CN114723271A - Power transmission project quality detection method and system based on image recognition - Google Patents

Abstract

The invention provides a method and system for detecting the quality of a power transmission project based on image recognition. The method includes autonomous flight after take-off of an unmanned aerial vehicle, and obtaining a point cloud by scanning a power transmission line project; Identify the model, obtain the construction progress corresponding to the current point cloud data, call the preset route corresponding to the current construction progress, and obtain the image of the target point; compare the second identification model of the key points in the image to obtain the current power transmission project quality situation. The invention adopts the UAV as a mobile platform, automatically collects visible light images and point cloud data, and realizes efficient acquisition of information on the construction site; the front-end computing unit integrates AI technology to automatically judge the difference between the engineering entity and the identification model, so as to realize the complete power transmission project. Intelligent management of the process; in addition, based on the analysis and induction of existing data, intelligent prediction of the subsequent construction progress.

Description

A method and system for quality detection of power transmission projects based on image recognition

technical field

The invention relates to the technical field of power transmission project quality and progress analysis, in particular to a power transmission project quality detection method and system based on image recognition.

Background technique

Accurate monitoring of quality and progress during the construction of a power transmission project is crucial to ensuring the safe and efficient completion of the project. At this stage, most of the methods used are manual on-site verification. This method is easily affected by factors such as the environment and personnel subjectivity, and it is difficult to provide real and effective feedback. information, which seriously affects the relevant predictive decision-making.

In recent years, AI technology has appeared more and more in people's daily life. The application of advanced technology such as autonomous driving and smart logistics is subtly improving people's life and work style. As a highly maneuverable mobile platform, the related application technology is improving day by day. Combined with the automatic analysis function of edge computing technology and the high maneuverability of UAVs, it is possible to intelligently analyze and determine the quality and progress of power transmission projects without the need for personnel to arrive at the construction site, so as to achieve the purpose of remote surveying the construction site.

SUMMARY OF THE INVENTION

The invention provides a method and system for detecting the quality of a power transmission project based on image recognition, which is used to solve the problem that the verification result of the project quality is not objective in the construction process of the existing power transmission project.

To achieve the above object, the present invention adopts the following technical solutions:

A first aspect of the present invention provides an image recognition-based power transmission project quality detection method, the detection method comprising the following steps:

After the drone takes off, it flies autonomously, and obtains point clouds by scanning the transmission line project;

Comparing the trained first recognition model based on the point cloud data, obtaining the construction progress corresponding to the current point cloud data, calling the preset route corresponding to the current construction progress, and acquiring the image of the target point;

According to the second identification model comparison of the key points in the image, the quality of the current power transmission project is obtained.

Further, the training process of the first target recognition model is specifically:

Determine the tower type of the transmission line and the milestone nodes in the construction process, collect point cloud data for the line towers under different milestone nodes, and build a point cloud sample library;

Classify and label the point cloud data according to the construction progress to obtain a project progress data set;

Convolutional neural network training is performed on the engineering progress data set to obtain a first identification model based on engineering progress identification.

Further, the classification includes classification based on the time period of the construction period, and the labeling includes labeling of tower types in different construction periods.

Further, the preset route is planned through the line tower point cloud data and the target point position data under different milestone nodes, and the route file is obtained.

Further, the route file, the first identification model and the second identification model are all deployed on the UAV.

Further, the training process of the second target recognition model is specifically:

Determine the tower type of the transmission line and the milestone nodes in the construction process, and collect images of the target points under different milestone nodes to build an image sample library;

Label the key points in the image data to obtain a quality discrimination data set;

Convolutional neural network training is performed on the quality identification data set to obtain a second identification model based on engineering quality identification.

Further, the labeling of the key points in the image data includes labeling of the tower foundation, the tower plate, the hanging points at both ends of the insulator, and the grounding points.

A second aspect of the present invention provides a power transmission project quality detection system based on image recognition, the system includes an unmanned aerial vehicle, a comparison processing unit and a quality discrimination unit,

The drone flies autonomously after taking off, and obtains point clouds by scanning the transmission line project;

The comparison processing unit compares the trained first recognition model based on the point cloud data, obtains the construction progress corresponding to the current point cloud data, calls the preset route corresponding to the current construction progress, and obtains the image of the target point ;

The quality judging unit obtains the quality of the current power transmission project according to the second identification model comparison of the key points in the image.

Further, the system further includes a first model training unit, and the first model training unit is trained based on a convolutional neural network to obtain a first recognition model;

The first model training unit includes:

The first information collection sub-unit defines the tower type of the transmission line and each milestone node in the construction process, and collects point cloud data for the line towers under different milestone nodes, and builds a point cloud sample library;

The first data processing sub-unit classifies and labels the point cloud data according to the construction progress, and obtains a project progress data set;

The first training subunit performs convolutional neural network training on the project progress data set to obtain a first recognition model based on project progress recognition.

Further, the system further includes a second model training unit, and the second model training unit is trained based on a convolutional neural network to obtain a second recognition model;

The second model training unit includes:

The second information collection sub-unit defines the tower type of the transmission line and each milestone node in the construction process, and collects images of target points under different milestone nodes to build an image sample library;

The second data processing sub-unit marks key points in the image data to obtain a quality discrimination data set;

The second training subunit performs convolutional neural network training on the quality identification data set to obtain a second identification model based on engineering quality identification.

The power transmission engineering quality detection system of the second aspect of the present invention can implement the methods in the first aspect and the implementation manners of the first aspect, and achieve the same effect.

The effects provided in the summary of the invention are only the effects of the embodiments, rather than all the effects of the invention. One of the above technical solutions has the following advantages or beneficial effects:

The invention adopts the UAV as a mobile platform, automatically collects visible light images and point cloud data, and realizes efficient acquisition of information on the construction site; the front-end computing unit integrates AI technology to automatically judge the difference between the engineering entity and the identification model, so as to realize the complete power transmission project. Intelligent management of the process; in addition, based on the analysis and induction of existing data, intelligently predict the subsequent construction progress, which is more objective and accurate than the manual detection method in the existing technology, and identifies and alarms irregularities and defects in the construction process. , to bring convenience to the quality inspection personnel of transmission line engineering.

Description of drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the accompanying drawings that need to be used in the description of the embodiments or the prior art. In other words, other drawings can also be obtained based on these drawings without creative labor.

1 is a schematic flowchart of a method embodiment of the present invention;

2 is a schematic flowchart of a specific implementation of the embodiment of the present invention;

Fig. 3 is the training flow schematic diagram of the neural network model in the method embodiment of the present invention;

FIG. 4 is a schematic structural diagram of an embodiment of the system according to the present invention.

Detailed ways

In order to clearly illustrate the technical features of the solution, the present invention will be described in detail below through specific embodiments and in conjunction with the accompanying drawings. The following disclosure provides many different embodiments or examples for implementing different structures of the present invention. In order to simplify the disclosure of the present invention, the components and arrangements of specific examples are described below. Furthermore, the present invention may repeat reference numerals and/or letters in different instances. This repetition is for the purpose of simplicity and clarity and does not in itself indicate a relationship between the various embodiments and/or arrangements discussed. It should be noted that the components illustrated in the figures are not necessarily drawn to scale. Descriptions of well-known components and processing techniques and processes are omitted from the present invention to avoid unnecessarily limiting the present invention.

As shown in FIG. 1 , an embodiment of the present invention provides a method for detecting the quality of a power transmission project based on image recognition. The detecting method includes the following steps:

S1, the drone flies autonomously after taking off, and obtains the point cloud by scanning the transmission line project;

S2, compare the trained first recognition model based on the point cloud data, obtain the construction progress corresponding to the current point cloud data, call the preset route corresponding to the current construction progress, and obtain the image of the target point;

S3, according to the second identification model comparison of the key points in the image, to obtain the quality situation of the current power transmission project.

As shown in Figure 2, the UAV takes off and autonomously flies above the line under construction. The UAV scans the current tower point cloud through the airborne LiDAR, and the airborne AI computing module analyzes and judges the construction progress based on the first recognition model, and obtains Corresponding construction progress, record the AI calculation module to call the route suitable for the current construction progress for inspection and shooting. The airborne AI calculation module analyzes and judges the construction quality based on the second identification model, and outputs the construction progress and quality report. After completing the task The drone returns.

As shown in Figure 3, the training ideas of the first recognition model and the second recognition model are similar, specifically, the tower type of the transmission line and the milestone node in the construction process are defined, and the point cloud and key point image data of the line tower under the milestone node are obtained. Based on the image data, the route planning is carried out according to the point cloud under each milestone node, and the route file is obtained, and the point cloud and image are marked and trained based on the image data, and the model is obtained, and the model and route are deployed on the recording AI module.

Specifically, the training process of the first target recognition model is as follows:

Determine the tower type of the transmission line and each milestone node in the construction process, collect LiDAR point cloud data for the line towers under different milestone nodes, and build a point cloud sample library, which requires good point cloud data accuracy;

Classifying and labeling the point cloud data according to the construction progress, to obtain a project progress data set, the classification includes classification based on the time stage of the construction period, and the labeling includes the labeling of tower types in different construction periods;

Convolutional neural network training is performed on the engineering progress data set to obtain a first identification model based on engineering progress identification. The training of the model is performed using the YOLO V5 algorithm.

The preset route is planned through the line tower point cloud data and target point position data under different milestone nodes, and the route file is obtained.

The route file, the first recognition model and the second recognition model are all deployed on the AI module of the UAV.

The route planning in the route file is based on the construction progress settings. For example, when the tower is just starting to build the foundation, it is only necessary to photograph whether the tower foot is compliant, and the aircraft can zoom and shoot above the route channel without lowering the height to ensure safety. For example, after the completion of the tower, it is necessary to take pictures of the hardware hanging points. At this time, the route strategy becomes a refined inspection, and the hanging points need to be taken on both sides of the line. At this time, the zoom shooting method can also be used to achieve single-camera multi-angle shooting. , to avoid the accident caused by the plane flying into the tower bed.

The AI module relies on the hardware of the AI computing module, such as DJI's Magic Calculation, the picture of the technology picture is the smart box, etc., but the AI computing module with the GPU core can be used for image analysis and calculation.

The training process of the second target recognition model is specifically:

The tower type of the transmission line and the milestone nodes in the construction process are defined, and images are collected for the target points under different milestone nodes to build an image sample library. The images are required to be clear and consistent with the resolution of the actual application;

Marking the key points in the image data to obtain a quality discrimination data set, the marking of the key points in the image data includes marking the tower foundation, the tower plate, the hanging points at both ends of the insulator, and the grounding point;

Convolutional neural network training is performed on the quality identification data set to obtain a second identification model based on engineering quality identification.

As shown in FIG. 4 , an embodiment of the present invention further provides a power transmission project quality detection system based on image recognition, the system includes an unmanned aerial vehicle 1 , a comparison processing unit 2 and a quality discrimination unit 3 .

The unmanned aerial vehicle 1 flies autonomously after taking off, and obtains a point cloud by scanning the transmission line project; the comparison processing unit 2 compares the trained first recognition model based on the point cloud data, and obtains the corresponding data of the current point cloud data. The construction progress is called the corresponding preset route under the current construction progress, and the image of the target point is obtained; the quality judgment unit 3 compares the second identification model of the key points in the image to obtain the quality of the current power transmission project.

The system also includes a first model training unit 4, the first model training unit is trained based on a convolutional neural network to obtain a first recognition model;

The first model training unit 4 includes a first information collection subunit 41 , a first data processing subunit 42 and a first training subunit 43 .

The first information collection subunit 41 specifies the tower type of the transmission line and each milestone node in the construction process, and collects point cloud data for the line towers under different milestone nodes to construct a point cloud sample library; the first data processing subunit 42 Classify and label the point cloud data according to the construction progress to obtain a project progress data set; the first training subunit 43 performs convolutional neural network training on the project progress data set to obtain a first recognition model based on project progress recognition .

The system further includes a second model training unit 5, which is trained based on a convolutional neural network to obtain a second recognition model;

The second model training unit 5 includes a second information collection subunit 51 , a second data processing subunit 52 and a second training subunit 53 .

The second information collection subunit 51 specifies the tower type of the transmission line and each milestone node in the construction process, and collects images of target points under different milestone nodes to construct an image sample library; the second data processing subunit 52 analyzes the image data Mark the key points in the data set to obtain a quality discrimination data set; the second training subunit 53 performs convolutional neural network training on the quality discrimination data set to obtain a second recognition model based on engineering quality recognition.

Although the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, they do not limit the scope of protection of the present invention. Those skilled in the art should understand that on the basis of the technical solutions of the present invention, those skilled in the art do not need to pay creative work. Various modifications or deformations that can be made are still within the protection scope of the present invention.

Claims (10)

1. A power transmission project quality detection method based on image recognition is characterized by comprising the following steps:

the unmanned aerial vehicle flies autonomously after taking off and acquires point cloud through scanning power transmission line engineering;

comparing the point cloud data with the trained first recognition model to obtain a construction progress corresponding to the current point cloud data, calling a corresponding preset route under the current construction progress, and acquiring an image of a target point;

and comparing according to the second identification model of the key points in the image to obtain the quality condition of the current power transmission project.

2. The method for detecting the quality of the power transmission project based on the image recognition as claimed in claim 1, wherein the training process of the first target recognition model is specifically as follows:

determining the tower type of the power transmission line and each milestone node in the construction process, and carrying out point cloud data acquisition on the line towers under different milestone nodes to construct a point cloud sample library;

classifying and labeling the point cloud data according to the construction progress to obtain an engineering progress data set;

and carrying out convolutional neural network training on the project progress data set to obtain a first recognition model based on project progress recognition.

3. The method of claim 2, wherein the classification includes classification based on time period of the project, and the labeling includes labeling of tower types in different projects.

4. The image recognition-based transmission project quality detection method according to claim 2, wherein the preset route is planned through line tower point cloud data and target point position data under different milestone nodes to obtain a route file.

5. The image recognition-based transmission project quality detection method according to claim 4, wherein the route file, the first recognition model and the second recognition model are all deployed on an unmanned aerial vehicle.

6. The method for detecting the quality of the power transmission project based on the image recognition as claimed in claim 1, wherein the training process of the second target recognition model is specifically as follows:

defining the tower type of the power transmission line and each milestone node in the construction process, and carrying out image acquisition on target points under different milestone nodes to construct an image sample library;

marking key points in the image data to obtain a quality judgment data set;

and performing convolutional neural network training on the quality discrimination data set to obtain a second recognition model based on engineering quality recognition.

7. The image recognition-based transmission project quality detection method according to claim 6, wherein the labeling of the key points in the image data includes labeling of a tower foundation, a tower plate, insulator both-end hanging points, and ground wire hanging points.

8. A power transmission project quality detection system based on image recognition is characterized by comprising an unmanned aerial vehicle, a comparison processing unit and a quality judging unit,

the unmanned aerial vehicle flies autonomously after taking off and acquires point cloud through scanning power transmission line engineering;

the comparison processing unit is used for comparing the trained first identification model based on the point cloud data to obtain a construction progress corresponding to the current point cloud data, calling a corresponding preset air route under the current construction progress and acquiring an image of a target point;

and the quality judgment unit obtains the quality condition of the current power transmission project according to the comparison of the second identification models of the key points in the image.

9. The system according to claim 8, wherein the system further comprises a first model training unit, wherein the first model training unit is trained based on a convolutional neural network to obtain a first recognition model;

the first model training unit includes:

the first information acquisition subunit defines the tower type of the power transmission line and each milestone node in the construction process, and carries out point cloud data acquisition on the line towers under different milestone nodes to construct a point cloud sample library;

the first data processing subunit classifies and labels the point cloud data according to the construction progress to obtain an engineering progress data set;

and the first training subunit performs convolutional neural network training on the engineering progress data set to obtain a first recognition model based on engineering progress recognition.

10. The system for detecting the quality of the power transmission project based on the image recognition as claimed in claim 8, wherein the system further comprises a second model training unit, the second model training unit is trained based on a convolutional neural network to obtain a second recognition model;

the second model training unit includes:

the second information acquisition subunit defines the tower type of the power transmission line and each milestone node in the construction process, and acquires images of target points under different milestone nodes to construct an image sample library;

the second data processing subunit labels key points in the image data to obtain a quality judgment data set;

and the second training subunit performs convolutional neural network training on the quality discrimination data set to obtain a second recognition model based on engineering quality recognition.

Images