CN115809986A - Multi-sensor fusion type intelligent external damage detection method for power transmission corridor - Google Patents
Multi-sensor fusion type intelligent external damage detection method for power transmission corridor Download PDFInfo
- Publication number
- CN115809986A CN115809986A CN202211341101.6A CN202211341101A CN115809986A CN 115809986 A CN115809986 A CN 115809986A CN 202211341101 A CN202211341101 A CN 202211341101A CN 115809986 A CN115809986 A CN 115809986A
- Authority
- CN
- China
- Prior art keywords
- power transmission
- model
- transmission corridor
- image
- camera
- 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
Landscapes
- Image Processing (AREA)
Abstract
The invention relates to a multi-sensor fusion intelligent outer broken detection method for a power transmission corridor, which is based on a YOLOv3 algorithm, utilizes a deep neural network model compression technology, greatly compresses the model volume, greatly reduces the model parameter quantity and the calculated quantity on the premise of reducing the precision loss as much as possible, accelerates the model reasoning speed, ensures the detection real-time performance, simultaneously fuses a millimeter wave radar sensor and a camera radar sensor, accurately obtains the depth distance of a target in an image, performs coordinate conversion according to the depth distance, further calculates the distance between the image target and a power line, and uses the distance as an alarm judgment condition. The invention can effectively reduce the false alarm rate on the premise of ensuring the detection precision, realize the real-time monitoring of the external damage of the power transmission corridor, improve the online video monitoring quality of the power transmission corridor, ensure the safety of the power transmission corridor and improve the operation safety and stability of power equipment.
Description
Technical Field
The invention relates to the technical field of power transmission corridor external damage monitoring, in particular to a multi-sensor fusion power transmission corridor intelligent external damage detection method.
Background
Because the power transmission corridor which is one of the important components of the power system has the characteristics of complex structure, huge scale, environment and complexity, peripheral foreign matters, trees and illegal buildings have high possibility of invading into the safe area of the power transmission corridor to cause short-circuit accidents. When a power transmission corridor serving as a main channel for bearing power transmission fails, equipment on a line is easily damaged, tripped and shut down, stable operation of a power grid is seriously influenced, normal production and living power supply is interfered, and production and living operation of the society is disturbed.
The existing transmission line external damage prevention technology starts from multiple angles, potential external damage threats are found, certain positive effects are achieved in specific time periods and application occasions, and the existing transmission corridor external damage monitoring method mainly comprises manual line patrol, airborne laser measurement and video online monitoring:
(1) In the line patrol process, power workers often judge whether the problem of outburst invasion exists in a patrol section through experience or measurement by using a total station, however, workers are usually difficult to reach places with multiple accidents due to complex line patrol environment, and meanwhile, the workers are difficult to effectively and accurately judge whether the safety distance between the outburst and a power transmission corridor reaches the standard due to interference of trees and buildings and the influence of visual deviation. With the continuous progress and development of the power grid, the disadvantages of the method are more prominent, and the existing high-efficiency, advanced and scientific power transmission line external damage monitoring method is a problem to be solved urgently for the power grid.
(2) At present, the safe distance of the power transmission line is identified by using a laser radar mainly by an airborne laser measurement technology. The method comprises the steps of firstly scanning a power transmission corridor by an unmanned aerial vehicle loaded with a laser radar, obtaining more accurate point cloud data of the environment of the power transmission corridor, analyzing the point cloud data, extracting the point cloud data of a power transmission line and a surface object, and calculating the Euclidean distance between the point cloud data and the point cloud data, so that the monitoring of the outer damage of the power transmission corridor is completed. However, the method has high precision and high cost, needs unmanned aerial vehicle inspection, cannot perform real-time inspection, and is not beneficial to real-time inspection of potential safety hazards. In addition, the laser point cloud data processing has higher difficulty. In the power transmission corridor external damage monitoring scene, the airborne laser radar method still has a certain distance from large-scale use.
(3) Because the channel environment is complex and changeable, the hidden dangers threatening the safe operation of the line are various in types, fast in change and complex in correlation, and the requirement of high-precision detection is difficult to meet simply depending on simple target detection at present. The video online monitoring of the power transmission line mainly has the following problems at present: the method has the advantages that the distance cannot be accurately measured, the method is limited by two-dimensional images, various angles and the like, the distance between an outer break and a power line is measured by the current target detection algorithm, difficulty is brought to the discrimination of hidden dangers, and the missing report rate and the false report rate of the identification algorithm are seriously influenced.
Disclosure of Invention
In order to overcome the defects in the prior art, the invention provides a multi-sensor fusion intelligent external damage detection method for a power transmission corridor. The outer broken real-time monitoring of transmission corridor is realized, the online video monitoring quality of transmission corridor is improved, and the safety of transmission corridor is guaranteed.
A multi-sensor fused intelligent external damage detection method for a power transmission corridor comprises the following steps:
firstly, calibrating a camera by adopting a Zhang Zhengyou method, collecting dozens of images with different shooting visual angles by using a calibration plate with black and white checkerboard grids according to the calibration plate, extracting angular points of the image checkerboard, analyzing internal and external parameters and distortion coefficients of the camera, and correcting the distortion of the camera by using the parameters;
secondly, calibrating the millimeter wave radar in a combined mode, placing a camera and the millimeter wave radar in a posture that an optical axis is parallel to the ground, placing a metal plate at different distance angles, recording the distance and the angle returned by the millimeter wave radar, and simultaneously recording a camera shooting image; aiming at the positions of radar data and target data on an image, a millimeter wave radar coordinate system is converted into a world coordinate system, the world coordinate system is converted into a pixel coordinate system by calibrating parameters obtained by a camera in the first step, and the conversion relation among the five coordinate systems of the millimeter wave radar, a three-dimensional world, the camera, the image and the pixel is accurately obtained, so that target point information returned by the millimeter wave radar can be accurately projected to the correct position of the image, and the fusion of the millimeter wave radar and the camera is realized;
thirdly, collecting and processing power transmission corridor video image data, uniformly cutting and storing the collected power transmission corridor video data into image data through manpower, taking the safety standard of the power transmission corridor as a basis, taking an external force damage object mainly related to the standard as a corresponding label, matching and labeling the label and an object in a data set, labeling the object to be detected with a rectangular frame, and manufacturing into a data set required by model training;
fourthly, constructing a model, constructing a YOLOv3 network by using a Keras framework, extracting a main structure of the network by taking Darknet53 as a feature, connecting a feature pyramid FPN module, realizing the extraction of shape features, texture features and edge information of the image, inputting the extracted feature information into a deep neural network, calculating model parameters by the deep neural network, and outputting a power transmission corridor external damage identification result based on the image;
fifthly, pre-training a model, inputting the image data with labels into a deep neural network for pre-training by adopting a data enhancement and balanced sampling method comprising Gaussian blur, random illumination and horizontal inversion, iteratively optimizing model parameters through each round of prediction results, performing grouped sampling on samples with unbalanced classes to balance the number of the classes, continuously training the model by utilizing a balanced sampling mode, and performing auxiliary calculation through a high-performance calculation accelerator card until the model parameters are converged to obtain an optimal model;
sixthly, performing sparse training, namely endowing sparse factors to batch normalization layers in the deep neural network, manually designing a sparse rate, and then performing sparse training on an original data set to obtain deep neural network models with different batch normalization layer weight values;
seventhly, channel pruning, namely sequencing the maximum values of the weights of all the batches of normalization layers according to the size of the sparse factor, taking the minimum value to calculate pruning rate, and adaptively searching the optimal pruning rate within the range of 10% above and below according to the identification effect of the verification set;
step eight, fine tuning training, namely retraining the pruned model to recover the accuracy of the model prediction result and obtain an optimal compression model;
ninthly, detecting the power line, namely extracting the edge of the linear power line by utilizing Hough change, extracting the edge of the power line and realizing efficient and quick power line detection while detecting a target;
step ten, converting a target point coordinate system, and converting an image coordinate identified by a YOLOv3 algorithm and a power line image coordinate into a three-dimensional coordinate under a world coordinate system according to the camera internal and external parameters obtained in the step one and the pixel depth value measured by the millimeter wave radar;
step eleven, calculating a target point safe distance, calculating Euclidean distances between an outer broken coordinate identified by a YOLOv3 algorithm under a world coordinate system and all power line coordinates, and using the minimum value as a safe distance for subsequent alarm judgment;
and a twelfth step of alarming the outer break, wherein the shortest distance between the outer break and the power line identified by the YOLOv3 algorithm calculated in the eleventh step is judged, whether the outer break can generate a security threat to the power transmission corridor or not is judged, and an outer break identification result smaller than the preset alarm distance is alarmed.
Preferably, the detection method is composed of an image acquisition and identification module and a millimeter wave radar ranging module.
Preferably, the fourth step of constructing a YOLOv3 network model, performing downsampling through a continuous convolution module to obtain feature maps of different scales, performing cross-layer connection, stacking features between different layers to obtain feature maps of different granularities, effectively obtaining shape features, texture features and edge information in an original picture, extracting a main structure by taking Darknet53 as a feature, connecting a feature pyramid FPN module, integrating the features of different scales into the same size through convolution operation, stacking and combining the features into a whole through a channel, inputting the whole into a prediction module of Yolo-Head, and performing position and category prediction on possible targets through the convolution operation.
Preferably, the fifth step of model training performs grouping sampling on samples with unbalanced categories, and adopts a data enhancement and balanced sampling method to balance the number of categories, wherein the enhancement method comprises gaussian blurring, random illumination and horizontal inversion.
Preferably, the sixth step of sparse training is to train the weight of the batch normalization layer by using an introduced manually designed sparse factor on the premise that the pre-training model has obtained the optimal model parameters, and the randomly initialized sparse factor reduces the precision loss of the sparse model on the verification set and makes part of the weight values close to 0 through iterative optimization, so as to obtain the convolution channel index which has less influence on the model prediction result.
Preferably, the channel pruning in the seventh step is to perform a channel clipping test on the convolution channel index obtained in the fifth step according to the ratio by using a step length set by a person according to the pruning ratio range obtained by calculation, verify the pruning effect on a verification set, and screen out the optimal pruning ratio and model.
Preferably, the eighth step of fine tuning training is to retrain the pruned model, and at this time, the learning rate needs to be adjusted according to the pruning rate, and the accuracy of the model prediction result is restored by repeatedly trying.
Preferably, the ninth step of power line detection adopts Hough change to extract edges of linear power lines, converts pixel points into curves of parameter space, and when pixel points on an image form a straight line, the curve corresponding to the parameter space is intersected at a point corresponding to a pixel point on the image forming a straight line, and edge features of the power lines divided into approximate straight lines are found by finding intersection points of the parameter space.
Preferably, four coordinate systems are involved: a world coordinate system describing the real position of the camera; a camera coordinate system with an origin as an optical center; an image coordinate system, wherein the origin is the midpoint of an imaging plane; and a pixel coordinate system with an origin at the upper left corner of the image, wherein a formula for converting a point from the pixel coordinate system to a world coordinate system can be obtained through the conversion of the above four coordinate systems:
in the formula, Z c The coordinate value of the camera coordinate system can be obtained by millimeter wave radar ranging, wherein f is the focal length of the camera, f x And f y Is the length and width of a single pixel (u) 0 ,v 0 ) Coordinates of the central point of the imaging plane belong to camera internal parameters; and R is a rotation matrix, and T is an offset matrix, and the first step is used for obtaining.
Preferably, the eleventh step of calculating the safe distance between the target points and calculating yolov3 algorithm identifies the three-dimensional coordinates (X, Y, Z) after the conversion of the external fault and the three-dimensional coordinates (X) after the conversion of all the points detected by the power line k ,Y k ,Z k ) The Euclidean distance between coordinate points is obtained by three-dimensional coordinates of two spatial points:
using the minimum distance obtained by calculation as the power line for identifying the outer broken lineThe safety distance between them.
Preferably, the alarm of the external break in the twelfth step is performed according to the shortest distance between the external break identified by the YOLOv3 algorithm obtained by the eleventh step and the power line, and whether the external break can pose a security threat to the power transmission corridor or not, and an alarm is performed on an external break identification result smaller than a preset alarm distance.
The method is based on deep learning image recognition, a model compression technology and a sensor fusion technology, a deep neural network model is constructed by utilizing a computer vision theory and a deep learning framework, the model compression technology is adopted, and the miniaturization and the lightweight of the deep neural network model are realized through sparse training, channel pruning and fine tuning training, so that the intelligent recognition of the power transmission corridor video is realized. Meanwhile, a millimeter wave radar sensor and a camera sensor are fused, the depth distance of a target in an image is accurately acquired, coordinate system conversion is carried out according to the depth distance, the shortest distance between an outer broken target and a power line is further calculated and identified, and the shortest distance is taken as an alarm judging condition, so that the false alarm rate is effectively reduced on the premise of ensuring the detection precision, the outer broken real-time monitoring of a power transmission corridor is realized, the safety of the power transmission corridor is guaranteed, and the safety and stability of the operation of power equipment are improved.
Drawings
FIG. 1 is a schematic diagram of the model structure of the present invention.
Detailed Description
The following further describes a specific embodiment of the present invention with reference to the drawings, but the present invention is not limited to this embodiment.
As shown in fig. 1, a multi-sensor integrated intelligent external damage detection method for a power transmission corridor includes the following steps:
marking a monocular camera, collecting dozens of images with different shooting visual angles by using a black and white checkerboard calibration board according to the calibration board, extracting angular points of image squares, calculating internal and external parameters and a distortion coefficient of the camera, and finally correcting lens distortion of the camera;
secondly, calibrating the millimeter wave radar in a combined mode, placing a camera and the millimeter wave radar in a posture that an optical axis is parallel to the ground, placing a metal plate at different distance angles, recording the distance and the angle returned by the millimeter wave radar, and simultaneously recording a camera shooting image; converting a millimeter wave radar coordinate system into a world coordinate system according to the radar data and the position of target data on the image, converting the world coordinate system into a pixel coordinate system by calibrating parameters obtained by the camera in the first step, realizing the accurate standard of the millimeter wave radar for returning the target and the target in the camera image, and completing the fusion of the millimeter wave radar and the camera sensor;
thirdly, collecting and processing video image data of the power transmission corridor, uniformly cutting the video data collected by the power transmission corridor monitoring camera and storing the video data as image data with the size of 1280 pixels x720 pixels, taking the safety standard of the power transmission corridor as a basis, using an external force damage object mainly related to the standard as a corresponding label, and using a rectangular frame to mark an object to be detected to manufacture a data set required by model training;
fourthly, constructing a model, constructing a YOLOv3 model by using a Keras framework, extracting shape features, texture features and edge information contained in the image by using the model, inputting the extracted feature information into a deep neural network, calculating model parameters by using the deep neural network, and outputting an electric power operator safety monitoring and identifying result based on the image;
fifthly, model pre-training, namely expanding the number of the image data with labels by adopting data enhancement means such as Gaussian blur, random illumination, horizontal inversion and the like, inputting the image data with labels into a deep neural network for pre-training, iteratively optimizing model parameters through each round of prediction results, performing grouped sampling on samples with unbalanced categories to balance the number of the categories, continuing training the model by utilizing a sampling mode after balancing, and performing auxiliary calculation through a high-performance calculation accelerator card until the model parameters are converged to obtain an optimal model;
the sixth step, sparse training, namely endowing sparse factors to batch normalization layers in the deep neural network, designing sparse rate manually, then carrying out sparse training on an original data set, carrying out iterative optimization on the randomly initialized sparse factors to obtain deep neural network models with different batch normalization layer weight values, enabling partial weight values to be close to 0, and selecting the model with the best sparse degree and the best verification set effect balance, so as to obtain a convolution channel index with less influence on the model prediction result;
seventhly, channel pruning, namely sequencing the maximum weight values of all the batches of normalization layers according to the size of the sparse factor, taking the minimum value to calculate pruning rate, performing channel pruning on the convolution channel index obtained in the sixth step according to the rate, and adaptively searching the optimal pruning rate within the range of 10% above and below according to the identification effect of the verification set;
step eight, fine tuning training, namely adjusting the learning rate of the model according to the pruning ratio, retraining the pruned model, and recovering the accuracy of the model prediction result through repeated attempts to obtain an optimal compression model;
ninth, power line detection, namely extracting the edge of a linear power line by using Hough change, extracting the edge of the power line and realizing efficient and quick power line detection while target detection is realized;
step ten, converting a target point coordinate system, and converting an image coordinate identified by a YOLOv3 algorithm and a power line image coordinate into a three-dimensional coordinate under a world coordinate system according to the internal and external parameters of the camera obtained in the step one and the pixel depth value measured by the millimeter wave radar;
step eleven, calculating a target point safe distance, calculating Euclidean distances between an outer broken coordinate identified by a YOLOv3 algorithm under a world coordinate system and all power line coordinates, and selecting a minimum value between the outer broken coordinate and the power line as a safe distance for subsequent alarm judgment;
and a twelfth step of alarming the outer break, namely judging the shortest distance between the outer break identified by the YOLOv3 algorithm obtained by the eleventh step and the power line, judging whether the outer break can threaten the safety of the power transmission corridor according to the safety specification of the power transmission corridor, and carrying out voice broadcast and alarm on the identification result of the outer break smaller than the preset alarm distance.
Claims (11)
1. The utility model provides a broken detection method outside transmission corridor intelligence of multisensor integration which characterized in that: the method comprises the following steps:
firstly, calibrating a camera by adopting a Zhang Zhengyou method, collecting dozens of images with different shooting visual angles by using a calibration plate with black and white checkerboard grids according to the calibration plate, extracting angular points of the image checkerboard, analyzing internal and external parameters and distortion coefficients of the camera, and correcting the distortion of the camera by using the parameters;
secondly, calibrating the millimeter wave radar in a combined mode, placing a camera and the millimeter wave radar in a posture that an optical axis is parallel to the ground, placing a metal plate at different distance angles, recording the distance and the angle returned by the millimeter wave radar, and simultaneously recording a camera shooting image; aiming at the positions of radar data and target data on an image, a millimeter wave radar coordinate system is converted into a world coordinate system, the world coordinate system is converted into a pixel coordinate system by calibrating parameters obtained by a camera in the first step, and the conversion relation among the five coordinate systems of the millimeter wave radar, a three-dimensional world, the camera, the image and the pixel is accurately obtained, so that target point information returned by the millimeter wave radar can be accurately projected to the correct position of the image, and the fusion of the millimeter wave radar and the camera is realized;
thirdly, collecting and processing power transmission corridor video image data, uniformly cutting and storing the collected power transmission corridor video data into image data through manpower, taking the safety standard of the power transmission corridor as a basis, taking an external force damage object mainly related to the standard as a corresponding label, matching and labeling the label and an object in a data set, labeling the object to be detected with a rectangular frame, and manufacturing into a data set required by model training;
fourthly, constructing a model, constructing a YOLOv3 network by using a Keras framework, extracting a main structure of the network by taking Darknet53 as a feature, connecting a feature pyramid FPN module, realizing the extraction of shape features, texture features and edge information of the image, inputting the extracted feature information into a deep neural network, calculating model parameters by the deep neural network, and outputting a power transmission corridor external damage identification result based on the image;
fifthly, model pre-training, namely inputting the image data with labels into a deep neural network for pre-training by adopting a data enhancement and balanced sampling method comprising Gaussian blur, random illumination and horizontal inversion, iteratively optimizing model parameters through each round of prediction results, performing grouped sampling on samples with unbalanced classes to balance the number of the classes, continuously training the model by using a sampling mode after balancing, and performing auxiliary calculation through a high-performance calculation accelerator card until model parameters converge to obtain an optimal model;
sixthly, performing sparse training, namely endowing sparse factors to batch normalization layers in the deep neural network, manually designing a sparse rate, and then performing sparse training on an original data set to obtain deep neural network models with different batch normalization layer weight values;
seventhly, channel pruning, namely sequencing the maximum weight values of the normalization layers of each batch according to the size of the sparse factor, taking the minimum value to calculate the pruning rate, and adaptively searching the optimal pruning rate within the range of 10% above and below according to the identification effect of the verification set;
step eight, fine tuning training, namely retraining the pruned model to recover the accuracy of the model prediction result and obtain an optimal compression model;
ninth, power line detection, namely extracting the edge of a linear power line by using Hough change, extracting the edge of the power line and realizing efficient and quick power line detection while target detection is realized;
step ten, converting a target point coordinate system, and converting an image coordinate identified by a YOLOv3 algorithm and a power line image coordinate into a three-dimensional coordinate under a world coordinate system according to the camera internal and external parameters obtained in the step one and the pixel depth value measured by the millimeter wave radar;
step eleven, calculating a target point safe distance, calculating Euclidean distances between an outer broken coordinate identified by a YOLOv3 algorithm under a world coordinate system and all power line coordinates, and using the minimum value as a safe distance for subsequent alarm judgment;
and a twelfth step of alarming the outer break, wherein the shortest distance between the outer break and the power line identified by the YOLOv3 algorithm calculated in the eleventh step is judged, whether the outer break can generate a security threat to the power transmission corridor or not is judged, and an outer break identification result smaller than the preset alarm distance is alarmed.
2. The intelligent outburst detection method for the multi-sensor fused power transmission corridor according to claim 1, characterized by comprising the following steps: the detection method is composed of an image acquisition and identification module and a millimeter wave radar ranging module.
3. The intelligent outburst detection method for the multi-sensor fused power transmission corridor according to claim 1, characterized by comprising the following steps: and fourthly, constructing a YOLOv3 network model, performing downsampling through a continuous convolution module to obtain feature maps of different scales, stacking features among different layers through cross-layer connection to obtain feature maps of different granularities, effectively obtaining shape features, texture features and edge information in an original picture, extracting a main structure by taking Darknet53 as a feature, connecting a feature pyramid FPN module, integrating the features of different scales into the same size through convolution operation, stacking and combining the features into a whole for output through a channel, finally inputting the whole into a prediction module of a Yolo-Head, and predicting positions and types of possible targets through the convolution operation.
4. The intelligent outburst detection method for the multi-sensor fused power transmission corridor according to claim 1, characterized by comprising the following steps: and the fifth step of model training is to perform grouping sampling on samples with unbalanced classes and adopt a data enhancement and balanced sampling method to balance the number of the classes, wherein the enhancement method comprises Gaussian blur, random illumination and horizontal turnover.
5. The intelligent outburst detection method for the multi-sensor fused power transmission corridor according to claim 1, characterized by comprising the following steps: and the sixth step of sparse training is to train the weight of the batch normalization layer by using an introduced manually designed sparse factor on the premise that the pre-training model obtains the optimal model parameter, and the randomly initialized sparse factor reduces the precision loss of the sparse model on the verification set and enables part of the weight values to be close to 0 through iterative optimization, so that a convolution channel index with less influence on the model prediction result is obtained.
6. The intelligent outburst detection method for the multi-sensor fused power transmission corridor according to claim 1, characterized by comprising the following steps: and the seventh step of channel pruning is to perform channel pruning test on the convolution channel index obtained from the fifth step according to the ratio by using the step length set artificially according to the pruning ratio range obtained by calculation, verify the pruning effect on the verification set and screen out the optimal pruning ratio and model.
7. The intelligent outburst detection method for the multi-sensor fused power transmission corridor according to claim 1, characterized by comprising the following steps: and the eighth step of fine tuning training is to retrain the pruned model, wherein the learning rate needs to be adjusted according to the pruning rate at the moment, and the accuracy of the model prediction result is restored by repeatedly trying.
8. The intelligent outburst detection method for the multi-sensor fused power transmission corridor according to claim 1, characterized by comprising the following steps: and in the ninth step, the power line detection adopts Hough change to extract the edge of the linear power line, the pixel points are converted into curves of parameter space, when the pixel points on the image form a straight line, the curve is intersected with one point corresponding to the parameter space to form the pixel points of the straight line on the image, and the edge characteristics of the power line divided into approximate straight lines are searched by searching the intersection points of the parameter space.
9. The intelligent outdamage detection method for the power transmission corridor based on the fusion of the multiple sensors as claimed in claim 1, is characterized by involving the conversion of four coordinate systems: a world coordinate system describing the real position of the camera; a camera coordinate system with an origin as an optical center; an image coordinate system, wherein the origin is the midpoint of an imaging plane; and a pixel coordinate system with an origin at the upper left corner of the image, wherein a formula for converting a point from the pixel coordinate system to a world coordinate system can be obtained through the conversion of the above four coordinate systems:
in the formula, Z c The coordinate value of the camera coordinate system can be obtained by millimeter wave radar ranging, wherein f is the focal length of the camera, f x And f y Is the length and width of a single pixel (u) 0 ,v 0 ) Coordinates of the central point of the imaging plane belong to camera internal parameters; and R is a rotation matrix, and T is an offset matrix, and the first step is used for obtaining the rotation matrix.
10. The intelligent outburst detection method for the multi-sensor fused power transmission corridor according to claim 1, characterized by comprising the following steps: and the eleventh step of calculating the safe distance of the target point, namely calculating yolov3 algorithm to identify the three-dimensional coordinates (X, Y and Z) after the external damage conversion and the three-dimensional coordinates (X) after all the points of the power line detection are converted k ,Y k ,Z k ) The Euclidean distance between coordinate points is obtained by three-dimensional coordinates of two spatial points:
and taking the minimum distance obtained by calculation as the safety distance between the identification outcrop and the power line.
11. The intelligent outbreak detection method for the multi-sensor fused power transmission corridor according to claim 1, characterized by comprising the following steps: and the twelfth step of external damage alarming judges whether the external damage can generate a security threat to the power transmission corridor according to the shortest distance between the external damage identified by the YOLOv3 algorithm obtained by eleven-step calculation and the power line, and alarms the external damage identification result smaller than the preset alarming distance.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211341101.6A CN115809986A (en) | 2022-10-30 | 2022-10-30 | Multi-sensor fusion type intelligent external damage detection method for power transmission corridor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211341101.6A CN115809986A (en) | 2022-10-30 | 2022-10-30 | Multi-sensor fusion type intelligent external damage detection method for power transmission corridor |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115809986A true CN115809986A (en) | 2023-03-17 |
Family
ID=85482948
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211341101.6A Pending CN115809986A (en) | 2022-10-30 | 2022-10-30 | Multi-sensor fusion type intelligent external damage detection method for power transmission corridor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115809986A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117092631A (en) * | 2023-10-19 | 2023-11-21 | 江苏翰林正川工程技术有限公司 | Target positioning and ranging method and system for power transmission channel construction machinery |
| CN117092641A (en) * | 2023-10-20 | 2023-11-21 | 江苏翰林正川工程技术有限公司 | Information fusion method based on radar ranging and image recognition |
| CN117554942A (en) * | 2024-01-11 | 2024-02-13 | 江苏翰林正川工程技术有限公司 | Method and device for monitoring tree line distance of transmission line based on millimeter wave radar |
-
2022
- 2022-10-30 CN CN202211341101.6A patent/CN115809986A/en active Pending
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117092631A (en) * | 2023-10-19 | 2023-11-21 | 江苏翰林正川工程技术有限公司 | Target positioning and ranging method and system for power transmission channel construction machinery |
| CN117092631B (en) * | 2023-10-19 | 2024-04-19 | 江苏翰林正川工程技术有限公司 | Target positioning and ranging method and system for power transmission channel construction machinery |
| CN117092641A (en) * | 2023-10-20 | 2023-11-21 | 江苏翰林正川工程技术有限公司 | Information fusion method based on radar ranging and image recognition |
| CN117092641B (en) * | 2023-10-20 | 2023-12-15 | 江苏翰林正川工程技术有限公司 | Information fusion method based on radar ranging and image recognition |
| CN117554942A (en) * | 2024-01-11 | 2024-02-13 | 江苏翰林正川工程技术有限公司 | Method and device for monitoring tree line distance of transmission line based on millimeter wave radar |
| CN117554942B (en) * | 2024-01-11 | 2024-03-22 | 江苏翰林正川工程技术有限公司 | Method and device for monitoring tree line distance of transmission line based on millimeter wave radar |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106356757B (en) | A kind of power circuit unmanned plane method for inspecting based on human-eye visual characteristic | |
| CN109034018B (en) | Low-altitude small unmanned aerial vehicle obstacle sensing method based on binocular vision | |
| CN115809986A (en) | Multi-sensor fusion type intelligent external damage detection method for power transmission corridor | |
| CN108009515B (en) | Power transmission line positioning and identifying method of unmanned aerial vehicle aerial image based on FCN | |
| CN107560592B (en) | Precise distance measurement method for photoelectric tracker linkage target | |
| CN108317953A (en) | A kind of binocular vision target surface 3D detection methods and system based on unmanned plane | |
| CN106529538A (en) | Method and device for positioning aircraft | |
| CN115439424A (en) | Intelligent detection method for aerial video image of unmanned aerial vehicle | |
| CN109255286A (en) | A kind of quick detection recognition method of unmanned plane optics based on YOLO deep learning network frame | |
| CN110197185B (en) | Method and system for monitoring space under bridge based on scale invariant feature transform algorithm | |
| CN113901874A (en) | Tea tender shoot identification and picking point positioning method based on improved R3Det rotating target detection algorithm | |
| CN104239899A (en) | Electric transmission line spacer identification method for unmanned aerial vehicle inspection | |
| CN114089329A (en) | Target detection method based on fusion of long and short focus cameras and millimeter wave radar | |
| CN116385958A (en) | Edge intelligent detection method for power grid inspection and monitoring | |
| CN115995058A (en) | Power transmission channel safety on-line monitoring method based on artificial intelligence | |
| CN110110618A (en) | A kind of SAR target detection method based on PCA and global contrast | |
| CN116071424A (en) | Fruit space coordinate positioning method based on monocular vision | |
| CN115880231A (en) | Power transmission line hidden danger detection method and system based on deep learning | |
| CN114842340A (en) | Robot binocular stereoscopic vision obstacle sensing method and system | |
| CN115761265A (en) | Method and device for extracting substation equipment in laser radar point cloud | |
| Wang et al. | A moving target detection and localization strategy based on optical flow and pin-hole imaging methods using monocular vision | |
| CN110910379B (en) | Incomplete detection method and device | |
| CN113932712A (en) | Melon and fruit vegetable size measuring method based on depth camera and key points | |
| CN114396921A (en) | Qiantanjiang river tidal bore height and propagation speed measuring method based on unmanned aerial vehicle | |
| Li et al. | Quality assessment of polarization analysis images in foggy conditions |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication |