file, wherein the marked information comprises the width, the height and the name of the picture, and the coordinates of the upper left corner and the lower right corner of a target external rectangle, the coordinates of the upper left corner are (xmin, ymin), and the coordinates of the lower right corner are (xmax, ymax);
building a deep learning model yolov5 neural network model based on tensierflow-gpu by using python, setting the size of an input image of the yolov5 neural network model to be 640 multiplied by 640, and setting an output image to be three characteristic layers with different sizes;
and expanding an original data set of the model input picture by a data enhancement method, then inputting the picture and the labeling information in the xml file into a yolov5 network, and training to obtain the yolov5 model.
In this embodiment, the data enhancement method includes scaling, flipping, splicing, clipping, or color space transformation.
In this embodiment, the trained yolov5 model is deployed to the detection terminal by using C + + language.
In this embodiment, in step S2, the step of detecting whether a hidden danger target exists through the yolov5 model, and classifying, scoring, and marking the hidden danger target includes:
s31, inputting the model input picture into a yolov5 model to obtain 3 4-dimensional output characteristic layers:
output0：(1，20，20，42)
output1：(1，40，40，42)
output2：(1，80，80，42)，
in the above array, since the number of detected pictures is 1, the number of pictures in the first dimension is 1;
the second dimension and the third dimension are preset frames anchors of each group, the three preset frames anchors with different sizes are 20, 40 and 80, targets with different sizes can be detected, the middle dimension of output0 is 20, which represents a large anchor, the middle dimension of output1 is 40, which represents a middle anchor, and the middle dimension of output2 is 80, which represents a small anchor;
the fourth dimension is information predicted by the anchors, specifically, each anchor has 3 anchors with different length-width ratios, targets with different length-width ratios can be predicted, and each anchor predicts 14-dimensional information:
(t x ，t y ，t w ，t h ，conf，class_prob)，
wherein:
t x ，t y is the offset of the center coordinate of the prediction box from the current anchor
t w ，t h Is the offset of the predicted frame width height relative to the current anchor width height
conf is the probability that the current anchor is responsible for prediction
class _ prob is the score of each category of the detection target, and the number of the detected categories is 9 in this embodiment;
each anchor predicts a 14-dimensional vector, each anchor has 3 anchors, each anchor outputs a 42-dimensional vector;
s32, setting confidence coefficient threshold values conf _ threshold, and screening out detection results of which all threshold values conf are greater than the conf _ threshold;
s33, converting the coordinate offset of the prediction frame into the relative coordinate of the feature layer of the picture:
in the formula (I), the compound is shown in the specification,
t x ，t y ，t w ，t h predicted box center coordinates and width-to-height offsets, c, for each anchor prediction x ，c y ，p w ，p h For the coordinates of the grid of the characteristic layer where each anchor is located and the preset width and height of the anchor, b x ，b y ，b w ，b h The central coordinate and the width and the height of the prediction frame relative to the current feature layer are calculated;
s34, converting the relative coordinates of the height and the width of the center of the characteristic layer of the prediction frame into the relative coordinates of the height and the width of the center of the original picture
In the formula (I), the compound is shown in the specification,
nw, nh is the width and height grid of the original image after data enhancement w ，grid h Is the number of meshes of the current feature layer, r x ，r y ，r w ，r h Relative coordinates of the prediction frame relative to the width and height of the center of the original picture;
s35, converting the relative coordinates of the center width and the height of the original picture of the prediction frame into real coordinates of the upper left corner and the lower right corner
In the formula (I), the compound is shown in the specification,
w, h is the width and height of the original picture, T xmin ，T ymin ，T xmax ，T ymax True coordinates of the upper left corner and the lower right corner of the prediction frame;
s36, finding out a category maximum value for the score class _ prob of each prediction frame, representing that the probability of the prediction frame belonging to which category is the maximum, classifying the prediction frame into the category, then sorting the scores of all the prediction frames of each category, wherein the higher the score in the front is, the higher the probability of belonging to the category is, finally setting an nms _ threshold non-maximum value inhibition threshold, and removing the prediction frames with higher coincidence degree in the same category from all the prediction frames sorted in the same category according to the condition that the intersection ratio iou is smaller than the nms _ threshold;
and S37, coordinates of the upper left corner and the lower right corner of the prediction frame, the category of the prediction frame and the score of the prediction frame are obtained, and then the information is marked on the original image to obtain the picture and the category, the score and the coordinates of the hidden danger target in the picture.
In this embodiment, in step S3, an alarm is issued according to whether the manual rechecking condition is true or not and after the manual rechecking condition is true.
In this embodiment, in step S3, a nearby person is notified to travel to a target location through an intranet.
In this embodiment, the hidden danger picture includes mountain fire, smoke, or engineering machinery.
Example 2
The power transmission line hidden danger detection early warning system disclosed in the embodiment comprises a storage medium, wherein a program code is stored in the storage medium, and when the program code is executed by a processor, the power transmission line hidden danger detection method based on deep learning in embodiment 1 is realized.
The system has low equipment cost and low maintenance cost, can be remotely upgraded and can greatly save the cost of enterprises.
The relative arrangement of the components and steps, the numerical expressions, and numerical values set forth in these embodiments do not limit the scope of the present application unless specifically stated otherwise. Meanwhile, it should be understood that the sizes of the respective portions shown in the drawings are not drawn in an actual proportional relationship for the convenience of description. Techniques, methods, and apparatus known to one of ordinary skill in the relevant art may not be discussed in detail but are intended to be part of the specification where appropriate. In all examples shown and discussed herein, any particular value should be construed as exemplary only and not as limiting. Thus, other examples of the exemplary embodiments may have different values. It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, further discussion thereof is not required in subsequent figures.
In the description of the present application, it is to be understood that the orientation or positional relationship indicated by the directional terms such as "front, rear, upper, lower, left, right", "lateral, vertical, horizontal" and "top, bottom", etc., are generally based on the orientation or positional relationship shown in the drawings, and are used for convenience of description and simplicity of description only, and in the case of not making a reverse description, these directional terms do not indicate and imply that the device or element being referred to must have a particular orientation or be constructed and operated in a particular orientation, and therefore, should not be considered as limiting the scope of the present application; the terms "inner and outer" refer to the inner and outer relative to the profile of the respective component itself.
For ease of description, spatially relative terms such as "over … …", "over … …", "over … …", "over", etc. may be used herein to describe the spatial positional relationship of one device or feature to another device or feature as shown in the figures. It will be understood that the spatially relative terms are intended to encompass different orientations of the device in use or operation in addition to the orientation depicted in the figures. For example, if a device in the figures is turned over, devices described as "above" or "on" other devices or configurations would then be oriented "below" or "under" the other devices or configurations. Thus, the exemplary term "above … …" may include both orientations of "above … …" and "below … …". The device may be otherwise variously oriented (rotated 90 degrees or at other orientations) and the spatially relative descriptors used herein interpreted accordingly.
It should be noted that the terms "first", "second", and the like are used to define the components, and are only used for convenience of distinguishing the corresponding components, and the terms have no special meanings unless otherwise stated, and therefore, the scope of protection of the present application is not to be construed as being limited.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A method for detecting hidden danger of a power transmission line based on deep learning is characterized by comprising the following steps: the method comprises the following steps:
s1, shooting a hidden danger picture of the power transmission line, converting the hidden danger picture into a model input picture, training the model input picture by using a deep learning model yolov5 neural network to obtain a yolov5 model for detecting a hidden danger target of the power transmission line, and deploying the yolov5 model at a detection terminal;
s2, regularly acquiring images around the power transmission line, converting the images into model input pictures, inputting the model input pictures into a yolov5 model, detecting whether hidden danger targets exist in the model input pictures through the yolov5 model, and if yes, classifying, grading and marking the hidden danger targets to obtain the pictures and the types, the scores and the coordinates of the hidden danger targets in the pictures;
and S3, uploading the picture and the data of the category, the score and the coordinate of the hidden danger target in the picture to a detection background, outputting a detection result by the detection background if the hidden danger occurs, rechecking the detection result, and informing nearby personnel of a hidden danger place to handle the potential safety hazard problem if the situation is true.
2. The deep learning-based power transmission line hidden danger detection method according to claim 1, characterized in that: the step of converting the picture or image into a model input picture comprises:
and under the condition that the size scale of the original image is kept unchanged and the long edge is zoomed to 640, zooming the original image, pasting the zoomed image to the middle of a 640 multiplied by 640 preset image, and obtaining a model input image.
3. The deep learning-based power transmission line hidden danger detection method according to claim 1, characterized in that: the method for training the model input picture by the deep learning model yolov5 neural network comprises the following steps:
marking the hidden danger target of each picture to generate a corresponding xml file, wherein the marked information comprises the width, the height and the name of the picture, and the coordinates of the upper left corner and the lower right corner of a target external rectangle, the coordinates of the upper left corner are (xmin, ymin), and the coordinates of the lower right corner are (xmax, ymax);
building a deep learning model yolov5 neural network model based on tensoflow-gpu by using python, setting the size of an input image of the yolov5 neural network model to be 640 multiplied by 640, and setting an output image to be three feature layers with different sizes;
and expanding the original data set of the model input picture by a data enhancement method, then inputting the picture and the labeling information in the xml file into a yolov5 network, and training to obtain the yolov5 model.
4. The deep learning-based power transmission line hidden danger detection method according to claim 3, characterized in that: the data enhancement method comprises scaling, turning, splicing, clipping or color space transformation.
5. The deep learning-based power transmission line hidden danger detection method according to claim 1, characterized in that: and deploying the trained yolov5 model to a detection terminal by adopting C + + language.
6. The deep learning-based power transmission line hidden danger detection method according to claim 1, characterized in that: in step S2, the step of detecting whether a hidden danger target exists through the yolov5 model and classifying, scoring and marking the hidden danger target includes:
s31, inputting the model input picture into a yolov5 model to obtain 3 4-dimensional output characteristic layers:
output0：(1，20，20，42)
output1：(1，40，40，42)
output2：(1，80，80，42)，
in the array, the first dimension is the number 1 of pictures in the group, and the second dimension and the third dimension are preset frames anchors of each group; the fourth dimension is information predicted by the anchor:
(t x ，t y ，t w ，t h ，conf，class_prob)，
wherein:
t x ，t y is the offset of the center coordinate of the prediction box from the current anchor
t w ，t h Is the offset of the predicted frame width and height relative to the current anchor width and height
conf is the probability that the current anchor is responsible for prediction
class _ prob is each category score of the detection target;
each anchor predicts a 14-dimensional vector, each anchor has 3 anchors, each anchor outputs a 42-dimensional vector;
s32, setting confidence coefficient threshold values conf _ threshold, and screening out detection results of which all threshold values conf are greater than the conf _ threshold;
s33, converting the coordinate offset of the prediction frame into the relative coordinate of the feature layer of the picture:
in the formula (I), the compound is shown in the specification,
t x ，t y ，t w ，t h predicted box center coordinates and width-to-height offsets, c, for each anchor prediction x ，c y ，p w ，p h For the coordinates of the grid of the characteristic layer where each anchor is located and the preset width and height of the anchor, b x ，b y ，b w ，b h The central coordinate and the width and the height of the prediction frame relative to the current characteristic layer are obtained;
s34, converting the relative coordinates of the height and the width of the center of the characteristic layer of the prediction frame into the relative coordinates of the height and the width of the center of the original picture
In the formula (I), the compound is shown in the specification,
nw, nh is the width and height grid of the original image after data enhancement w ，grid h Is the number of meshes of the current feature layer, r x ，r y ，r w ，r h Relative coordinates of the prediction frame relative to the width and height of the center of the original picture;
s35, converting the relative coordinates of the center width and the height of the original picture of the prediction frame into real coordinates of the upper left corner and the lower right corner
In the formula (I), the compound is shown in the specification,
w, h is the width and height of the original picture, T amin ，T ymin ，T xmax ，T ymax True coordinates of the upper left corner and the lower right corner of the prediction frame;
s36, finding out a category maximum value for the score class _ prob of each prediction frame, wherein the category maximum value indicates the category to which the prediction frame belongs, classifying the prediction frame into the category, then sorting the scores of all the prediction frames of each category, the higher the score arranged in the front is, the higher the possibility of belonging to the category is, finally setting an nms _ threshold non-maximum value inhibition threshold, and removing the prediction frames with higher coincidence degree of the same category from all the prediction frames sorted by the same category under the condition that the coincidence ratio iou is less than nmS _ threshold;
and S37, coordinates of the upper left corner and the lower right corner of the prediction frame, the category of the prediction frame and the score of the prediction frame are obtained, and then the information is marked on the original image to obtain the picture and the category, the score and the coordinates of the hidden danger target in the picture.
7. The deep learning-based power transmission line hidden danger detection method according to claim 1, characterized in that: and step S3, sending an alarm according to whether the manual rechecking condition is true or not and after the manual rechecking condition is true.
8. The deep learning-based power transmission line hidden danger detection method according to claim 1, characterized in that: in step S3, nearby persons are notified to travel to the target location through the intranet.
9. The deep learning-based power transmission line hidden danger detection method according to claim 1, characterized in that: the hidden danger pictures comprise mountain fire, smog or engineering machinery.
10. The utility model provides a transmission line hidden danger detects early warning system which characterized by: the storage medium is stored with program codes, and when the program codes are executed by a processor, the method for detecting the hidden danger of the power transmission line based on the deep learning as claimed in any one of claims 1 to 9 is realized.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211461867.8A CN115880231A (en) | 2022-11-17 | 2022-11-17 | Power transmission line hidden danger detection method and system based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211461867.8A CN115880231A (en) | 2022-11-17 | 2022-11-17 | Power transmission line hidden danger detection method and system based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115880231A true CN115880231A (en) | 2023-03-31 |
Family
ID=85760431
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211461867.8A Pending CN115880231A (en) | 2022-11-17 | 2022-11-17 | Power transmission line hidden danger detection method and system based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115880231A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116152863A (en) * | 2023-04-19 | 2023-05-23 | 尚特杰电力科技有限公司 | Personnel information identification method and device, electronic equipment and storage medium |
| CN117081237A (en) * | 2023-07-19 | 2023-11-17 | 珠海市深瑞智联科技有限公司 | Method and system for identifying hidden danger of power transmission line based on radar scanning and image acquisition |
-
2022
- 2022-11-17 CN CN202211461867.8A patent/CN115880231A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116152863A (en) * | 2023-04-19 | 2023-05-23 | 尚特杰电力科技有限公司 | Personnel information identification method and device, electronic equipment and storage medium |
| CN117081237A (en) * | 2023-07-19 | 2023-11-17 | 珠海市深瑞智联科技有限公司 | Method and system for identifying hidden danger of power transmission line based on radar scanning and image acquisition |
| CN117081237B (en) * | 2023-07-19 | 2024-04-02 | 珠海市深瑞智联科技有限公司 | Method and system for identifying hidden danger of power transmission line based on radar scanning and image acquisition |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20210073573A1 (en) | Ship identity recognition method based on fusion of ais data and video data | |
| CN111444809B (en) | Power transmission line abnormal target detection method based on improved YOLOv3 | |
| CN110674746B (en) | Method and device for realizing high-precision cross-mirror tracking by using video spatial relationship assistance, computer equipment and storage medium | |
| CN115880231A (en) | Power transmission line hidden danger detection method and system based on deep learning | |
| CN108009515B (en) | Power transmission line positioning and identifying method of unmanned aerial vehicle aerial image based on FCN | |
| CN113052876B (en) | Video relay tracking method and system based on deep learning | |
| CN108648169A (en) | The method and device of high voltage power transmission tower defects of insulator automatic identification | |
| CN112101088B (en) | Unmanned aerial vehicle electric power automatic inspection method, device and system | |
| CN113903081A (en) | Visual identification artificial intelligence alarm method and device for images of hydraulic power plant | |
| CN107590456A (en) | Small micro- mesh object detection method in a kind of high-altitude video monitoring | |
| CN113808282A (en) | Multi-navigation-factor data fusion method | |
| CN110703760B (en) | Newly-added suspicious object detection method for security inspection robot | |
| CN109828267A (en) | The Intelligent Mobile Robot detection of obstacles and distance measuring method of Case-based Reasoning segmentation and depth camera | |
| CN112183472A (en) | Method for detecting whether test field personnel wear work clothes or not based on improved RetinaNet | |
| CN114240868A (en) | Unmanned aerial vehicle-based inspection analysis system and method | |
| CN109977840A (en) | A kind of airport scene monitoring method based on deep learning | |
| CN115809986A (en) | Multi-sensor fusion type intelligent external damage detection method for power transmission corridor | |
| CN110852179A (en) | Method for detecting suspicious personnel intrusion based on video monitoring platform | |
| CN113822217A (en) | Ship tail gas monitoring method based on AIS and video image analysis | |
| CN116846059A (en) | Edge detection system for power grid inspection and monitoring | |
| CN109684982B (en) | Flame detection method based on video analysis and combined with miscible target elimination | |
| CN114812403A (en) | Large-span steel structure hoisting deformation monitoring method based on unmanned aerial vehicle and machine vision | |
| CN114998846A (en) | Method and system for realizing intelligent system of offshore drilling platform and related components | |
| CN113076899B (en) | High-voltage transmission line foreign matter detection method based on target tracking algorithm | |
| CN112926426A (en) | Ship identification method, system, equipment and storage medium based on monitoring video |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |