An ATP-YOLOv8-based transmission line passage hidden danger remote sensing image target detection algorithm

Abstract

The application discloses a kind of hidden danger remote sensing image target detection algorithm of transmission line channel based on ATP-YOLOv8, and specific steps are as follows: unmanned aerial vehicle aerial acquisition transmission line channel hidden danger remote sensing dataset is divided into training set, verification set and test set according to the proportion of 8:1:1;Design MAM network to replace the C2f module of original model, expand the context receptive field of model;Build a kind of mixed attention MHA module, adjust the weight of features in channel and space;Design a TBD module more suitable for remote sensing targets, reduce the feature loss of transmission line hazard size target in the down-sampling process.The application significantly enhances the ability of the model to capture multi-scale target features from image feature maps by designing MAM network, MHA module and TBD module.These improvements enable the model to have stronger perception capabilities when detecting transmission line channel hidden danger remote sensing images in complex backgrounds.

Description

Technical Field

[0001] The present invention relates to the technical field of combining deep learning target detection with remote sensing images of power transmission lines, and specifically to an ATP-YOLOv8-based target detection algorithm for remote sensing images of hidden dangers in power transmission line channels. Background Art

[0002] Transmission channel hazards primarily include bird nests and foreign objects on conductors (such as balloons, kites, and garbage). These objects can cause short circuits, ground faults, and other faults in transmission lines, leading to power outages and impacting the reliability and stability of power supply. In severe cases, they can even cause major accidents, resulting in the loss of critical facilities and personnel. Transmission channel inspections face complex scenarios and numerous interfering factors, such as weather changes, background conditions, equipment performance, and lighting, which significantly impact image quality. Furthermore, factors such as the different models and ages of towers and equipment types can also lead to variations in target types and shapes. Therefore, detecting transmission channel hazards has become a top priority for power line inspections. In recent years, the use of drones to capture aerial images or videos for automated identification and detection of transmission channel hazards has become increasingly common.

[0003] Target detection in remote sensing images is fundamental to remote sensing image analysis and understanding, and can determine the location, range, and category of targets contained in remote sensing images. With the advancement of remote sensing technology, it has become possible to obtain higher resolution and larger quantities of remote sensing images, laying the foundation for target detection in remote sensing images to play a more critical role in applications. Target detection in remote sensing images requires processing large amounts of remote sensing image data, and the challenges of distinguishing different target categories and handling target overlap are significant. Effective implementation requires the use of computer vision and machine learning methods. In recent years, deep learning technology has developed rapidly, and various deep learning-based target detection algorithms have achieved significant performance improvements. These algorithms have made outstanding progress in remote sensing image target detection, providing more accurate and efficient support for many applications such as resource management, environmental monitoring, and security.

[0004] Deep learning object detection technology has brought new development opportunities for object detection in remote sensing images of transmission line corridor hazards. Currently, the main technical approaches are divided into two categories: one is two-stage detection methods, such as RCNN and Faster-RCNN. These methods generally have high detection accuracy and can handle multi-scale and complex scenes, but their high computational complexity and slow speed make them less than ideal for object detection in remote sensing images of corridor hazards. The other is single-stage detection methods, such as the YOLO series of methods, which achieve object detection through a single forward propagation, are faster and perform well in real-time applications.

[0005] While single-stage detection methods have achieved some success in object detection in remote sensing images of transmission line hazards, they also face challenges such as varying target orientations and complex backgrounds. To address these issues, researchers have proposed several improved methods. For example, some researchers have proposed the YOLO-FSD algorithm, which reduces prediction bias caused by small and densely packed objects by introducing a Swin-CSP structure and a DWC head into each layer of the YOLOv7 network, improving the algorithm's performance in complex backgrounds. On the other hand, some researchers have proposed the DCN-YOLO one-stage object detection algorithm, which optimizes the YOLOv7 backbone network using DCNv2 and a residual structure, enhancing the model's feature extraction capabilities. However, these algorithms still cannot address the problem of densely packed small objects and the varying scales of large and small objects in transmission line remote sensing images, which can lead to feature loss. Therefore, a target detection algorithm for transmission line channel hazards remote sensing images based on ATP-YOLOv8 was developed. Summary of the Invention

[0006] The present invention provides a target detection algorithm for remote sensing images of transmission line channel hidden dangers based on ATP-YOLOv8, aiming to solve the problem of dense small targets in remote sensing images of transmission line channel hidden dangers to a certain extent while maintaining high accuracy, and taking into account the detection of both large and small targets.

[0007] To solve the above technical problems, according to one aspect of the present invention, the present invention provides the following technical solutions:

[0008] An ATP-YOLOv8-based remote sensing image target detection algorithm for transmission line channel hidden dangers includes the following specific steps:

[0009] S1: Use high-precision drones to capture transmission line corridors from all directions and angles, acquiring a large number of remote sensing images. After denoising and geometric correction preprocessing, a high-quality dataset is obtained. This dataset is then divided into training, validation, and test sets in a ratio of 8:1:1.

[0010] S2: Design a MAM network to replace the C2f module of the original YOLOv8 network and expand the context receptive field of the model;

[0011] S3: Construct a hybrid attention MHA module to adjust the weights of features in channels and space;

[0012] S4: Design a TBD module that is more suitable for remote sensing targets to reduce the feature loss of large and small targets of transmission line hidden dangers during downsampling;

[0013] S5: The preprocessed training set is input into the ATP-YOLOv8 network training. During training, the validation set is used to continuously evaluate the model performance. Finally, the test set is tested using the optimal training weights to obtain the final test results.

[0014] As a preferred solution of the ATP-YOLOv8-based transmission line channel hidden danger remote sensing image target detection algorithm described in the present invention, wherein: the data set in S1 is 6382 pictures of different heights, orientations, and times, of which the training set is 5106 pictures, the validation set is 638 pictures, and the test set is 638 pictures.

[0015] As a preferred solution of the ATP-YOLOv8-based transmission line channel hidden danger remote sensing image target detection algorithm described in the present invention, wherein: the MAM network in S2 is a neural network structure designed specifically for image processing tasks, and the MAM network is composed of multiple dilated convolutions, depthwise separable convolutions, average pooling and softmax layers;

[0016] The multi-dilated convolution is used to integrate feature information of different receptive fields; the depth-wise separable convolution is used to reduce computational costs while promoting information flow; the average pooling is used to aggregate feature information in the spatial dimension into a representative numerical value; the Softmax layer is used to map vector values ​​to the probability space to obtain specific spatial adaptive weights for feature fusion.

[0017] As a preferred solution of the ATP-YOLOv8-based remote sensing image target detection algorithm for transmission line channel hidden dangers described in the present invention, the depthwise separable convolution includes the following steps:

[0018] Step 1, channel-by-channel convolution: Perform a separate convolution operation on each input channel to capture local features;

[0019] Step 2, point-by-point convolution: Use 1×1 convolution kernel for linear combination to map the output of depthwise convolution to the final output channel;

[0020] The parameter amount DP and computation amount DF of the depth-wise separable convolution are as follows:

[0021] DP=D κ ×D κ ×M+M×N

[0022] DF=D κ ×D κ ×M×D W ×D H +M×N×D W ×D H

[0023] Among them D K represents the convolution kernel, M represents the number of input channels, N represents the number of output channels, and D W Represents the output feature width, D HIndicates that the output feature is high.

[0024] As a preferred solution of the ATP-YOLOv8-based transmission line channel hidden danger remote sensing image target detection algorithm described in the present invention, the operation process of the MHA module in S3 combined with the coordinate direction and spatial channel is as follows:

[0025] Process 1: Each feature map is weighted in the spatial direction through the coordinate attention module to emphasize the key spatial position information;

[0026] Process 2: Use the channel attention module to adjust the weights in the channel direction to highlight important feature channels.

[0027] As a preferred solution of the ATP-YOLOv8-based transmission line channel hidden danger remote sensing image target detection algorithm described in the present invention, the operation process of the TBD module in S4 is as follows:

[0028] Process 1: For small targets: The TBD module captures the details of small targets through small convolution kernels and strides;

[0029] Process 2: For large targets: The TBD module processes input images of the same scale through mixed downsampling and calculates feature maps with a wider receptive field to prevent detection difficulties caused by target overlap and large-area target occlusion;

[0030] Among them, the TBD module also introduces the SE module to model the relationship between features, thereby strengthening the correlation between features of different resolutions.

[0031] As a preferred solution of the ATP-YOLOv8-based transmission line channel hidden danger remote sensing image target detection algorithm described in the present invention, wherein: when performing network training in S5, the experimental hardware and software and system parameters are set as follows: CPU is i9-10900X; memory size is 32G; GPU is NVIDIA GeForce RTX 3090; operating system is Ubuntu20.04; the target detection framework uses mmrotate0.3.3 based on pytorch1.12; Python version is 3.7; the network parameter optimizer used in the experiment is AdamW; the learning rate is 0.0001; the weight decay is 0.0001; the step learning rate adjustment strategy is used: the batch size is 16, and a total of 100 epochs are iterated.

[0032] As a preferred solution of the ATP-YOLOv8-based transmission line channel hidden danger remote sensing image target detection algorithm described in the present invention, wherein: the parameters for evaluating the model performance in S5 include: model accuracy, recall rate, average precision mean and transmission frames per second;

[0033] The model accuracy and recall are calculated as follows:

[0034]

[0035]

[0036] Among them, TP represents samples correctly judged as positive examples by the model; FP represents the situation where the model mistakenly judges negative examples as positive examples; FN represents samples that actually have hidden dangers but are mistakenly judged as negative examples by the model.

[0037] As a preferred solution of the ATP-YOLOv8-based transmission line channel hidden danger remote sensing image target detection algorithm described in the present invention, the average precision mean is a comprehensive evaluation indicator of the detection performance of multiple categories. After calculating the AP of each category, the average precision mean is obtained by averaging the AP of all categories. The calculation method of the AP and the average precision mean is as follows:

[0038]

[0039]

[0040] Where precision(r) is the precision when the recall rate is r; recall(r) is the recall rate when the recall rate is r; N is the number of categories; AP i is the average precision of the i-th category.

[0041] As a preferred solution of the ATP-YOLOv8-based transmission line channel hidden danger remote sensing image target detection algorithm described in the present invention, the calculation method of the number of frames per second is as follows:

[0042]

[0043] Among them, frame is the number of images detected by the model; time is the total time of detection.

[0044] Compared with existing technologies:

[0045] The present invention replaces the C2f module of the traditional model by designing a MAM network, and further expands the context receptive field of the model by designing an MHA module and a TMD module. This not only enhances the feature fusion capability, but also improves the representation capability of the model. It solves the problem of dense small targets and loss of feature information in the detection of hidden dangers in transmission line channels in remote sensing images, and provides a feasible solution for target detection in remote sensing images of hidden dangers in transmission line channels in complex scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 is a flow chart of the method of the present invention;

[0047] Figure 2 This is the overall network structure diagram of the present invention;

[0048] Figure 3 This is a partial dataset image of the present invention;

[0049] Figure 4 The MAM network of the present invention;

[0050] Figure 5 is the depth-separable convolutional DSC of the present invention;

[0051] Figure 6 The MHA module of the present invention;

[0052] Figure 7 The TBD module of the present invention;

[0053] Figure 8 This is the SE module of the present invention. DETAILED DESCRIPTION

[0054] To make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0055] This paper provides a remote sensing image target detection algorithm for transmission line channel hidden dangers based on ATP-YOLOv8. Figures 1-8 , including the following specific steps:

[0056] S1: Use high-precision drones to capture transmission line corridors from all directions and angles, acquiring a large number of remote sensing images. After preprocessing such as denoising and geometric correction, a high-quality dataset is obtained. This dataset is then divided into training, validation, and test sets in a ratio of 8:1:1.

[0057] S2: Design a MAM network to replace the C2f module of the original YOLOv8 network and expand the context receptive field of the model;

[0058] S3: Construct a hybrid attention MHA module to adjust the weights of features in channels and space;

[0059] S4: Design a TBD module that is more suitable for remote sensing targets to reduce the feature loss of large and small targets of transmission line hidden dangers during downsampling;

[0060] S5: The preprocessed training set is input into the ATP-YOLOv8 network training. During the training, the validation set is used to continuously evaluate the model performance. Finally, the optimal training weight is used to test the test set to obtain the final test results.

[0061] In this embodiment, after the above steps, the overall network structure of ATP-YOLOv8 constructed by the present invention is as follows: Figure 2 shown.

[0062] The remote sensing image dataset of transmission line channel hidden dangers in step S1 comes from remote sensing images of channel hidden dangers taken by drones above transmission lines. It contains 6,382 images at different heights, orientations, and times, and is divided into a training set of 5,106 images, a validation set of 638 images, and a test set of 638 images for experiments.

[0063] The MAM network in step S2 is specifically:

[0064] A MAM network was designed to replace the C2f module of ATP-YOLOv8 and expand the model's contextual information. The MAM network is a neural network structure specifically designed for image processing tasks. It primarily consists of multi-dilated convolutions, depthwise separable convolutions, average pooling, and a softmax layer. In this structure, multi-dilated convolutions integrate information from features with different receptive fields, while depthwise separable convolutions reduce computational cost and facilitate information flow. Average pooling aggregates spatial feature information into a representative numerical value. The softmax layer maps vector values ​​to a probability space, obtaining specific spatial adaptive weights, and finally performs feature fusion. The integration of these modules effectively improves the efficiency of feature learning and information transfer. By adjusting the adaptive weights, it highlights features useful for remote sensing image detection of transmission line channel hazards and reduces the impact of redundant and conflicting information.

[0065] The depthwise separable convolution (DSC) in step S2 is specifically:

[0066] DSC is a commonly used operation in convolutional neural networks, which is designed to reduce the number of parameters and computation while maintaining model performance. Figure 5 As shown, it is mainly divided into two steps:

[0067] (1) Channel-by-channel convolution, performing a separate convolution operation on each input channel to capture local features;

[0068] (2) Point-by-point convolution, using a 1x1 convolution kernel for linear combination to map the output of the depthwise convolution to the final output channel.

[0069] The parameter amount DP and computation amount DF formula of depth-separable convolution are as follows:

[0070] DP=D κ ×D κ ×M+M×N

[0071] DF=D κ ×D κX M X D W X D H + M X N X D W X D H

[0072] wherein D K represents a convolution kernel, M represents an input channel number, N represents an output channel number, D W represents an output feature width, D H represents an output feature height.

[0073] In step S3, the MHA module is specifically:

[0074] The MHA module is mainly used to solve the problem that the small target dense power line channel hidden danger in the remote sensing image cannot extract features. The MHA module combines the coordinate direction and the spatial channel. First, each feature map is adjusted in the spatial direction by the coordinate attention module, and then adjusted in the channel direction by the channel attention module. This operation mode realizes the mutual correlation and complementarity of spatial information and channel information in the small target dense power line channel hidden danger image, can more comprehensively understand the image features, comprehensively consider the information in the spatial and channel dimensions, improve the expression ability of the features in the small target dense scene, and more accurately extract the features of the small target hidden danger.

[0075] In step S4, the TBD module is specifically:

[0076] The TBD module is used to solve the problem of feature loss of different size targets in the remote sensing image in the down-sampling process. For small targets such as bird nests and kites on the power line, the TBD captures more detailed information of the small target through a small size convolution kernel and a step. For large targets such as balloons and garbage, the TBD processes the input image of the same scale through mixed down-sampling, calculates a feature map with a wider receptive field, prevents detection difficulties caused by target overlap and large area target shielding, and then models the relationship between features through the SE module to strengthen the relevance of features of different resolutions.

[0077] In step S4, the SE module is specifically:

[0078] The SE module is a squeeze and excitation attention module that can input a feature map to strengthen the channel features without changing the size of the input feature map. This is very effective for improving the accuracy and reliability of target detection of different scales in the remote sensing image of the power line channel hidden danger.

[0079] In step S5, the model training is specifically:

[0080] The S5 model training was conducted in an Ubuntu 16.04 environment, using Python 3.8 and PyTorch 1.9.0 as the primary development tools. The hardware configuration used was a GeForce RTX 3090 GPU and a Core i9-10900 CPU equipped with CUDA 11.1, with 64GB of system memory. The model was trained with an image size of 640×640, an SGD optimizer, a batch size of 16, and 100 training iterations.

[0081] The experimental hardware, software, and system parameters were as follows: i9-10900X CPU, 32GB RAM, NVIDIA GeForce RTX 3090 GPU, Ubuntu 20.04 operating system, mmrotate 0.3.3 based on PyTorch 1.12 as the object detection framework, and Python 3.7. The network parameter optimizer used in the experiment was AdamW, with a learning rate of 0.0001 and a weight decay of 0.0001. The step learning rate adjustment strategy was used, with a batch size of 16 and 100 epochs. Data augmentation methods used during training included multi-scale training, random flipping, random rotation, and affine transformation.

[0082] In step S5, the model detection result is obtained, specifically:

[0083] The improved algorithm is compared with different algorithms, and the model accuracy (Precision), recall rate (Recall), mean average precision (mAP) and frames per second (Fps) are selected to evaluate the model performance.

[0084] Precision and recall are calculated as follows:

[0085]

[0086]

[0087] Among them, TP represents samples correctly judged as positive examples by the model, for example, samples that are accurately identified as having hidden dangers (such as balloons, garbage, bird nests, kites, etc.) in the detection of hidden dangers in transmission line channels; FP represents the situation where the model mistakenly judges negative examples as positive examples, that is, areas that actually have no hidden dangers are mistakenly judged to have hidden dangers; FN represents those samples that actually have hidden dangers but are mistakenly judged as negative examples by the model (that is, hidden dangers are not detected).

[0088] mAP is a comprehensive evaluation indicator of the detection performance of multiple categories. After calculating the AP of each category, mAP is obtained by taking the average of the AP of all categories. The calculation method of AP and mAP is as follows:

[0089]

[0090]

[0091] Where precision(r) is the precision when the recall rate is r, recall(r) is the recall rate when the recall rate is r. N is the number of categories, AP i is the average precision of the i-th category.

[0092] Fps is calculated as follows:

[0093]

[0094] Where frame is the number of images detected by the model; time is the total time of detection.

[0095] In step S5, the improved algorithm is compared with different algorithms, specifically:

[0096] To accurately verify the performance of the ATP-YOLOv8 algorithm in the task of detecting hidden dangers in transmission line corridors using remote sensing images, we compared it with the YOLOv3, YOLOv4, YOLOv5, YOLOX, YOLOv7, and YOLOv8 algorithms. Each algorithm was trained and tested under the same conditions. The final experimental results are shown in Table 1.

[0097] ATP-YOLOv8 achieves a mAP of 91.52%, significantly higher than other comparison algorithms (e.g., YOLOv3's 85.21%, YOLOv4's 84.76%, YOLOv5's 81.17%, YOLOX's 86.96%, and YOLOv7's 89.76%). This indicates that ATP-YOLOv8 is more capable of accurately detecting hidden dangers in power line corridors (e.g., balloons, garbage, bird nests, kites, etc.), and can more accurately locate and identify different types of targets, effectively reducing false detections. The model achieves the highest recall rate of 92.01%, enabling more comprehensive detection of various transmission line corridor hazards. It also offers improved detection capabilities for small, partially occluded, or complex targets, ensuring the detection of as many hazards as possible and reducing safety risks. The FPS is 24.00, which, while lower than YOLOX's 37.00 and YOLOv4's 28.18, is a significant improvement over YOLOv8's 10.56 and YOLOv7's 13.40. In real-world transmission line corridor hazard detection scenarios, it is necessary to process large amounts of image data as quickly as possible while maintaining a certain level of detection accuracy to achieve real-time monitoring. While improving detection accuracy, ATP-YOLOv8 also strikes a good balance in speed, meeting certain real-time requirements. For example, in drone inspections or real-time monitoring systems, it can quickly inspect power lines, promptly identify potential hazards, and take timely action. Compared with other object detection algorithms, ATP-YOLOv8 demonstrates a good balance between accuracy (mAP, Precision, and Recall) and detection speed (FPS). Its high mAP and Recall indicate strong and comprehensive detection capabilities for transmission line corridor hazards. High Precision ensures the accuracy of detection results, while relatively high FPS enables it to function more effectively in practical applications, providing a better solution for transmission line corridor safety monitoring.

[0098] Table 1 Performance comparison with different target detection algorithms

[0099]

[0100] Among them: Bold indicates the optimal value.

[0101] Although the present invention has been described above with reference to embodiments, various modifications may be made thereto and equivalent components may be substituted without departing from the scope of the present invention. In particular, as long as there are no structural conflicts, the various features of the embodiments disclosed herein may be combined with each other in any manner, and the omission of an exhaustive description of such combinations in this specification is solely for the sake of space and resource conservation. Therefore, the present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A transmission line channel hidden danger remote sensing image target detection algorithm based on ATP-YOLOv8, characterized by: The specific steps are as follows: S1: Use high-precision drones to capture transmission line corridors from all directions and angles, acquiring a large number of remote sensing images. After denoising and geometric correction preprocessing, a high-quality dataset is obtained. This dataset is then divided into training, validation, and test sets in a ratio of 8:1:

1. S2: Design a MAM network to replace the C2f module of the original YOLOv8 network and expand the context receptive field of the model; S3: Construct a hybrid attention MHA module to adjust the weights of features in channels and space; S4: Design a TBD module that is more suitable for remote sensing targets to reduce the feature loss of large and small targets of transmission line hidden dangers during downsampling; S5: The preprocessed training set is fed into the ATP-YOLOv8 network for training. During training, the validation set is used to continuously evaluate the model performance. Finally, the optimal training weights are used to test the test set to obtain the final test results. The MAM network in S2 is a neural network structure designed specifically for image processing tasks. The MAM network consists of multiple dilated convolutions, depthwise separable convolutions, average pooling, and softmax layers. The multi-dilated convolution is used to integrate feature information of different receptive fields; The depthwise separable convolution is used to reduce computational costs while promoting information flow; the average pooling is used to aggregate feature information in the spatial dimension into a representative value; the softmax layer is used to map vector values ​​to a probability space to obtain specific spatial adaptive weights for feature fusion; The operation process of the TBD module in S4 is as follows: Process 1: For small targets: The TBD module captures the details of small targets through small convolution kernels and strides; Process 2: For large targets: The TBD module processes input images of the same scale through mixed downsampling and calculates feature maps with a wider receptive field to prevent detection difficulties caused by target overlap and large-area target occlusion; Among them, the TBD module also introduces the SE module to model the relationship between features, thereby strengthening the correlation between features of different resolutions.

2. The ATP-YOLOv8-based transmission line channel hidden danger remote sensing image target detection algorithm according to claim 1 is characterized in that: The dataset in S1 consists of 6382 images at different heights, orientations, and times, of which 5106 are for training, 638 for validation, and 638 for testing.

3. The ATP-YOLOv8-based transmission line channel hidden danger remote sensing image target detection algorithm according to claim 1 is characterized in that: The depth-wise separable convolution includes the following steps: Step 1, channel-by-channel convolution: Perform a separate convolution operation on each input channel to capture local features; Step 2, point-by-point convolution: Use 1×1 convolution kernel for linear combination to map the output of depthwise convolution to the final output channel; The parameter amount DP and computation amount DF of the depth-wise separable convolution are as follows: DP=D κ ×D κ ×M+M×N DF=D κ ×D κ ×M×D W ×D H +M×N×D W ×D H Among them D K represents the convolution kernel, M represents the number of input channels, N represents the number of output channels, and D W Represents the output feature width, D H Indicates that the output feature is high.

4. The ATP-YOLOv8-based remote sensing image target detection algorithm for transmission line channel hidden dangers according to claim 1 is characterized in that: The operation process of the MHA module in S3 combining coordinate direction and spatial channel is as follows: Process 1: Each feature map is weighted in the spatial direction through the coordinate attention module to emphasize the key spatial position information; Process 2: Use the channel attention module to adjust the weights in the channel direction to highlight important feature channels.

5. The ATP-YOLOv8-based remote sensing image target detection algorithm for transmission line channel hidden dangers according to claim 1 is characterized in that: During network training in S5, the experimental hardware and system parameter settings are as follows: CPU is i9-10900X; memory size is 32G; GPU is NVIDIA GeForce RTX 3090; operating system is Ubuntu 20.04; the target detection framework uses mmrotate0.3.3 based on pytorch1.12; Python version is 3.7; the network parameter optimizer used in the experiment is AdamW; the learning rate is 0.0001; the weight decay is 0.0001; the step learning rate adjustment strategy is used: the batch size is 16, and a total of 100 epochs are iterated.

6. The ATP-YOLOv8-based remote sensing image target detection algorithm for transmission line channel hidden dangers according to claim 1 is characterized in that: The parameters for evaluating the model performance in S5 include: model accuracy, recall rate, average precision and transmission frames per second; The model accuracy and recall are calculated as follows: Among them, TP represents samples correctly judged as positive examples by the model; FP represents the situation where the model mistakenly judges negative examples as positive examples; FN represents samples that actually have hidden dangers but are mistakenly judged as negative examples by the model.

7. The ATP-YOLOv8-based transmission line channel hidden danger remote sensing image target detection algorithm according to claim 6, characterized in that: The mean average precision is a comprehensive evaluation indicator of the detection performance of multiple categories. After calculating the AP of each category, the mean average precision is obtained by averaging the AP of all categories. The calculation method of the AP and the mean average precision is as follows: Where precision(r) is the precision when the recall rate is r; recall(r) is the recall rate when the recall rate is r; N is the number of categories; AP i is the average precision of the i-th category.

8. The ATP-YOLOv8-based transmission line channel hidden danger remote sensing image target detection algorithm according to claim 7, characterized in that: The number of frames per second is calculated as follows: Among them, frame is the number of images detected by the model; time is the total time of detection.

Images