Power transmission line engineering construction progress identification method and system based on deep learning

Abstract

The invention belongs to the technical field of power transmission line engineering, and particularly relates to a power transmission line engineering construction progress identification method and system based on deep learning, and the method comprises the steps: firstly constructing an improved YOLOv8 model, and then carrying out the feature extraction of a power transmission line construction site image through the improved YOLOv8 model, and finally, carrying out feature matching on the feature vector of the power transmission line construction site image and the building feature vectors under different construction progresses, and obtaining the current construction progress of the power transmission line construction site based on a feature matching result. According to the method, the improved YOLOv8 model is utilized to perform feature extraction on the power transmission line construction site image, the construction progress identification accuracy in a complex environment can be improved, and the power transmission line engineering construction efficiency and quality are further improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of transmission line engineering, and particularly relates to a method and system for identifying the construction progress of transmission line engineering based on deep learning. Background Technique

[0002] In the construction project of transmission line engineering, the identification and tracking of construction progress are crucial. An efficient and accurate construction progress identification method can not only ensure the smooth progress of the project, but also guarantee the construction quality to a certain extent. However, the traditional construction progress identification is mainly manual, which requires specialized personnel to measure and record the daily construction situation, and has the disadvantages of time-consuming and low efficiency, and is prone to cause time waste and cost overrun. With the rapid development of informatization and digitalization, in recent years, digital twin technology has been gradually widely applied in various fields of the construction industry, solving many problems in the industry, and at the same time showing significant advantages in the identification of construction progress in engineering construction.

[0003] The invention patent with the application number 202211499857.3 provides a method and device for monitoring engineering progress based on UAV 3D modeling. This method establishes a complete project 3D model, uses a UAV to collect and draw the project 3D model of the current project status, and compares and analyzes the project 3D model of the current project status with the previously determined project 3D model to determine the current project progress information and project situation. This method has the characteristics of easy image acquisition and high recognition efficiency, and at the same time can more realistically simulate the actual environment, so as to better understand and predict the changes and impacts during the construction process. However, the accuracy of construction progress judgment is poor, and it is applicable to construction projects with a small construction area and a simple construction environment. Therefore, the existing technology has the problem of poor accuracy in identifying construction progress. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for identifying the construction progress of transmission line engineering based on deep learning, which can improve the accuracy of identifying construction progress in complex environments, and further improve the construction efficiency and quality of transmission line engineering, aiming at the above problems existing in the prior art.

[0005] To achieve the above purpose, the technical solution of the present invention is as follows:

[0006] In the first aspect, the present invention provides a method for identifying the construction progress of transmission line engineering based on deep learning, and the method for identifying the construction progress of transmission line engineering includes:

[0007] S1. Construct an improved YOLOv8 model;

[0008] S2. Use the improved YOLOv8 model to extract features from the images of the transmission line construction site to obtain the feature vectors of the images of the transmission line construction site;

[0009] S3. Feature match the feature vector of the power transmission line construction site image with the building feature vectors under different construction progress, and obtain the current construction progress of the power transmission line construction site based on the feature match result.

[0010] In the backbone network of the improved YOLOv8 model, use the sparse sampling strategy and learnable offsets to improve DCNv2 to form C-IDCNv2, and replace the ordinary convolution in the C2f module with C-IDCNv2; the calculation formula of C-IDCNv2 is:

[0011]

[0012] In the above formula, y(p) is the output feature value; p is the sampling point position; N is the total number of sampling points; Ε is the sparse sampling set, containing the positions of all sampling points; M is the number of scales; w k is the weight of the k-th sampling point; x is the input feature value; p k is the original offset; is the learnable offset; is the weight modulation parameter.

[0013] In the backbone network of the improved YOLOv8 model, introduce a BRA attention module before the last conv module.

[0014] The improved YOLOv8 model uses the FloU loss function for object detection, and the expression of the FloU loss function is:

[0015]

[0016] In the above formula, L FloU is the FloU loss function; L IoU is the IOU loss function; L 2loss is the L2 norm loss function; IoU is the intersection over union of the predicted box and the ground truth box; is the upper left coordinate of the ground truth box; is the lower right coordinate of the ground truth box; is the upper left coordinate of the predicted box; is the lower right coordinate of the predicted box; w and h are the width and height of the ground truth box respectively.

[0017] The feature match in S3 is based on the fusion similarity metric method, and the fusion similarity metric method includes: calculating the fusion similarity between the feature vector of the power transmission line construction site image and the building feature vectors under different construction progress, and selecting the construction progress with the highest fusion similarity as the feature match result; the calculation formula of the fusion similarity is:

[0018]

[0019]

[0020] In the above formula, S is the fusion similarity; S C is the cosine similarity; S D is the Euclidean distance similarity; A is the building feature vector corresponding to a certain construction progress; B is the feature vector of the image of the transmission line construction site; ||A|| and ||B|| are the norms of the feature vectors A and B respectively; ||A - B|| is the actual distance between the feature vectors A and B; a i and b i are the i-th element components of the feature vectors A and B respectively.

[0021] In a second aspect, the present invention provides a transmission line project construction progress recognition system based on deep learning. The transmission line project construction progress recognition system includes:

[0022] A model construction module for constructing an improved YOLOv8 model;

[0023] A feature extraction module for using the improved YOLOv8 model to extract features from the image of the transmission line construction site to obtain the feature vector of the image of the transmission line construction site;

[0024] A construction progress recognition module for performing feature matching between the feature vector of the image of the transmission line construction site and the building feature vectors under different construction progress, and obtaining the current construction progress of the transmission line construction site based on the feature matching result.

[0025] The model construction module is used to improve DCNv2 to form C-IDCNv2 by using a sparse sampling strategy and a learnable offset in the backbone network of the improved YOLOv8 model, and replace the ordinary convolution in the C2f module with C-IDCNv2; the calculation formula of C-IDCNv2 is:

[0026]

[0027] In the above formula, y(p) is the output feature value; p is the sampling point position; N is the total number of sampling points; Ε is the sparse sampling set, including the positions of all sampling points; M is the number of scales; w k is the weight of the k-th sampling point; x is the input feature value; p k is the original offset; is the learnable offset; is the weight modulation parameter.

[0028] The feature extraction module is used to introduce a BRA attention module before the last conv module in the backbone network of the improved YOLOv8 model.

[0029] The model construction module is used to perform object detection using the FloU loss function in the improved YOLOv8 model. The expression of the FloU loss function is:

[0030]

[0031] In the above formula, L FloU is the FloU loss function; L IoU is the IOU loss function; L 2loss is the L2 norm loss function; IoU is the intersection over union of the predicted bounding box and the ground truth bounding box; is the upper left coordinate of the ground truth bounding box; is the lower right coordinate of the ground truth bounding box; is the upper left coordinate of the predicted bounding box; is the lower right coordinate of the predicted bounding box; w and h are the width and height of the ground truth bounding box respectively.

[0032] The construction progress recognition module is used to perform feature matching based on the fusion similarity measurement method. The fusion similarity measurement method includes: calculating the fusion similarity between the feature vector of the transmission line construction site image and the feature vectors of buildings under different construction progress, and selecting the construction progress with the highest fusion similarity as the feature matching result; the calculation formula of the fusion similarity is:

[0033]

[0034] In the above formula, S is the fusion similarity; S C is the cosine similarity; S D is the Euclidean distance similarity; A is the feature vector of a building corresponding to a certain construction progress; B is the feature vector of the transmission line construction site image; ||A|| and ||B|| are the norms of the feature vectors A and B respectively; ||A - B|| is the actual distance between the feature vectors A and B; a i and b i are the i-th element components of the feature vectors A and B respectively.

[0035] Compared with the prior art, the beneficial effects of the present invention are:

[0036] 1. The construction progress recognition method for transmission line projects based on deep learning according to the present invention first constructs an improved YOLOv8 model, then uses the improved YOLOv8 model to extract features from the images of the transmission line construction site, obtains the feature vectors of the images of the transmission line construction site, and finally performs feature matching between the feature vectors of the images of the transmission line construction site and the building feature vectors under different construction progress. Based on the feature matching results, the current construction progress of the transmission line construction site is obtained. The present invention uses the improved YOLOv8 model to extract features from the images of the transmission line construction site, which can improve the accuracy of construction progress recognition in complex environments, and further improve the construction efficiency and quality of transmission line projects. Therefore, the present invention can improve the accuracy of construction progress recognition in complex environments, and further improve the construction efficiency and quality of transmission line projects.

[0037] 2. The construction progress recognition method for transmission line projects based on deep learning according to the present invention, in the improved YOLOv8 model, uses a sparse sampling strategy and learnable offsets to improve DCNv2 to form C-IDCNv2, and replaces the ordinary convolution in the backbone network C2f module with C-IDCNv2. By introducing learnable offsets to adjust the shape of the convolution kernel, it can better extract the features of irregular targets, and a modulation mechanism is added to learn the weights of each sampling point. By setting higher weights for key features, the learned offsets of the sampling points are effectively modulated to further improve the model's ability to extract the features of irregular targets. Therefore, the present invention can further improve the performance and efficiency when processing construction scenarios of different scales and complexities.

[0038] 3. The construction progress recognition method for transmission line projects based on deep learning according to the present invention, in the improved YOLOv8 model, introduces a BRA attention module before the last conv module of the backbone network. By using the sparsity operation of the BRA attention module, it reduces the convolution operations of the conv module and the C-IDCNv2 module in the last layer of the backbone network, as well as the processing burden of the SPPF module on unimportant information, thereby saving computational resources and improving computational efficiency. Therefore, the present invention can save computational resources and improve computational efficiency.

[0039] 4. The construction progress recognition method for transmission line projects based on deep learning according to the present invention, in the improved YOLOv8 model, forms a FloU loss function by combining the L2 norm and the IOU loss function, and replaces the traditional CloU loss function with the FloU loss function for object detection. It can avoid problems such as the traditional CloU loss function being insensitive to small target detection and the low training efficiency caused by the complex calculation of the loss function, thereby improving the model's localization accuracy and the convergence speed of the algorithm. Therefore, the present invention can improve the model's localization accuracy and the convergence speed of the algorithm.

[0040] 5. The construction progress recognition method for transmission line projects based on deep learning according to the present invention performs feature matching on the feature vectors of the images at the construction sites of transmission lines and the feature vectors of buildings under different construction progress through a data control platform based on a fusion similarity measurement method, and calculates the fusion similarity using the harmonic mean fusion method. It can effectively balance the two metrics (cosine similarity and Euclidean distance) and avoid the influence of extreme values, and can improve the reliability of the final construction progress recognition. Therefore, the present invention can improve the reliability of the final construction progress recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a flowchart of the construction progress recognition method for transmission line projects according to the present invention.

[0042] Figure 2 It is the network structure of the improved YOLOv8 model constructed according to the present invention.

[0043] Figure 3 It is a structural block diagram of the construction progress recognition system for transmission line projects according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0044] The present invention will be further described in detail below in conjunction with the specific embodiments and the drawings.

[0045] Example 1:

[0046] Refer to Figure 1 , a construction progress recognition method for transmission line projects based on deep learning, which is carried out in the following steps in sequence:

[0047] S1. Based on the construction plan and architectural drawings of the transmission line project, establish a three-dimensional model of the main buildings of the transmission line project through BIM technology; the three-dimensional model is used to simulate the building states at different construction stages of the transmission line project, convert the construction progress features of each building into corresponding feature vectors, and store them in the data control platform classified by project type and construction progress to ensure that each component in the building three-dimensional model contains necessary attribute information such as materials, processes, and construction stages; the main buildings of the transmission line project include pouring projects, tower projects, wire stringing projects, grounding projects, and power equipment installation, etc.; considering factors such as the large amount of building modeling work, rich types of components, and software cost in the transmission line project, Revit software is selected as the BIM modeling tool in this embodiment;

[0048] S2. Use a drone to take pictures of the construction site of the transmission line project to obtain images of the transmission line construction site. The specific steps include: conduct a preliminary survey of the construction site, formulate a reasonable flight path that should cover all key construction areas, including transmission lines, equipment layout, and the positions of operating personnel; adopt a grid flight mode to ensure that there are overlapping images in each area and improve the accuracy of subsequent feature extraction; adjust the flight height, speed, and camera angle of the drone. Set the flight height between 30 - 50m to obtain clear images without affecting construction operations, and set the camera to take pictures at a 45-degree angle downward to capture ground details more comprehensively; select a time period with good lighting conditions for shooting to reduce the impact of shadows on image quality; make the drone automatically capture an image every set distance (such as every 5 meters or every 10 meters) to obtain sufficient image data; after shooting, check the obtained images and perform processing such as cropping, adjusting contrast, and brightness to improve image quality;

[0049] S3. Construct an improved YOLOv8 model with the network structure as shown in Figure 2 . First, since there are various irregular targets at the construction site of the transmission line project, and the shapes, sizes, and arrangements of these targets are different. To further improve the performance and efficiency when processing construction scenarios with different scales and complexities, use the sparse sampling strategy and learnable offsets to improve DCNv2 to form C-IDCNv2, and replace the ordinary convolution in the backbone network C2f module with C-IDCNv2; reduce the number of points processed by introducing the sparse sampling strategy, thereby improving the computational efficiency of the model without affecting the model's ability to extract features; adjust the shape of the convolution kernel by introducing learnable offsets to better extract the features of irregular targets, and add a modulation mechanism to learn the weights of each sampling point, so as to set higher weights for key features and effectively modulate the offsets of the learned sampling points to further improve the model's ability to extract the features of irregular targets; define the sparse sampling set Ε, which contains all selected sampling positions. For the sampling point position p, the C-IDCNv2 calculates its output feature value, and the calculation formula is as follows:

[0050]

[0051] In the above formula, y(p) is the output feature value; Ε is the sparse sampling set, which contains the positions of all sampling points; p is the sampling point position; w k is the weight of the kth sampling point, and k is used to represent the kth eigenvalue in a sampling point index; x is the input feature value, that is, the feature value of the sampling point position p; p k is the original offset; is the learnable offset; is the weight modulation parameter; N is the total number of sampling points; M is the number of scales, and each scale corresponds to targets of different sizes or shapes;

[0052] At the initial stage of training is initialized to a random value, and then according to update where is the updated offset; η is the learning rate; is the gradient of the loss function with respect to At high-resolution scales the adjustment of will be more refined, aiming to capture the minute changes and subtle features in the image; at low-resolution scales the adjustment is relatively rough, mainly used to capture larger feature regions:

[0053] Secondly, a BRA attention module is introduced before the last conv module of the backbone network; by utilizing the sparsity operation of the BRA attention module, the convolution operations of the conv module and the C-IDCNv2 module in the last layer of the backbone network, as well as the processing burden of the SPPF module on unimportant information, are reduced, thereby saving computational volume and improving computational efficiency; at the same time, the position of the BRA module cannot be too forward to avoid excessive low-level features from affecting the enhancement effect of the BRA attention module on key features;

[0054] Finally, since the traditional YOLOv8 model uses a weighted CloU loss function, there are problems such as insensitivity to small target detection and low training efficiency caused by the complex calculation of the loss function. To improve the model's localization accuracy and the algorithm's convergence speed, a FloU loss function is formed by combining the L2 norm and the IOU loss function, and the FloU loss function is used to replace the traditional CloU loss function for object detection; the expression of the FloU loss function is:

[0055]

[0056] In the above formula, L FloU is the FloU loss function; L IoU is the IOU loss function; L 2loss is the L2 norm loss function; IoU is the intersection over union of the predicted box and the ground truth box; is the upper left coordinate of the ground truth box; is the lower right coordinate of the ground truth box; is the upper left coordinate of the predicted box; is the lower right coordinate of the predicted box; w and h are the width and height of the ground truth box respectively;

[0057] S4. Input the image of the transmission line construction site finally obtained in S2 into the improved YOLOv8 model for feature extraction, and input the feature vector of the image of the transmission line construction site extracted by the improved YOLOv8 model into the data management and control platform;

[0058] S5. Through the data management and control platform, perform feature matching on the feature vector of the image of the transmission line construction site and the building feature vectors under different construction progress based on the fusion similarity measurement method, and obtain the current construction progress of the transmission line construction site based on the feature matching result; the fusion similarity measurement method includes: for each image of the transmission line construction site, calculate the fusion similarity between its feature vector and the building feature vectors under all construction progress, and select the construction progress with the highest fusion similarity as the feature matching result; use the harmonic mean fusion method to calculate the fusion similarity, and the harmonic mean is more sensitive to smaller values, which can effectively balance the two metrics (cosine similarity and Euclidean distance) and avoid the influence of extreme values; the calculation formula for the fusion similarity is:

[0059]

[0060] In the above formula, S is the fusion similarity; S C is the cosine similarity; S D is the Euclidean distance similarity; A is the building feature vector corresponding to a certain construction progress; B is the feature vector of the image of the transmission line construction site; ||A|| and ||B|| are the norms of the feature vectors A and B respectively; ||A - B|| is the actual distance between the feature vectors A and B; a i , b i are the i-th element components of the feature vectors A and B respectively;

[0061] S6. Combine the feature matching result with the BIM three-dimensional model for visual display and generate a construction progress report for project management and monitoring.

[0062] Performance verification:

[0063] To verify the effectiveness and superiority of the improved YOLOv8 model proposed in the present invention in the real-time image recognition of the transmission line construction site, 3 image recognition models are set for comparative analysis, namely: Faster-RCNN, YOLOv7, YOLOv8, and four indicators of accuracy, precision, recall rate, and average time consumption are used to evaluate the image recognition performance of the models for the transmission line construction site. The performance comparison results are shown in Table 1:

[0064] Table 1 Performance comparison results

[0065] Model Accuracy / % Precision / % Recall / % Average time consumption / s Faster-RCNN 81.73 82.37 74.26 0.52 YOLOv7 84.47 86.26 77.35 0.43 YOLOv8 88.91 89.48 83.74 0.39 Improved YOLOv8 90.89 91.36 88.78 0.32

[0066] As can be seen from Table 1, compared with Faster-RCNN, YOLOv7, and YOLOv8, the improved YOLOv8 model proposed in the present invention has increased the accuracy of image recognition by 11.21%, 7.6%, and 2.23% respectively; the precision has increased by 10.91%, 5.91%, and 2.1% respectively; the recall rate has increased by 19.55%, 14.78%, and 6.02% respectively; and the average time consumption has decreased by 0.2s, 0.11s, and 0.07s respectively. It can be seen that the improved YOLOv8 model proposed in the present invention has higher accuracy and recall rate, and at the same time, the time consumed for training and testing is shorter.

[0067] Example 2:

[0068] See Figure 3 , a construction progress recognition system for transmission line projects based on deep learning, including a model construction module, a feature extraction module, and a construction progress recognition module; the model construction module is used to construct an improved YOLOv8 model; specifically, in the backbone network of the improved YOLOv8 model, the DCNv2 is improved by using a sparse sampling strategy and a learnable offset to form C-IDCNv2, and the C-IDCNv2 is used to replace the ordinary convolution in the C2f module; a BRA attention module is introduced before the last conv module in the backbone network; the FloU loss function is used for object detection; the calculation formula of the C-IDCNv2 is:

[0069]

[0070] In the above formula, y(p) is the output feature value; p is the sampling point position; N is the total number of sampling points; Ε is the sparse sampling set, which contains the positions of all sampling points; M is the number of scales, and each scale corresponds to targets of different sizes or shapes; w k is the weight of the kth sampling point; x is the input feature value; p k is the original offset; is the learnable offset; is the weight modulation parameter;

[0071] The expression of the FloU loss function is:

[0072]

[0073] In the above formula, L FloU is the FloU loss function; L IoU is the IOU loss function; L 2loss is the L2 norm loss function; IoU is the intersection over union of the predicted box and the ground truth box; is the upper left coordinate of the ground truth box; is the lower right coordinate of the ground truth box; is the upper left coordinate of the predicted box; is the lower right corner coordinate of the prediction box; w and h are the width and height of the ground truth box respectively;

[0074] The feature extraction module is used to extract features from the image of the transmission line construction site by using the improved YOLOv8 model to obtain the feature vector of the image of the transmission line construction site; the construction progress recognition module is used to perform feature matching on the feature vector of the image of the transmission line construction site and the building feature vectors under different construction progress based on the fusion similarity measurement method, and obtain the current construction progress of the transmission line construction site based on the feature matching result; the fusion similarity measurement method includes: calculating the fusion similarity between the feature vector of the image of the transmission line construction site and the building feature vectors under different construction progress, and selecting the construction progress with the highest fusion similarity as the feature matching result; the calculation formula of the fusion similarity is:

[0075]

[0076] In the above formula, S is the fusion similarity; S C is the cosine similarity; S D is the Euclidean distance similarity; A is the building feature vector corresponding to a certain construction progress; B is the feature vector of the image of the transmission line construction site; ||A|| and ||B|| are the norms of the feature vectors A and B respectively; ||A - B|| is the actual distance between the feature vectors A and B; a i and b i are the i-th element components of the feature vectors A and B respectively.

[0077] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0078] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowcharts and / or block diagrams can be implemented by computer program instructions, and the combination of the processes and / or blocks in the flowcharts and / or block diagrams. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate for implementation in the process Figure 1one process or multiple processes and / or boxes Figure 1 means for the functions specified in one box or multiple boxes.

[0079] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured article including an instruction device that implements the process Figure 1 one process or multiple processes and / or boxes Figure 1 the functions specified in one box or multiple boxes.

[0080] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one process or multiple processes and / or boxes Figure 1 one box or multiple boxes.

[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: modifications or equivalent replacements can still be made to the specific implementation manners of the present invention, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention shall be covered by the protection scope of the claims of the present invention.

Claims

1. A method for identifying the construction progress of transmission line projects based on deep learning, characterized in that: The method for identifying the construction progress of transmission line projects includes: S1. Construct an improved YOLOv8 model; S2. Use the improved YOLOv8 model to extract features from the images of the transmission line construction site, and obtain the feature vectors of the images of the transmission line construction site; S3. Perform feature matching between the feature vectors of the images of the transmission line construction site and the building feature vectors under different construction progress, and obtain the current construction progress of the transmission line construction site based on the feature matching results.

2. The method for identifying the construction progress of transmission line projects based on deep learning according to claim 1, characterized in that: In the backbone network of the improved YOLOv8 model, use the sparse sampling strategy and learnable offsets to improve DCNv2 to form C-IDCNv2, and replace the ordinary convolution in the C2f module with C-IDCNv2; the calculation formula of C-IDCNv2 is: In the above formula, y(p) is the output eigenvalue; p is the position of the sampling point; N is the total number of sampling points; Ε is the sparse sampling set, including the positions of all sampling points; M is the number of scales; wk is the weight of the k-th sampling point; x is the input eigenvalue; pk is the original offset; is the learnable offset; is the weight modulation parameter.

3. The method for identifying the construction progress of transmission line projects based on deep learning according to claim 1, characterized in that: In the backbone network of the improved YOLOv8 model, a BRA attention module is introduced before the last conv module.

4. The method for identifying the construction progress of transmission line projects based on deep learning according to claim 1, characterized in that: The improved YOLOv8 model uses the FloU loss function for object detection, and the expression of the FloU loss function is: In the above formula, L FloU is the FloU loss function; L IoU is the IOU loss function; L 2loss is the L2 norm loss function; IoU is the intersection over union of the predicted bounding box and the ground truth bounding box; is the top - left coordinate of the ground truth bounding box; is the bottom - right coordinate of the ground truth bounding box; is the top - left coordinate of the predicted bounding box; is the bottom - right coordinate of the predicted bounding box; w and h are the width and height of the ground truth bounding box respectively.

5. The method for identifying the construction progress of transmission line projects based on deep learning according to claim 1, characterized in that: The feature matching in S3 is based on the fusion similarity measurement method, and the fusion similarity measurement method includes: calculating the fusion similarity between the feature vectors of the images of the transmission line construction site and the building feature vectors under different construction progress, and selecting the construction progress with the highest fusion similarity as the feature matching result; the calculation formula of the fusion similarity is: In the above formula, S is the fusion similarity; S C is the cosine similarity; S D is the Euclidean distance similarity; A is the building feature vector corresponding to a certain construction progress; B is the feature vector of the power transmission line construction site image; ||A|| and ||B|| are the norms of the feature vectors A and B respectively; ||A - B|| is the actual distance between the feature vectors A and B; ai and bi are the i-th element components of the feature vectors A and B respectively.

6. A system for identifying the construction progress of transmission line projects based on deep learning, characterized in that: The system for identifying the construction progress of transmission line projects includes: A model construction module for constructing an improved YOLOv8 model; A feature extraction module for using the improved YOLOv8 model to extract features from the images of the transmission line construction site, and obtaining the feature vectors of the images of the transmission line construction site; A construction progress identification module for performing feature matching between the feature vectors of the images of the transmission line construction site and the building feature vectors under different construction progress, and obtaining the current construction progress of the transmission line construction site based on the feature matching results.

7. The system for identifying the construction progress of transmission line projects based on deep learning according to claim 6, characterized in that: The model construction module is used to use the sparse sampling strategy and learnable offsets to improve DCNv2 to form C-IDCNv2 in the backbone network of the improved YOLOv8 model, and replace the ordinary convolution in the C2f module with C-IDCNv2; the calculation formula of C-IDCNv2 is: In the above formula, y(p) is the output eigenvalue; p is the position of the sampling point; N is the total number of sampling points; Ε is the sparse sampling set, which contains the positions of all sampling points; M is the number of scales; w k is the weight of the k-th sampling point; x is the input eigenvalue; pk is the original offset; is the learnable offset; is the weight modulation parameter.

8. The construction progress recognition system for transmission line projects based on deep learning according to claim 6, characterized in that: The feature extraction module is used to introduce a BRA attention module before the last conv module in the backbone network of the improved YOLOv8 model.

9. The construction progress recognition system for transmission line projects based on deep learning according to claim 6, characterized in that: The model construction module is used to perform object detection using the FloU loss function in the improved YOLOv8 model, and the expression of the FloU loss function is: In the above formula, L FloU is the FloU loss function; L IoU is the IOU loss function; L 2loss is the L2 norm loss function; IoU is the intersection over union of the predicted bounding box and the ground truth bounding box; is the top-left coordinate of the ground truth bounding box; is the bottom-right coordinate of the ground truth bounding box; is the top-left coordinate of the predicted bounding box; is the bottom-right coordinate of the predicted bounding box; w and h are the width and height of the ground truth bounding box respectively.

10. The construction progress recognition system for transmission line projects based on deep learning according to claim 6, characterized in that: The construction progress recognition module is used to perform feature matching based on the fusion similarity measurement method, and the fusion similarity measurement method includes: calculating the fusion similarity between the feature vector of the transmission line construction site image and the building feature vectors under different construction progress, and selecting the construction progress with the highest fusion similarity as the feature matching result; the calculation formula of the fusion similarity is: In the above formula, S is the fusion similarity; S C is the cosine similarity; S D is the Euclidean distance similarity; A is the building feature vector corresponding to a certain construction progress; B is the feature vector of the image of the transmission line construction site; ||A|| and ||B|| are the norms of the feature vectors A and B respectively; ||A - B|| is the actual distance between the feature vectors A and B; ai and bi are the i-th element components of the feature vectors A and B respectively.

Images