CN116301046A - Unmanned aerial vehicle aerial photographing safety distance automatic positioning method based on electric power pole tower identification - Google Patents
Unmanned aerial vehicle aerial photographing safety distance automatic positioning method based on electric power pole tower identification Download PDFInfo
- Publication number
- CN116301046A CN116301046A CN202310297767.4A CN202310297767A CN116301046A CN 116301046 A CN116301046 A CN 116301046A CN 202310297767 A CN202310297767 A CN 202310297767A CN 116301046 A CN116301046 A CN 116301046A
- Authority
- CN
- China
- Prior art keywords
- aerial vehicle
- unmanned aerial
- tower
- target
- electric power
- 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
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course or altitude of land, water, air, or space vehicles, e.g. automatic pilot
- G05D1/10—Simultaneous control of position or course in three dimensions
- G05D1/101—Simultaneous control of position or course in three dimensions specially adapted for aircraft
- G05D1/106—Change initiated in response to external conditions, e.g. avoidance of elevated terrain or of no-fly zones
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
- G06V10/765—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects using rules for classification or partitioning the feature space
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/17—Terrestrial scenes taken from planes or by drones
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
-
- 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 automatic positioning, in particular to an unmanned aerial vehicle aerial photographing safety distance automatic positioning method based on electric power pole and tower recognition, which aims at the problems that the current manual measurement causes large data error, the measurement is easily influenced by environmental factors, the manual measurement efficiency is low, the automatic positioning of the safety distance cannot be performed, and the like, and the scheme is provided as follows, and comprises the following steps: s1: the control platform gives an instruction, and the unmanned aerial vehicle moves according to a standard inspection path; the invention aims to adopt unmanned aerial vehicle aerial photography to automatically position the safety distance, reduce the data error caused by manual measurement, send the data information to a background computer in real time, carry out large-scale inspection on an electric power tower, improve the real-time accuracy of the data, reduce the data error, reduce the loss, and combine the technical means of three-dimensional modeling and the like to carry out unmanned aerial vehicle planning route autonomous flight, thereby realizing the operation and maintenance digitization, the intellectualization and the informatization management of the transmission line.
Description
Technical Field
The invention relates to the technical field of automatic positioning, in particular to an unmanned aerial vehicle aerial photographing safety distance automatic positioning method based on electric power pole and tower recognition.
Background
The existing inspection of the power transmission tower line system is mainly completed by manual operation of an maintainer, the labor intensity is high, the time consumption is long, the danger is high, the efficiency is low, and the inspection of the unmanned power transmission tower line system is a non-contact power tower line system detection technology mainly based on machine vision. Detection methods based on unmanned aerial vehicle vision technology are currently receiving more and more attention and application. However, at present, the unmanned aerial vehicle is mainly responsible for shooting a tower line on a flight path, and the stored image and video information are delivered to professionals for identification and statistics to finally obtain a patrol report. Therefore, the machine vision and artificial intelligence technology is applied to unmanned aerial vehicle detection of the transmission tower line system, so that the unmanned aerial vehicle can effectively identify and position the target component of the power line in real time, and the intelligent level and the working efficiency of inspection can be greatly improved.
Along with popularization and high-speed development of unmanned aerial vehicle technology, the unmanned aerial vehicle technology is gradually increased in application in aspects of power grid line inspection, emergency response, design construction and the like, so that the cost budget of power enterprises can be effectively reduced, the working efficiency of a power system is improved, and the failure rate can be greatly reduced. Therefore, we propose an unmanned aerial vehicle aerial safety distance automatic positioning method based on electric power pole tower recognition.
Disclosure of Invention
The invention aims to solve the problems that the manual measurement causes large data error, the measurement is easily influenced by environmental factors, the manual measurement efficiency is low, the automatic positioning of the safety distance cannot be performed, and the like, and provides an unmanned aerial vehicle aerial safety distance automatic positioning method based on power tower identification.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
an unmanned aerial vehicle aerial photographing safety distance automatic positioning method based on electric power tower identification comprises the following steps:
s1: the control platform gives an instruction, the unmanned aerial vehicle moves according to a standard inspection path, and the transmission line corridor and the tower body are inspected at a multi-angle in a short distance through the carried high-definition camera;
s2: target identification, integrating YOLO into a single neural network model, and storing and transmitting picture information to a console computer in real time;
s3: YOLOV3 performs target detection and identification, divides an image, and performs identification and classification of objects in the image;
s4: YOLOV3 trains and tests the image dataset, and judges the accuracy of identification by using the accuracy and recall;
s5: introducing a compression activation network into the f-AnoGAN encoder, extracting significance information of the visible light picture, and reducing background noise interference;
S6: and carrying out safe distance positioning by data space acquisition, and carrying out reverse three-dimensional modeling of the electric power facilities based on the point cloud data.
Preferably, in the step S1, an operator issues an instruction from a management and control platform, the unmanned aerial vehicle quickly reaches a preset tower, moves according to a standard inspection path, accurately locates inspection emphasis, automatically rounds to the other side to continue inspection after inspection of one side of the tower, completes inspection of a first base tower after 10min, automatically flies to the next tower, automatically and safely returns to an unmanned aerial vehicle warehouse after the unmanned aerial vehicle completes all inspection tasks, and stores energy to wait for the next flight task.
Preferably, in the step S1, the unmanned aerial vehicle automatically takes off, automatically patrols and examines and automatically returns to land according to a preset route and task, and through the carried high-definition camera, the transmission line corridor and the tower body are patrolled and examined at a short distance at multiple angles, the tower part is accurately positioned through three-dimensional modeling, the patrol route is accurately planned, each patrol point position is set, and the compiled patrol route is uploaded to the unmanned aerial vehicle.
Preferably, in the step S2, the target recognition mainly includes determination of a target position and discrimination of a target class, the early target detection and recognition method divides a target recognition task into target region prediction and target class prediction, YOLO integrates the target region prediction and the target class prediction into a single neural network model, and the whole image training model is used to directly regress out the frame and the class of the target at a plurality of positions of the image, thereby realizing end-to-end target detection and recognition.
Preferably, in the step S2, a camera with extremely high definition is arranged on the unmanned aerial vehicle to shoot pictures and videos of the inspection site, the picture information is stored and transmitted to a console computer in real time, before the computer, a worker judges whether the running state of the line is normal according to the transmitted information, if the line is out of order, the line can be removed in time, and a visual detection technology is adopted to remove some faults reflected on the surface of the line.
Preferably, in the step S3, YOLOV3 performs object detection and recognition, divides an image, predicts and classifies bounding boxes, YOLOV3 divides an input image into 5×5 grids, predicts a plurality of bounding boxes and confidence levels thereof at three different scales and probability information of a plurality of objects belonging to a certain category, the bounding box information is offset, width and height of a center position of the object relative to the position of the grids, the confidence levels reflect whether the object is contained and the accuracy of the position in the case of the object being contained, YOLOV3 predicts a score of one object for each bounding box through logistic regression, CNN extracts features in grids, performs a series of convolution operation on the whole image to obtain feature graphs, extracts features in each frame on the feature graphs to form high-dimensional feature vectors, predicts a plurality of bounding boxes for each grid, and discriminates the object according to the calculated classification errors, confidence levels, category probabilities and the like to recognize and classify the object in the image.
Preferably, in the step S4, YOLOV3 trains and tests an image dataset, the training set has 2000 images, wherein 500 images are respectively used for the power transmission tower, the damper, the insulator and the equalizing ring, and the testing set has 1000 images respectively, 250 images are respectively used for the power transmission tower, the damper, the insulator and the equalizing ring, each image in the training set clearly shows the power component, and the accuracy of identification is judged by the accuracy and the recall, wherein the accuracy is the number of peripheral frames with correct target category marks divided by the number of all marked peripheral frames; the recall rate is the number of peripheral frames of the target class label that are correct divided by the number of peripheral frames of all class targets.
Preferably, in the step S5, a compression activation network is introduced into the f-AnoGAN encoder, the saliency information in the image is extracted, and then, unsupervised learning of the antagonism network and supervised learning of the two classifiers are generated and organically combined, so that the generalization performance of the model on a large-scale data set is further effectively improved by means of an optimization training strategy based on transfer learning.
Preferably, in the step S5, a compressed activation network based on channel attention is introduced into an encoder of the original f-AnoGAN network, so as to extract the significance information of the visible light picture, reduce the background noise interference, and effectively extract the features of the tower.
Preferably, in the step S6, the data space is acquired to perform safe distance positioning, and the vertical distance between two points in space is suitable for inquiring the crossing distance of a line, the distance between the lowest suspended point of the line and the ground or the tree, the vertical distance between any position of a pole tower and the ground, and the height difference between the pole towers; the method comprises the steps of marking three-dimensional coordinates of any point, marking spatial position positioning information of insulators of pole and tower hardware, enabling the vertical distance from the point to a straight line, inquiring the vertical distance between two mutually parallel wires, inquiring an included angle formed by any three vertexes, inquiring an included angle formed by connecting wires of adjacent poles and towers, carrying out reverse three-dimensional modeling on electric power facilities based on point cloud data, loading electric power asset information into an information management platform in a model form, and mounting attribute information for the model.
The beneficial effects of the invention are as follows:
1. the traditional manual power line inspection mode is limited by environmental topography, has low efficiency, and the unmanned aerial vehicle has the characteristics of light and reliability, compact structure, excellent performance and the like, is not limited by geographic conditions and environmental conditions, and is particularly suitable for executing tasks in complex environments.
2. The unmanned aerial vehicle can find important key defects in the line inspection process, provide accurate information for an operation unit in time, avoid line accident power failure and reduce high-power failure loss cost. The position of the potential accident can be rapidly mastered by rapid inspection of the large-scale flight.
The invention aims to adopt unmanned aerial vehicle aerial photography to automatically position the safety distance, reduce the data error caused by manual measurement, enable unmanned aerial vehicle inspection to be free from the limitation of environment and geographical conditions, send data information to a background computer in real time, enable the unmanned aerial vehicle inspection to be carried out on a large scale, enable a power tower to be rapidly checked, improve the real-time accuracy of the data, reduce the data error, enable the background to maintain equipment in time, reduce power failure risk, reduce loss, improve operation efficiency, and enable unmanned aerial vehicle planning route to fly autonomously in combination with the technical means of three-dimensional modeling and the like, so as to realize the digital, intelligent and informationized management of the operation and maintenance of a power transmission line.
Drawings
Fig. 1 is a schematic flow chart of an unmanned aerial vehicle aerial safety distance automatic positioning method based on electric power tower recognition.
Detailed Description
The following description of the technical solutions in the embodiments of the present invention will be clear and complete, and it is obvious that the described embodiments are only some embodiments of the present invention, but not all embodiments.
Example 1
Referring to fig. 1, an unmanned aerial vehicle aerial photographing safety distance automatic positioning method based on power tower identification comprises the following steps:
S1: the control platform gives an instruction, the unmanned aerial vehicle moves according to a standard inspection path, and the transmission line corridor and the tower body are inspected at a multi-angle in a short distance through the carried high-definition camera;
s2: target identification, integrating YOLO into a single neural network model, and storing and transmitting picture information to a console computer in real time;
s3: YOLOV3 performs target detection and identification, divides an image, and performs identification and classification of objects in the image;
s4: YOLOV3 trains and tests the image dataset, and judges the accuracy of identification by using the accuracy and recall;
s5: introducing a compression activation network into the f-AnoGAN encoder, extracting significance information of the visible light picture, and reducing background noise interference;
s6: and carrying out safe distance positioning by data space acquisition, and carrying out reverse three-dimensional modeling of the electric power facilities based on the point cloud data.
In this embodiment, operating personnel gives the instruction from managing and controlling the platform, and unmanned aerial vehicle reaches predetermined shaft tower fast, and the route removes according to the standard inspection, and the focus is patrolled to accurate location, and automatic the crossing continues to patrol to the opposite side after having patrolled and examined shaft tower one side, and 10min later unmanned aerial vehicle accomplish first basic shaft tower and patrol and examine, fly to next shaft tower automatically, and unmanned aerial vehicle accomplishes the whole task of patrolling and examining after, automatic safety returns to unmanned aerial vehicle hangar to energy storage waits for the next flight task.
In this embodiment, take off independently according to predetermined route and task, independently patrol and examine, independently return to the home and land to through the high definition digtal camera who carries on, the transmission line corridor is patrolled and examined closely to the multi-angle, shaft tower body, through three-dimensional modeling accurate location shaft tower position, the accurate planning is patrolled and examined the route, sets up each and patrol the position, with the good route of patrolling and examining of compiling uploading to unmanned aerial vehicle.
In this embodiment, target recognition mainly includes determination of a target position and discrimination of a target class, an early target detection and recognition method divides a target recognition task into target region prediction and target class prediction, YOLO integrates the target region prediction and the target class prediction into a single neural network model, and uses a whole image training model to directly regress out a frame and a class of a target at a plurality of positions of an image, thereby realizing end-to-end target detection and recognition.
In this embodiment, a camera with extremely high definition is arranged on the unmanned aerial vehicle to shoot a patrol scene picture and a video, picture information is stored and transmitted to a console computer in real time, a worker before the computer judges whether the running state of a line is normal according to the transmitted information, if the line is in failure, the line can be removed in time, and a visual detection technology is adopted to remove some failures reflected on the surface of the line.
In this embodiment, YOLOV3 performs object detection and recognition, image segmentation, prediction and classification of bounding boxes, YOLOV3 divides an input image into 5×5 grids, each grid predicts a plurality of bounding boxes and confidence levels thereof at three different scales, and probability information of a plurality of objects belonging to a certain class, the bounding box information is offset, width and height of a center position of an object relative to the position of the grid, the confidence levels reflect whether the object is contained and the accuracy of the position of the object in the case that the object is contained, YOLOV3 predicts the score of one object for each bounding box through logistic regression, CNN extracts features in grids, performs a series of convolution operations on the whole image to obtain a feature map, takes out features in each frame on the feature map to form a high-dimensional feature vector, predicts a plurality of bounding boxes, and discriminates the object according to the calculated classification errors, confidence levels, class probabilities and the like.
In the embodiment, the YOLOV3 trains and tests an image dataset, the training set has 2000 images, wherein 500 images are respectively used for a power transmission tower, a damper, an insulator and an equalizing ring, and the testing set has 1000 images respectively, 250 images are respectively used for the power transmission tower, the damper, the insulator and the equalizing ring, each image in the training set clearly shows a power component, and the accuracy of identification is judged by using the accuracy and the recall, wherein the accuracy is the number of peripheral frames with correct target category marks divided by the number of all marked peripheral frames; the recall rate is the number of peripheral frames of the target class label that are correct divided by the number of peripheral frames of all class targets.
In the embodiment, a compression activation network is introduced into an f-AnoGAN encoder, significance information in an image is extracted, then unsupervised learning of an antagonistic network and supervised learning of a two-classifier are generated and organically combined, and generalization performance of a model on a large-scale data set is further effectively improved by means of an optimized training strategy based on transfer learning.
In the embodiment, a compression activation network based on channel attention is introduced into an encoder of an original f-AnoGAN network, the significance information of visible light pictures is extracted, the background noise interference is reduced, and the features of a pole tower are effectively extracted.
In the embodiment, the data space is acquired to perform safe distance positioning, the vertical distance between two points in space is suitable for inquiring the crossing distance of a line, the distance between the lowest suspended point of the line and the ground or the distance between trees, the vertical distance between any position of a pole tower and the ground, and the height difference between the pole towers; the method comprises the steps of marking three-dimensional coordinates of any point, marking spatial position positioning information of insulators of pole and tower hardware, enabling the vertical distance from the point to a straight line, inquiring the vertical distance between two mutually parallel wires, inquiring an included angle formed by any three vertexes, inquiring an included angle formed by connecting wires of adjacent poles and towers, carrying out reverse three-dimensional modeling on electric power facilities based on point cloud data, loading electric power asset information into an information management platform in a model form, and mounting attribute information for the model.
Example two
Referring to fig. 1, an unmanned aerial vehicle aerial photographing safety distance automatic positioning method based on power tower identification comprises the following steps:
s1: the control platform gives an instruction, the unmanned aerial vehicle moves according to a standard inspection path, and the transmission line corridor and the tower body are inspected at a multi-angle in a short distance through the carried high-definition camera;
s2: target identification, integrating YOLO into a single neural network model, and storing and transmitting picture information to a console computer in real time;
s3: YOLOV3 performs target detection and identification, divides an image, and performs identification and classification of objects in the image;
s4: YOLOV3 trains and tests the image dataset, and judges the accuracy of identification by using the accuracy and recall;
s5: introducing a compression activation network into the f-AnoGAN encoder, extracting significance information of the visible light picture, and reducing background noise interference;
s6: and carrying out safe distance positioning by data space acquisition, and carrying out reverse three-dimensional modeling of the electric power facilities based on the point cloud data.
In this embodiment, operating personnel gives the instruction from managing and controlling the platform, and unmanned aerial vehicle reaches predetermined shaft tower fast, and the route removes according to the standard inspection, and accurate location inspection is important, and automatic flies to next shaft tower, and unmanned aerial vehicle accomplish all inspection tasks after, automatic safe return to unmanned aerial vehicle hangar to energy storage waits for the next flight task.
In this embodiment, the unmanned aerial vehicle automatically takes off, automatically patrols and examines, and automatically returns to the home to land according to a preset route and task, and accurately plans and patrols and examines the route through the high definition camera that carries on, sets up each and patrols and examines the point location, and uploads the route of patrolling and examining that compiles to the unmanned aerial vehicle.
In this embodiment, target recognition mainly includes determination of a target position and discrimination of a target class, and an early target detection and recognition method divides a target recognition task into target region prediction and target class prediction, uses a whole image training model, and directly regresses a frame and a class of a target at a plurality of positions of an image, thereby realizing end-to-end target detection and recognition.
In this embodiment, a camera with extremely high definition is arranged on the unmanned aerial vehicle to shoot pictures and videos of the inspection scene, the picture information is stored and transmitted to the console computer in real time, if faults occur, the faults reflected on the surface of the line can be eliminated in time by adopting a visual detection technology.
In this embodiment, YOLOV3 performs object detection and recognition, segments an image, predicts and classifies bounding boxes, YOLOV3 divides an input image into 5×5 grids, predicts a plurality of bounding boxes and confidence levels thereof and probability information that a plurality of objects belong to a certain class in three different scales, YOLOV3 predicts a score of an object through logistic regression for each bounding box, CNN extracts features in a grid, performs a series of convolution operations on the whole image to obtain a feature map, and discriminates objects according to calculated classification errors, confidence levels, class probabilities and the like to perform recognition classification of the objects in the image.
In the embodiment, the YOLOV3 trains and tests an image dataset, the training set has 2000 images, the testing set has 1000 images, and the power transmission tower, the damper, the insulator and the equalizing ring have 250 images respectively, wherein the accuracy is the number of peripheral frames with correct target class marks divided by the number of all marked peripheral frames; the recall rate is the number of peripheral frames of the target class label that are correct divided by the number of peripheral frames of all class targets.
In the embodiment, a compression activation network is introduced into an f-AnoGAN encoder, significance information in an image is extracted, then unsupervised learning of an antagonistic network and supervised learning of a two-classifier are generated and organically combined, and generalization performance of a model on a large-scale data set is further effectively improved by means of an optimized training strategy based on transfer learning.
In the embodiment, a compression activation network based on channel attention is introduced into an encoder of an original f-AnoGAN network, the significance information of visible light pictures is extracted, the background noise interference is reduced, and the features of a pole tower are effectively extracted.
In the embodiment, the data space is acquired to perform safe distance positioning, the vertical distance between two points in space is suitable for inquiring the crossing distance of a line, the distance between the lowest suspended point of the line and the ground or the distance between trees, the vertical distance between any position of a pole tower and the ground, and the height difference between the pole towers; inquiring the vertical distance between two mutually parallel wires, inquiring the included angle formed by any three vertexes, inquiring the included angle of the connecting wires of the adjacent towers, carrying out reverse three-dimensional modeling on the electric power facilities based on point cloud data, loading electric power asset information into an information management platform in a model form, and mounting attribute information for the model.
Example III
Referring to fig. 1, an unmanned aerial vehicle aerial photographing safety distance automatic positioning method based on power tower identification comprises the following steps:
s1: the control platform gives an instruction, the unmanned aerial vehicle moves according to a standard inspection path, and the transmission line corridor and the tower body are inspected at a multi-angle in a short distance through the carried high-definition camera;
s2: target identification, integrating YOLO into a single neural network model, and storing and transmitting picture information to a console computer in real time;
s3: YOLOV3 performs target detection and identification, divides an image, and performs identification and classification of objects in the image;
s4: YOLOV3 trains and tests the image dataset, and judges the accuracy of identification by using the accuracy and recall;
s5: introducing a compression activation network into the f-AnoGAN encoder, extracting significance information of the visible light picture, and reducing background noise interference;
s6: and carrying out safe distance positioning by data space acquisition, and carrying out reverse three-dimensional modeling of the electric power facilities based on the point cloud data.
In this embodiment, operating personnel gives the instruction from managing and controlling the platform, and unmanned aerial vehicle reaches predetermined shaft tower fast, and the route removes according to the standard inspection, and accurate location inspection is important, and automatic flies to next shaft tower, and unmanned aerial vehicle accomplish all inspection tasks after, automatic safe return to unmanned aerial vehicle hangar to energy storage waits for the next flight task.
In this embodiment, take off independently according to predetermined route and task, independently patrol and examine, independently return to the home and land to through the high definition digtal camera who carries on, the transmission line corridor is patrolled and examined closely to the multi-angle, shaft tower body, through three-dimensional modeling accurate location shaft tower position, the accurate planning is patrolled and examined the route, sets up each and patrol the position, with the good route of patrolling and examining of compiling uploading to unmanned aerial vehicle.
In this embodiment, target recognition mainly includes determination of a target position and discrimination of a target class, and an early target detection and recognition method divides a target recognition task into target region prediction and target class prediction, uses a whole image training model, and directly regresses a frame and a class of a target at a plurality of positions of an image, thereby realizing end-to-end target detection and recognition.
In this embodiment, a camera with extremely high definition is arranged on the unmanned aerial vehicle to shoot pictures and videos of the inspection scene, the picture information is stored and transmitted to the console computer in real time, if faults occur, the faults reflected on the surface of the line can be eliminated in time by adopting a visual detection technology.
In this embodiment, YOLOV3 performs object detection and identification, segments an image, predicts and classifies bounding boxes, YOLOV3 divides an input image into 5×5 grids, each grid predicts a plurality of bounding boxes and confidence levels thereof at three different scales, the confidence levels reflect whether an object is contained and the accuracy of the position of the object in the case that the object is contained, YOLOV3 predicts the score of one object for each bounding box through logistic regression, CNN extracts features in the grids, performs a series of convolution operations on the whole image to obtain a feature map, takes out features in each frame on the feature map to form a high-dimensional feature vector, and determines objects such as confidence levels, class probability and the like to perform recognition classification of the objects in the image.
In the embodiment, the YOLOV3 trains and tests an image dataset, the training set has 2000 images, 250 images of a power transmission tower, a damper, an insulator and a grading ring respectively, each image in the training set clearly shows a power component, and the accuracy of identification is judged by using the accuracy and the recall rate, wherein the accuracy is the number of peripheral frames with correct target category marks divided by the number of all marked peripheral frames; the recall rate is the number of peripheral frames of the target class label that are correct divided by the number of peripheral frames of all class targets.
In the embodiment, a compression activation network is introduced into an f-AnoGAN encoder, significance information in an image is extracted, then unsupervised learning of an antagonistic network and supervised learning of a two-classifier are generated and organically combined, and generalization performance of a model on a large-scale data set is further effectively improved by means of an optimized training strategy based on transfer learning.
In the embodiment, a compression activation network based on channel attention is introduced into an encoder of an original f-AnoGAN network, the significance information of visible light pictures is extracted, the background noise interference is reduced, and the features of a pole tower are effectively extracted.
In this embodiment, the data space acquisition performs safe distance positioning, the vertical distance between two points in space is suitable for inquiring the line crossing distance, the three-dimensional coordinates of any point, marking the spatial position positioning information of the insulator of the pole tower hardware, the vertical distance from the point to the straight line, inquiring the vertical distance between two mutually parallel wires, the included angle formed by any three vertexes, inquiring the included angle of the connecting wires of the adjacent pole towers, performing reverse three-dimensional modeling of the electric power facilities based on the point cloud data, loading the electric power asset information into the information management platform in the form of a model, and mounting attribute information for the model.
Example IV
Referring to fig. 1, an unmanned aerial vehicle aerial photographing safety distance automatic positioning method based on power tower identification comprises the following steps:
s1: the control platform gives an instruction, the unmanned aerial vehicle moves according to a standard inspection path, and the transmission line corridor and the tower body are inspected at a multi-angle in a short distance through the carried high-definition camera;
s2: target identification, integrating YOLO into a single neural network model, and storing and transmitting picture information to a console computer in real time;
s3: YOLOV3 performs target detection and identification, divides an image, and performs identification and classification of objects in the image;
s4: YOLOV3 trains and tests the image dataset, and judges the accuracy of identification by using the accuracy and recall;
s5: introducing a compression activation network into the f-AnoGAN encoder, extracting significance information of the visible light picture, and reducing background noise interference;
s6: and carrying out safe distance positioning by data space acquisition, and carrying out reverse three-dimensional modeling of the electric power facilities based on the point cloud data.
In this embodiment, operating personnel gives the instruction from managing and controlling the platform, and unmanned aerial vehicle reaches predetermined shaft tower fast, and the route removes according to the standard inspection, and the key is patrolled to accurate location, and automatic the crossing continues to patrol to the opposite side after having patrolled and examined shaft tower one side, and unmanned aerial vehicle accomplishes first basic shaft tower and patrol and examine after 10min, and automatic safety returns to unmanned aerial vehicle hangar to energy storage waits for next flight task.
In this embodiment, the unmanned aerial vehicle automatically takes off, automatically patrols and examines, and automatically returns to the home to land according to a preset route and task, and accurately plans and patrols and examines the route through the high definition camera that carries on, sets up each and patrols and examines the point location, and uploads the route of patrolling and examining that compiles to the unmanned aerial vehicle.
In this embodiment, the target recognition mainly includes determination of a target position and discrimination of a target class, and the early target detection and recognition method divides a target recognition task into target region prediction and target class prediction, and YOLO integrates the target region prediction and the target class prediction into a single neural network model to realize end-to-end target detection and recognition.
In this embodiment, a camera with extremely high definition is arranged on the unmanned aerial vehicle to shoot a patrol scene picture and a video, picture information is stored and transmitted to a console computer in real time, a worker before the computer judges whether the running state of a line is normal according to the transmitted information, if the line is in failure, the line can be removed in time, and a visual detection technology is adopted to remove some failures reflected on the surface of the line.
In this embodiment, YOLOV3 performs object detection and recognition, segments an image, predicts and classifies bounding boxes, YOLOV3 divides an input image into 5×5 grids, predicts a plurality of bounding boxes and their confidence levels in three different scales for each grid, YOLOV3 predicts the score of an object through logistic regression for each bounding box, CNN extracts features in grids, performs a series of convolution operations on the whole image to obtain a feature map, takes out features in each frame on the feature map to form a high-dimensional feature vector, predicts a plurality of bounding boxes for each grid, and discriminates objects according to calculated classification errors, confidence levels, class probabilities and the like to perform recognition classification of objects in the image.
In the embodiment, the YOLOV3 trains and tests an image dataset, wherein the training set has 2000 images, 500 images are respectively used for a power transmission tower, a damper, an insulator and a grading ring, and 250 images are respectively used for a testing set, wherein the accuracy is the number of peripheral frames with correct target category marks divided by the number of all marked peripheral frames; the recall rate is the number of peripheral frames of the target class label that are correct divided by the number of peripheral frames of all class targets.
In the embodiment, a compression activation network is introduced into an f-AnoGAN encoder, significance information in an image is extracted, then unsupervised learning of an antagonistic network and supervised learning of a two-classifier are generated and organically combined, and generalization performance of a model on a large-scale data set is further effectively improved by means of an optimized training strategy based on transfer learning.
In the embodiment, a compression activation network based on channel attention is introduced into an encoder of an original f-AnoGAN network, the significance information of visible light pictures is extracted, the background noise interference is reduced, and the features of a pole tower are effectively extracted.
In the embodiment, the data space is acquired to perform safe distance positioning, the vertical distance between two points in space is suitable for inquiring the crossing distance of a line, the distance between the lowest suspended point of the line and the ground or the distance between trees, the vertical distance between any position of a pole tower and the ground, and the height difference between the pole towers; inquiring the vertical distance between two mutually parallel wires, forming an included angle by any three vertexes, carrying out reverse three-dimensional modeling on the electric power facilities based on point cloud data, loading electric power asset information into an information management platform in a model form, and mounting attribute information for the model.
Example five
Referring to fig. 1, an unmanned aerial vehicle aerial photographing safety distance automatic positioning method based on power tower identification comprises the following steps:
s1: the control platform gives an instruction, the unmanned aerial vehicle moves according to a standard inspection path, and the transmission line corridor and the tower body are inspected at a multi-angle in a short distance through the carried high-definition camera;
s2: target identification, integrating YOLO into a single neural network model, and storing and transmitting picture information to a console computer in real time;
s3: YOLOV3 performs target detection and identification, divides an image, and performs identification and classification of objects in the image;
s4: YOLOV3 trains and tests the image dataset, and judges the accuracy of identification by using the accuracy and recall;
s5: introducing a compression activation network into the f-AnoGAN encoder, extracting significance information of the visible light picture, and reducing background noise interference;
s6: and carrying out safe distance positioning by data space acquisition, and carrying out reverse three-dimensional modeling of the electric power facilities based on the point cloud data.
In this embodiment, operating personnel gives the instruction from managing and controlling the platform, and unmanned aerial vehicle reaches predetermined shaft tower fast, and according to the route removal of patrolling and examining of rule, accurate location patrols the key, and the automatic other side that crosses after having patrolled and examined shaft tower one side continues to patrol and examine, and automatic safety returns to unmanned aerial vehicle hangar to energy storage waits for next flight mission.
In this embodiment, take off independently according to predetermined route and task, independently patrol and examine, independently return to the home and land to through the high definition digtal camera who carries on, the transmission line corridor is patrolled and examined closely to the multi-angle, shaft tower body, through three-dimensional modeling accurate location shaft tower position, the accurate planning is patrolled and examined the route, sets up each and patrol the position, with the good route of patrolling and examining of compiling uploading to unmanned aerial vehicle.
In this embodiment, the target recognition mainly includes determination of a target position and discrimination of a target class, and the early target detection and recognition method divides a target recognition task into target region prediction and target class prediction, and YOLO integrates the target region prediction and the target class prediction into a single neural network model, and uses the whole image training model to realize end-to-end target detection and recognition.
In this embodiment, a camera with extremely high definition is arranged on the unmanned aerial vehicle to shoot pictures and videos of the inspection scene, the picture information is stored and transmitted to the console computer in real time, and a visual detection technology is adopted to eliminate faults reflected on the surface of a line.
In this embodiment, YOLOV3 performs object detection and recognition, segments an image, predicts and classifies bounding boxes, YOLOV3 divides an input image into 5×5 grids, predicts a plurality of bounding boxes and their confidence levels in three different scales for each grid, YOLOV3 predicts the score of an object through logistic regression for each bounding box, CNN extracts features in grids, performs a series of convolution operations on the whole image to obtain a feature map, takes out features in each frame on the feature map to form a high-dimensional feature vector, predicts a plurality of bounding boxes for each grid, and discriminates objects according to calculated classification errors, confidence levels, class probabilities and the like to perform recognition classification of objects in the image.
In this embodiment, YOLOV3 trains and tests an image dataset, the training set has 2000 images, wherein 500 images are respectively used for the power transmission tower, the damper, the insulator and the equalizing ring, and 250 images are respectively used for the testing set, and the recall rate is the number of peripheral frames with correct target class marks divided by the number of peripheral frames of all class targets.
In the embodiment, a compression activation network is introduced into an f-AnoGAN encoder, significance information in an image is extracted, then unsupervised learning of an antagonistic network and supervised learning of a two-classifier are generated and organically combined, and generalization performance of a model on a large-scale data set is further effectively improved by means of an optimized training strategy based on transfer learning.
In the embodiment, a compression activation network based on channel attention is introduced into an encoder of an original f-AnoGAN network, the significance information of visible light pictures is extracted, the background noise interference is reduced, and the features of a pole tower are effectively extracted.
In the embodiment, the data space is acquired to perform safe distance positioning, the vertical distance between two points in space is suitable for inquiring the crossing distance of a line, the distance between the lowest suspended point of the line and the ground or the distance between trees, the vertical distance between any position of a pole tower and the ground, and the height difference between the pole towers; and marking the spatial position positioning information of the insulator of the pole tower hardware fitting by the three-dimensional coordinates of any point, inquiring the included angle of the connecting wires of the adjacent pole towers, carrying out reverse three-dimensional modeling on the electric power facilities based on the point cloud data, loading the electric power asset information into an information management platform in a model form, and mounting attribute information for the model.
Comparative example one
The first difference from the embodiment is that S1: the control platform gives an instruction, the unmanned aerial vehicle moves according to a standard inspection path, through the high-definition camera carried, the multi-angle short-distance inspection transmission line corridor and the tower body are provided, the operator gives the instruction from the control platform, the unmanned aerial vehicle quickly reaches a preset tower, the unmanned aerial vehicle moves according to the standard inspection path, the inspection key is precisely positioned, the unmanned aerial vehicle automatically walks to the other side after finishing inspecting one side of the tower to continue inspecting, the unmanned aerial vehicle completes the inspection of the first basic tower after 10min, the unmanned aerial vehicle automatically flies to the next tower, after completing all inspection tasks, the unmanned aerial vehicle automatically and safely returns to the unmanned aerial vehicle library, and the energy storage waits for the next flight tasks, the unmanned aerial vehicle automatically takes off, automatically inspects and autonomously returns to land according to preset navigation lines and tasks, and through the carried high-definition camera, the multi-angle short-distance inspection transmission line corridor and the tower body are precisely positioned, the navigation line is precisely positioned through three-dimensional modeling, each inspection navigation point is precisely planned, and the inspection navigation line is well is transferred to the unmanned aerial vehicle.
Comparative example two
The second difference from the embodiment is that S2: the method mainly comprises the steps of target recognition, namely integrating YOLO into a single neural network model, storing and transmitting picture information to a console computer in real time, wherein the target recognition mainly comprises determination of a target position and discrimination of a target class, an early target detection and recognition method divides a target recognition task into target region prediction and target class prediction, a camera with extremely high definition is arranged on an unmanned aerial vehicle to shoot a patrol scene picture and video, the picture information is stored and transmitted to the console computer in real time, if faults occur, the faults reflected on the surface of a line can be eliminated in time by adopting a visual detection technology.
Comparative example three
The difference from the third embodiment is that S3: YOLOV3 carries out target detection recognition, image segmentation, object recognition and classification in the image, YOLOV3 carries out target detection recognition, image segmentation, prediction and classification of boundary boxes, YOLOV3 divides an input image into 5×5 grids, each grid predicts a plurality of boundary boxes and confidence degrees thereof in three different scales, the confidence degrees reflect whether objects are contained or not and the accuracy of positions of the objects in the condition that the objects are contained, YOLOV3 predicts the score of one object on each boundary box through logistic regression, CNN extracts the characteristics in grids, carries out series convolution operation on the whole image to obtain a characteristic map, takes out the characteristics in each frame on the characteristic map to form high-dimensional characteristic vectors, and judges the objects to carry out recognition classification of the objects in the image, such as confidence degrees, class probabilities.
Experimental example
The automatic positioning method for the aerial photographing safety distance of the unmanned aerial vehicle based on the power tower identification is tested in the first embodiment, the second embodiment, the third embodiment, the fourth embodiment and the fifth embodiment, and the following result is obtained:
| example 1 | Example two | Example III | Existing methods | |
| Measurement efficiency | 91％ | 74％ | 71％ | 68％ |
| Data accuracy rate | 95％ | 61％ | 68％ | 71％ |
The automatic positioning method of the unmanned aerial vehicle aerial photographing safety distance based on the power tower identification in the first embodiment, the second embodiment, the third embodiment, the fourth embodiment and the fifth embodiment is compared with the existing automatic positioning method of the unmanned aerial vehicle aerial photographing safety distance based on the power tower identification, the measurement efficiency and the data accuracy are obviously improved, and the first embodiment is the best embodiment.
Detection report
Aiming at the problems that the data error is large, the measurement is easily influenced by environmental factors, the manual measurement efficiency is low, the automatic positioning of the safety distance cannot be performed, and the like, the invention provides the unmanned aerial vehicle aerial safety distance automatic positioning method based on the power tower identification.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may be modified or some technical features may be replaced with others, which may not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (10)
1. An unmanned aerial vehicle aerial photographing safety distance automatic positioning method based on electric power tower identification is characterized by comprising the following steps:
s1: the control platform gives an instruction, the unmanned aerial vehicle moves according to a standard inspection path, and the transmission line corridor and the tower body are inspected at a multi-angle in a short distance through the carried high-definition camera;
s2: target identification, integrating YOLO into a single neural network model, and storing and transmitting picture information to a console computer in real time;
s3: YOLOV3 performs target detection and identification, divides an image, and performs identification and classification of objects in the image;
s4: YOLOV3 trains and tests the image dataset, and judges the accuracy of identification by using the accuracy and recall;
s5: introducing a compression activation network into the f-AnoGAN encoder, extracting significance information of the visible light picture, and reducing background noise interference;
s6: and carrying out safe distance positioning by data space acquisition, and carrying out reverse three-dimensional modeling of the electric power facilities based on the point cloud data.
2. The unmanned aerial vehicle aerial photographing safe distance automatic positioning method based on electric power pole and tower recognition according to claim 1, wherein in the step S1, an operator issues an instruction from a management and control platform, the unmanned aerial vehicle quickly reaches a preset pole and tower, moves according to a standard inspection route, accurately positions inspection emphasis, automatically climbs to the other side after one side of the pole and tower is inspected to continue inspection, the unmanned aerial vehicle completes inspection of a first base pole and tower after 10min, automatically flies to the next pole and tower, automatically and safely returns to an unmanned aerial vehicle hangar after all inspection tasks are completed, and energy storage waits for the next flight task.
3. The unmanned aerial vehicle aerial safety distance automatic positioning method based on electric power tower recognition according to claim 2, wherein in the step S1, the unmanned aerial vehicle aerial safety distance automatic positioning method based on electric power tower recognition is characterized in that in the step S1, the unmanned aerial vehicle aerial safety distance automatic positioning method based on electric power tower recognition automatically takes off, automatically patrols and examines, and automatically returns to the home for landing according to preset routes and tasks, and through a carried high-definition camera, a power transmission line corridor and a tower body are patrolled and examined in a multi-angle short-distance mode, a tower position is accurately positioned through three-dimensional modeling, a patrol route is accurately planned, each patrol point is set, and the compiled patrol route is uploaded to an unmanned aerial vehicle.
4. The unmanned aerial vehicle aerial safety distance automatic positioning method based on power tower recognition according to claim 1, wherein in the step S2, target recognition mainly comprises determination of target positions and discrimination of target types, an early target detection and recognition method divides target recognition tasks into target region prediction and target type prediction, YOLO integrates the target region prediction and the target type prediction into a single neural network model, and a whole image training model is used for directly returning out frames and the belonging types of targets at a plurality of positions of an image so as to realize end-to-end target detection and recognition.
5. The unmanned aerial vehicle aerial photographing safety distance automatic positioning method based on the power tower identification according to claim 1, wherein in the step S2, a camera with extremely high definition is arranged on the unmanned aerial vehicle to shoot a patrol scene picture and a video, picture information is stored and transmitted to a console computer in real time, a worker before the computer judges whether the running state of a line is normal according to the transmitted information, if the line is in a normal state, the line can be removed in time, and a visual detection technology is adopted to remove the fault reflected on the surface of the line.
6. The unmanned aerial vehicle aerial safety distance automatic positioning method based on electric power tower recognition according to claim 1, wherein in the step S3, YOLOV3 performs target detection recognition, images are segmented, prediction and classification of boundary boxes are performed, YOLOV3 divides an input image into 5×5 grids, each grid predicts a plurality of boundary boxes and confidence levels thereof at three different scales, and probability information of a plurality of objects belonging to a certain category, the boundary box information is offset, width and height of a central position of an object relative to the grid positions, the confidence levels reflect whether the object is contained or not and the accuracy of the position under the condition that the object is contained, YOLOV3 predicts the score of one object through logistic regression for each boundary box, CNN extracts features in grids, performs series convolution operation on the whole image to obtain a feature map, takes out features in each frame on the feature map to form a high-dimensional feature vector, predicts a plurality of boundary boxes according to the calculated classification errors, confidence levels, and the like, and discriminates the object in the image are classified according to the classification errors and the category probabilities.
7. The unmanned aerial vehicle aerial photographing safety distance automatic positioning method based on power tower recognition according to claim 1, wherein in the step S4, a YOLOV3 training and testing image data set is provided, the training set has 2000 images, wherein 500 images are provided for a power transmission tower, a damper, an insulator and a grading ring respectively, the testing set has 1000 images, 250 images are provided for the power transmission tower, the damper, the insulator and the grading ring respectively, each image in the training set clearly shows a power component, and accuracy of recognition is judged by accuracy and recall, wherein the accuracy is the number of peripheral frames with correct target category marks divided by the number of peripheral frames marked by the mark; the recall rate is the number of peripheral frames of the target class label that are correct divided by the number of peripheral frames of all class targets.
8. The unmanned aerial vehicle aerial safety distance automatic positioning method based on electric power tower recognition according to claim 1, wherein in the step S5, a compression activation network is introduced into an f-AnoGAN encoder, significance information in an image is extracted, then unsupervised learning of an antagonistic network and supervised learning of a two classifier are generated and organically combined, and generalization performance of a model on a large-scale data set is further effectively improved by means of an optimization training strategy based on transfer learning.
9. The unmanned aerial vehicle aerial photographing safety distance automatic positioning method based on the power tower recognition according to claim 1, wherein in the step S5, a compressed activation network based on channel attention is introduced into an encoder of an original f-AnoGAN network, the significance information of visible light pictures is extracted, the background noise interference is reduced, and the tower characteristics are effectively extracted.
10. The automatic positioning method for the aerial photographing safety distance of the unmanned aerial vehicle based on the electric power pole and tower identification is characterized in that in the S6, the data space is acquired for positioning the safety distance, the vertical distance between two points in space is suitable for inquiring the crossing distance of a line, the distance between the lowest point of the line overhang and the ground and the tree, the vertical distance between any position of the pole and the ground, and the height difference between the pole and tower; the three-dimensional coordinates of any point are marked, the spatial position positioning information of insulators of the pole and tower fittings is marked, the vertical distance from the point to a straight line is obtained, the vertical distance between two mutually parallel wires is inquired, the included angle formed by any three vertexes is inquired, the included angle of connecting wires of adjacent poles and towers is inquired, reverse three-dimensional modeling of the electric power facilities is carried out based on point cloud data, electric power asset information is loaded into an information management platform in a model form, attribute information is mounted for the model, various tower type data of a power transmission line and a power distribution line are combined, a panoramic pole and tower is intelligently identified, and the aerial photographing safety distance of the unmanned aerial vehicle is automatically confirmed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310297767.4A CN116301046A (en) | 2023-03-24 | 2023-03-24 | Unmanned aerial vehicle aerial photographing safety distance automatic positioning method based on electric power pole tower identification |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310297767.4A CN116301046A (en) | 2023-03-24 | 2023-03-24 | Unmanned aerial vehicle aerial photographing safety distance automatic positioning method based on electric power pole tower identification |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116301046A true CN116301046A (en) | 2023-06-23 |
Family
ID=86827024
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310297767.4A Pending CN116301046A (en) | 2023-03-24 | 2023-03-24 | Unmanned aerial vehicle aerial photographing safety distance automatic positioning method based on electric power pole tower identification |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116301046A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117498225A (en) * | 2024-01-03 | 2024-02-02 | 山东黄金电力有限公司 | Unmanned aerial vehicle intelligent power line inspection system |
| CN117589177A (en) * | 2024-01-18 | 2024-02-23 | 青岛创新奇智科技集团股份有限公司 | Autonomous navigation method based on industrial large model |
-
2023
- 2023-03-24 CN CN202310297767.4A patent/CN116301046A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117498225A (en) * | 2024-01-03 | 2024-02-02 | 山东黄金电力有限公司 | Unmanned aerial vehicle intelligent power line inspection system |
| CN117498225B (en) * | 2024-01-03 | 2024-03-19 | 山东黄金电力有限公司 | Unmanned aerial vehicle intelligent power line inspection system |
| CN117589177A (en) * | 2024-01-18 | 2024-02-23 | 青岛创新奇智科技集团股份有限公司 | Autonomous navigation method based on industrial large model |
| CN117589177B (en) * | 2024-01-18 | 2024-04-05 | 青岛创新奇智科技集团股份有限公司 | Autonomous navigation method based on industrial large model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108037770B (en) | Unmanned aerial vehicle power transmission line inspection system and method based on artificial intelligence | |
| CN116301046A (en) | Unmanned aerial vehicle aerial photographing safety distance automatic positioning method based on electric power pole tower identification | |
| CN111311597B (en) | Unmanned aerial vehicle inspection method and system for defective insulator | |
| CN111537515A (en) | Iron tower bolt defect display method and system based on three-dimensional live-action model | |
| CN110033453A (en) | Based on the power transmission and transformation line insulator Aerial Images fault detection method for improving YOLOv3 | |
| CN112633535A (en) | Photovoltaic power station intelligent inspection method and system based on unmanned aerial vehicle image | |
| CN112327906A (en) | Intelligent automatic inspection system based on unmanned aerial vehicle | |
| CN111914813A (en) | Power transmission line inspection image naming method and system based on image classification | |
| CN111178148A (en) | Ground target geographic coordinate positioning method based on unmanned aerial vehicle vision system | |
| CN112013830B (en) | Accurate positioning method for inspection image detection defects of unmanned aerial vehicle of power transmission line | |
| CN112684806A (en) | Electric power inspection unmanned aerial vehicle system based on dual obstacle avoidance and intelligent identification | |
| CN113205116A (en) | Automatic extraction and flight path planning method for unmanned aerial vehicle inspection shooting target point of power transmission line | |
| CN115275870B (en) | Inspection system based on high-altitude line maintenance | |
| CN112527010B (en) | Indoor substation unmanned aerial vehicle multi-machine cooperative inspection method based on artificial potential field and particle optimization | |
| CN111561943A (en) | Robot inspection method and system | |
| CN116258980A (en) | Unmanned aerial vehicle distributed photovoltaic power station inspection method based on vision | |
| CN113674355A (en) | Target identification and positioning method based on camera and laser radar | |
| Chehri et al. | Accelerating power grid monitoring with flying robots and artificial intelligence | |
| CN114434456A (en) | Machine room inspection robot and inspection method thereof | |
| CN113885573A (en) | Unmanned aerial vehicle autonomous inspection method based on three-dimensional model and Beidou CORS differential positioning | |
| Pinney et al. | Drone path planning and object detection via QR codes; a surrogate case study for wind turbine inspection | |
| CN112269397A (en) | Unmanned aerial vehicle inspection method based on high-precision positioning and visual tracking | |
| CN115617080B (en) | Unmanned aerial vehicle inspection system and method for transformer substation | |
| CN116740833A (en) | Line inspection and card punching method based on unmanned aerial vehicle | |
| CN114937040B (en) | Train inspection method, device and system for rail transit vehicle section and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication |