Target detection method for electric power infrastructure scene based on Yolov8 improvement

Abstract

The invention provides a target detection method for an electric power infrastructure scene based on Yolov8 improvement, and belongs to the technical field of artificial intelligence and target detection. According to the invention, the artificial intelligence technology is integrated into a power operation scene, so that potential safety hazards in the operation process can be prevented; according to the method, a C2f network module in a Yolov8 model framework is replaced by constructing a module combining ConvFormer and a convolutional gated linear unit Convolutional Gated Line Unit network, so that the calculation parameters of the model can be effectively reduced on the premise of ensuring the detection accuracy of the algorithm; and meanwhile, a shared convolution module Feature Shared Conv is constructed to replace an SPPF network module in a Yolov8 model framework, so that finer features in the image can be captured, and the detection accuracy of the model in a complex scene is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of artificial intelligence and target detection, and specifically relates to a target detection method for an electric power infrastructure scene based on Yolov8 improvement. Background Art

[0002] With the expansion of infrastructure projects, UHV, EHV and other power construction projects have entered a new peak. The new requirements for power construction and safety management have brought many challenges to infrastructure safety management. The construction site is more complex, the risk of safety operations and the number of construction personnel have increased significantly, and the supervision is insufficient, making the safety issues at the infrastructure site increasingly prominent.

[0003] In the process of infrastructure construction, in order to timely discover and prevent potential safety risks, employees are usually given professional safety training to improve their operating skills and safety awareness, and manual on-site or video monitoring is used to ensure the safety of operations. However, this manual method not only consumes a lot of human resources, but also easily leads to omissions or misdetection during the monitoring process. Therefore, in order to strengthen the safety control of infrastructure sites and ensure the quality and safety of construction personnel and projects, it is necessary to introduce an efficient and intelligent detection technology. Accurately identifying the personnel, safety belts, and operating tools at infrastructure sites is a crucial task in the research.

[0004] With the development of deep learning technology, it can be well applied to the detection field by training neural networks to learn the internal structure of image data and extract deeper features of targets. Therefore, using deep learning methods to detect targets in complex infrastructure scenes has important application value and practical significance. At present, the most commonly used target detection algorithms, such as Yolo series, Faster-RCNN, SSD, etc., have achieved good performance and speed in target detection tasks. However, these methods have high requirements on device computing power and are difficult to deploy in embedded devices. At the same time, the detection accuracy is not high and the robustness is poor in complex scenes such as lighting, occlusion, fast motion, and background interference. For some lightweight models such as the ShuffleNet series, these models reduce the dependence on hardware, but their model accuracy is also greatly reduced. Therefore, in practical applications, how to reduce model calculation parameters and improve calculation speed without sacrificing the detection performance of the model is a problem that needs to be solved urgently.

[0005] The deep integration of artificial intelligence technology into power operation scenarios aims to improve the professional quality of operators, standardize their work behavior, and prevent potential safety hazards during the operation process. The application of this technology not only helps to build a more standardized and standardized power operation environment, but also promotes the development of power operation scenarios towards intelligent and refined management. Summary of the invention

[0006] In view of the above problems existing in the prior art, the purpose of the present invention is to provide a target detection method for power infrastructure scenes based on Yolov8 improvement, which can capture more subtle features in the image and improve the detection accuracy of the model. At the same time, by optimizing the network model and reducing the calculation parameters of the model, the problems raised in the above background technology are solved; specifically, the following steps are included: Step 1: Determine the detection target and collect the data set required for the research. Collect image data through monitoring video at a real power infrastructure site. Collect data under different lighting, weather, time periods and other conditions to increase data diversity; Step 2: Label and convert the collected data to obtain the processed data set, which is divided into experimental training set, test set and validation set according to the preset ratio; Step 3: Design a module that combines ConvFormer and Convolutional Gated Linear Unit (CGLU) network to replace the C2f network module in the Yolov8 model framework. This effectively reduces the model's calculation parameters while ensuring the accuracy of the algorithm detection. Step 4: Design a shared convolution module (Feature Shared Conv, FSConv) to replace the SPPF network module in the algorithm framework. This module uses convolution operations for feature extraction, which can capture more detailed features in the image and improve the detection accuracy of the model in complex scenes; Step 5: Use the dataset to train the improved model, obtain the optimal target detection model based on the performance evaluation indicators, detect the images in the test set, and complete the target detection task.

[0007] Furthermore, the specific process of step 1 is as follows: For complex scenes of power infrastructure, we identify the targets that need to be detected, such as workers, helmets, safety belts, excavators, fire extinguishers, roadblocks, cranes, etc. We obtain images by extracting frames from surveillance videos. To ensure the validity of the images, we reduce the frame extraction interval during normal working hours.

[0008] Furthermore, the specific process of step 2 is as follows: Remove blurry, overexposed and other low-quality images. Determine the detection target according to the needs of the application scenario, use the Labelme tool to annotate the image data, and convert the annotated data into the Yolo format required for model training. Divide the processed data set into training set, test set and validation set in a ratio of 8:1:1.

[0009] Furthermore, the specific process of step 3 is as follows: Based on the ConvFormer and CGLU network models, a separable convolution module C2f_CFGLU is designed to replace all C2f modules in the Yolov8 model. The core feature of ConvFormer is to use separable convolution to mix feature vector tokens. This convolution method consists of two steps: deep convolution and point-by-point convolution. Deep convolution is responsible for capturing the spatial features of the image, while point-by-point convolution is responsible for capturing the relationship between channels. CGLU is a channel mixer that combines channel attention and local feature extraction, which can effectively capture local features, thereby enhancing the model's understanding and expression of image details. Based on the ConvFormer structure (i.e., input embedding, feature mixer Token Mixer, and multi-layer perceptron MLP, etc.), CGLU is used to replace the multi-layer perceptron MPL in the ConvFormer network structure to obtain the separable convolution module C2f_CFGLU. This module can effectively enhance the robustness of the model, making the model better adaptable to different application scenarios. In addition, the module has lower parameters and computational complexity, making it easier to train and deploy.

[0010] Furthermore, the specific process of step 4 is as follows: A shared convolution module FSConv is designed. This module constructs a feature pyramid. First, it uses a 1x1 convolution layer to process the input feature map and extract the main features. Then, according to different expansion rates, it uses a shared convolution layer to process feature maps of different scales to capture richer multi-scale information. Finally, all feature maps are connected and output. This module is used to replace the SPPF module in the Yolov8 model. Compared with the pooling operation of SPPF, the convolution operation has higher flexibility and expression ability in feature extraction, and can better capture the details in the image. It further improves the model's multi-scale feature representation ability and enhances the model's ability to recognize targets in complex scenes.

[0011] Furthermore, the specific process of step 5 is as follows: The performance evaluation indicators specifically include precision, recall, mean average precision (mAP50), and mean average precision (mAP50-95).

[0012] Performance evaluation indicators make it easy to screen out the optimal target detection model to complete the target detection task.

[0013] Furthermore, the expression of the precision Precision is: ; It refers to the ratio of positive samples (TP) correctly detected by the classifier to all samples marked as positive by the classifier (TP + FP).

[0014] The expression of recall rate Recall is: ; Refers to the ratio of positive samples (TP) correctly detected by the classifier to all true positive samples (TP + FN).

[0015] The expressions of mAP50 and mAP50-95 are: ; ; Where IoU is the overlap between the predicted bounding box and the true bounding box. AP is the average precision calculated at different recall levels, usually the area under the PR curve. N is the total number of categories, and M is the number of IoU thresholds. mAP50 refers to the mean average precision when the IoU threshold is 0.5, and mAP50-95 calculates the mean average precision when the IoU threshold ranges from 0.5 to 0.95 (usually increasing in steps of 0.05).

[0016] The performance evaluation index of the detection model can be calculated through the above formula.

[0017] By adopting the above technology, compared with the prior art, the beneficial effects of the present invention are as follows: The present invention constructs a module combining ConvFormer and CGLU networks to replace the C2f network module in the Yolov8 model framework, which can effectively reduce the calculation parameters of the model while ensuring the detection accuracy of the algorithm. At the same time, by constructing the FSConv network module to replace the SPPF network module in the Yolov8 model framework, it can capture more detailed features in the image and improve the detection accuracy of the model in complex scenes. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is a flow chart of the method of the present invention; Figure 2 This is a diagram of the improved Yolov8 network structure in an embodiment of the present invention; Figure 3 : is a module diagram of a separable convolution module C2F_CFGLU in an embodiment of the present invention; Figure 4 It is a diagram of the MLP module in the separable convolution module C2F_CFGLU in an embodiment of the present invention; Figure 5 FSConv module diagram in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments of the specification. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0020] On the contrary, the present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention as defined by the claims. Further, in order to make the public have a better understanding of the present invention, some specific details are described in detail in the detailed description of the present invention below. Those skilled in the art can fully understand the present invention without the description of these details.

[0021] like Figure 1 As shown, a target detection method for a power infrastructure scene based on Yolov8, the specific steps include: Step 1: Use a camera to shoot a real power infrastructure scene, and obtain image data by extracting frames from the captured video. In order to enhance the validity of the data, the frame extraction interval during normal working hours is reduced.

[0022] Step 2: Filter, annotate, and convert the image data to generate a standard data set that conforms to the Yolo format; divide the standard data set into training set, test set, and validation set in a ratio of 8:1:1 for subsequent experiments.

[0023] Step 3: If Figure 2As shown in the figure, based on ConvFormer and CGLU networks, a C2F_CFGLU network is designed to replace all C2f modules in the Yolov8 model. The backbone network of the improved Yolov8 model includes sequentially connected convolutional layer 01 (Conv01) and convolutional layer 02 (Conv02), separable convolutional module 01 (C2F_CFGLU01), convolutional layer 03 (Conv03), separable convolutional module 02 (C2F_CFGLU02), convolutional layer 04 (Conv04), separable convolutional module 03 (C2F_CFGLU03), convolutional layer 05 (Conv05), and separable convolutional module 04 (C2F_CFGLU04). The neck network includes upsampling layer 01 (Upsample01), connection module 01 (Concat01), separable convolution module 05 (C2F_CFGLU05), upsampling layer 02 (Upsample02), connection module 02 (Concat02), separable convolution module 06 (C2F_CFGLU06), convolution layer 06 (Conv06), connection module 03 (Concat03), separable convolution module 07 (C2F_CFGLU07), convolution layer 07 (Conv07), connection module 04 (Concat04), and separable convolution module 08 (C2F_CFGLU08), which are connected in sequence.

[0024] Specifically, the separable convolution module C2F_CFGLU designed in step 3 has the following structure: Figure 3 As shown in the figure, the specific feature extraction process of the separable convolution module C2F_CFGLU is as follows: First, the input is normalized by the normalization layer to standardize the feature values, and the normalized feature map is passed through the SepConv mixer layer to fuse different features. This layer uses separable convolution to achieve feature mixing, decomposing the convolution operation into depth convolution and point-by-point convolution, thereby reducing the computational complexity and parameter quantity of the model. The output of the SepConv mixer is then processed by the scaling layer Scale and the regularization layer Dropout to prevent overfitting, and is added to the residual path. The output after the residual connection is processed by the multi-layer perceptron MLP, where the module diagram of the multi-layer perceptron MLP is shown in Figure 4 As shown in the figure. It processes the input through CGLU, which combines convolution and attention mechanisms to enhance the model's expressiveness. Convolution can effectively capture local features, while the attention mechanism can help the model focus on the most important parts of the input features. Finally, it passes through a regularization layer Dropout and a scaling layer Scale as the final feature map output. Compared with the original C2f module, the optimized module reduces the number of parameters and calculations, making it easier to deploy on hardware devices.

[0025] Step 4: Figure 2 As shown in FIG. 5 , a shared convolution module FSConv is designed to replace the SPPF module in the Yolov8 model, that is, the module is connected behind the separable convolution module C2F_CFGLU04 of the backbone network.

[0026] Furthermore, the FSConv module designed in step 4 has a specific structure as follows Figure 5 As shown in the figure, the specific operation process of the FSConv module is as follows: First, the input features are passed through a 1x1 convolution layer to reduce the number of channels and the amount of computation of the input features. Then a convolution kernel and the dilation rates of different atrous convolutions to be used are defined (the default is 1, 3, 5), and the input features are convolved using the same convolution kernel by looping through different dilation rates. Although this approach performs multiple convolution operations, all operations use the same convolution kernel weights, which helps to reduce the number of model parameters while maintaining the uniformity of the convolution operation, allowing the network to process features of different scales in the same way. Finally, after all convolution operations, the resulting feature maps are fused, and the number of channels is further adjusted to generate the final output features. Compared with the pooling operation in the original SPPF module, this module can reduce computing resources while also providing more refined image features for subsequent detection processes, thereby improving the detection accuracy of the model.

[0027] Step 5: The prepared standard data set, including the training set (8000 images), the test set (1000 images), and the validation set (1000 images), is used to train the improved Yolov8 model. The experimental hardware and software configuration is shown in Table 1. The Pytorch framework is used, the SGD optimizer is used, the initial learning rate is 0.001, the momentum factor is 0.973, the number of iterations is 300 rounds, and the input image size is scaled to 640×640. Based on the above experimental environment, the optimal object detection model is selected according to the performance evaluation indicators.

[0028] Table 1 Software and Hardware Configuration Table

[0029] In step 5, the performance evaluation indicators include precision, recall, mAP50, and mAP50-95.

[0030] The above settings make it easy to screen out the optimal target detection model through performance evaluation indicators to complete the target detection task.

[0031] Specifically, the expression of precision is: ; It refers to the ratio of positive samples (TP) correctly detected by the classifier to all samples marked as positive by the classifier (TP + FP).

[0032] The expression of recall rate Recall is: ; Refers to the ratio of positive samples (TP) correctly detected by the classifier to all true positive samples (TP + FN). The expressions of mAP50 and mAP50-95 are: ; ; Where IoU is the overlap between the predicted bounding box and the true bounding box. AP is the average precision calculated at different recall levels, usually the area under the PR curve. N is the total number of categories, and M is the number of IoU thresholds. mAP50 refers to the mean average precision when the IoU threshold is 0.5, and mAP50-95 calculates the mean average precision when the IoU threshold ranges from 0.5 to 0.95 (usually increasing in steps of 0.05).

[0033] In order to further verify the effects of the improved Yolov8 algorithm of the present invention and the existing Yolov8 algorithm, an ablation experiment is performed to compare the two algorithms. The experimental results are shown in Table 2: Table 2. Comparison of ablation experiment results

[0034] Experiments show that the improved Yolov8 algorithm in this embodiment has better performance in accuracy, recall, and mAP than the Yolov8 algorithm in target detection for complex infrastructure scenes. The indicator is 32.39%, and the algorithm in this embodiment The highest value is 35.07%, which is 2.6 percentage points higher than Yolov8. In the comparison of accuracy, the accuracy of the algorithm in this embodiment reaches 87.15%, which is 3.3 percentage points higher than Yolov8. Therefore, the overall performance of the algorithm in this embodiment is better.

[0035] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A target detection method for power infrastructure scenes based on Yolov8, characterized in that: The steps include: Step 1: Determine the detection target and collect the required data set: In the power infrastructure scenario, collect the corresponding image data through monitoring video; Step 2: Screen, annotate, and convert the collected image data to obtain a standard data set that conforms to the Yolo format, and divide it into a training set, a test set, and a validation set according to the set ratio; Step 3: Combine ConvFormer with the convolutional gated linear unit network to build a separable convolution module C2f_CFGLU, replacing the C2f network module in the Yolov8 model framework, so as to reduce the calculation parameters of the model while ensuring the accuracy of algorithm detection; Step 4, construct a shared convolution module Feature Shared Conv to replace the SPPF network module in the Yolov8 model framework; the shared convolution module Feature Shared Conv uses convolution operation to extract features; Step 5: Use the training set to train the improved Yolov8 model, obtain the optimal target detection model based on the performance evaluation index, detect the images in the test set, and complete the target detection task.

2. According to claim 1, a target detection method for electric power infrastructure scene based on Yolov8 improvement is characterized in that: The specific process of step 1 is as follows: For power infrastructure scenarios, determine the targets that need to be detected; obtain images by extracting frames from surveillance videos. To ensure the validity of the images, reduce the interval between frame extractions.

3. According to claim 2, a target detection method for electric power infrastructure scene based on Yolov8 improvement is characterized in that: The specific process of step 2 is as follows: Screen the collected images and remove images with defects; determine the detection target according to the needs of the application scenario, use the Labelme tool to annotate the image data, and convert the annotated data into the Yolo format required for model training; divide the processed data set into training set, test set, and validation set according to the set ratio.

4. According to claim 3, a target detection method based on Yolov8 improved power infrastructure scene is characterized in that: The specific process of step 3 is as follows: Based on the ConvFormer and Convolutional Gated Linear Unit network models, on the basis of the ConvFormer network structure, the convolutional gated linear unit Convolutional Gated Linear Unit is used to replace the multi-layer perceptron MPL in the ConvFormer network structure to obtain the separable convolution module C2f_CFGLU, and the separable convolution module C2f_CFGLU is used to replace all C2f modules in the Yolov8 model.

5. According to claim 4, a target detection method for electric power infrastructure scene based on Yolov8 improvement is characterized in that: In step 4, a shared convolution module Feature SharedConv is constructed by constructing a feature pyramid; the specific working process of the shared convolution module Feature SharedConv is as follows: 1) Use a 1x1 convolutional layer to process the input feature map and extract the required features; 2) According to different expansion rates, shared convolutional layers are used to process feature maps of different scales to capture multi-scale information; 3) Connect all feature maps and output them.

6. According to claim 5, a target detection method based on Yolov8 improved power infrastructure scene is characterized in that: In step 5, The performance evaluation indicators include precision, recall, mAP50 and mAP50-95; The expression of the precision Precision is: ； It represents the ratio of the positive samples TP correctly detected by the classifier to all the positive samples TP+FP marked by the classifier; The expression of recall rate Recall is: ； It represents the ratio of the positive sample TP correctly detected by the classifier to all true positive samples TP+FN; The expressions of mAP50 and mAP50-95 are: ； ； Where IoU is the overlap between the predicted bounding box and the true bounding box, AP is the average precision calculated at different recall levels, N is the total number of categories, M is the number of IoU thresholds, mAP50 refers to the mean average precision when the IoU threshold is 0.5, and mAP50-95 calculates the mean average precision when the IoU threshold ranges from 0.5 to 0.95.

Images