Lightweight aerial image target detection method based on improved YOLOv11

Abstract

The invention discloses an improved YOLOv11-based lightweight aerial image target detection method, which comprises the following steps of: firstly, acquiring an aerial image data set and converting the aerial image data set into a YOLO format, and then constructing an aerial image detection model which takes a YOLOv11 model as a reference network and comprises a backbone network, a neck network and a head network; in the backbone network of the aerial image detection model, a Conv module in the backbone network of the YOLOv11 model is replaced by a DBR Down module, and an MSFLB module is added in the neck network; and finally, performing iterative training on the constructed aerial image detection model to obtain a trained aerial image detection model, and detecting a to-be-detected target in the aerial image by using the trained aerial image detection model to obtain a detection result. The aerial image detection model constructed by the invention can effectively improve the precision of aerial image target detection under the condition that edge equipment resources are limited.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial image data processing, and in particular to a lightweight aerial image target detection method based on improved YOLOv11. Background Art

[0002] With the rapid development of computer technology and drone technology, the number of aerial images is increasing and their application scope is becoming increasingly wide. Nowadays, aerial images have been widely used in many fields such as ecological protection and monitoring, traffic management, disaster prevention and control, emergency rescue, and smart cities. In the target detection task, aerial image detection technology plays a vital role. It aims to accurately identify, locate and classify various targets of interest, such as vehicles, pedestrians, buildings, etc., from complex aerial images.

[0003] With the rapid development of deep learning technology, target detection has gradually shifted from traditional methods that rely on manually designed features to automated modes based on deep neural networks. These deep learning-based algorithms not only significantly improve detection accuracy and processing efficiency, but also show stronger generalization capabilities. At present, target detection algorithms are mainly divided into two categories: two-stage detection algorithms and single-stage detection algorithms. Two-stage algorithms (such as Mask R-CNN) have attracted much attention due to their excellent detection accuracy, but their complex network structure and high computational cost limit their use in real-time applications. In contrast, single-stage algorithms (such as YOLOv8, RetinaNet, and NanoDet) have significantly improved detection speed while maintaining high detection performance due to their simple and efficient design, and have therefore been widely used in real-time scenarios and resource-constrained embedded systems.

[0004] However, single-stage aerial image detection models usually need to run on resource-constrained edge devices, which poses severe challenges to computing power and storage capacity. In order to achieve efficient and accurate object detection in these constrained environments, it is urgent to develop more lightweight detection algorithms to meet the performance requirements of practical applications. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a lightweight aerial image target detection method based on improved YOLOv11. The constructed aerial image detection model can effectively improve the accuracy of aerial image target detection when edge device resources are limited.

[0006] The technical solution of the present invention is: The lightweight aerial image target detection method based on the improved YOLOv11 specifically includes the following steps: (1) Collect aerial image data sets and convert the format of the aerial image data sets into YOLO format; (2) Construct an aerial image detection model. The aerial image detection model uses the YOLOv11 model as the benchmark network, including a backbone network, a neck network, and a head network. In the backbone network of the aerial image detection model, the Conv module in the backbone network of the YOLOv11 model is replaced with a DBRDown module. The neck network of the aerial image detection model is the neck network of the YOLOv11 model with the MSFLB module added. The DBRDown module is a lightweight downsampling convolution module that reduces the amount of calculation and parameters while improving detection accuracy; The MSFLB module is a multi-scale feature learning module, which realizes the extraction and fusion of multi-scale features through a multi-branch structure and dilated convolutions with different dilation rates; (3) Iteratively train the constructed aerial image detection model to obtain a trained aerial image detection model, and use the trained aerial image detection model to detect the target to be detected in the aerial image to obtain the detection result.

[0007] The backbone network of the aerial image detection model includes a Conv module, four DBRDown modules, four C3K2 modules, a SPPF module and a C2PSA module. The processing process of the backbone network is shown in the following formula (1): (1); In formula (1), As the input of the backbone network, that is, the collected aerial images, , , Feature maps of three different scales extracted by the backbone network.

[0008] The processing process of the DBRDown module specifically includes the following steps: S11, the input feature map first extracts local features through the depthwise separable convolution DWConv module, and then the feature map output by the DWConv module is processed by the GeLU activation function and batch normalization, and the output feature map is then residually connected with the input feature map; S12. After the feature map after residual connection is segmented by the Split function, the features are split into two branches and processed through two paths respectively: one path extracts fine-grained local features through 3×3 convolution, and the other path successively realizes downsampling and global feature fusion through the maximum pooling layer and 1×1 convolution. Finally, the outputs of the two paths are concatenated and Concat processed to integrate multi-scale information and output the spliced ​​feature map.

[0009] The parameter calculation formula of the DBRDown module is as follows (2): (2); In formula (2), is the convolution kernel size, Represents the number of channels of the input feature map, Represents the number of channels of the output feature map.

[0010] The neck network of the aerial image detection model includes two Upsample modules, two MSFLB modules, four C3K2 modules, two Conv modules and four Concat modules. The processing process of the neck network is shown in the following formula (3): (3); In formula (3), is the intermediate feature map of the neck network, , and Three feature maps output by the neck network, , , Three feature maps of different scales extracted by the backbone network; Three feature maps output by the neck network , and Input into the head network for detection and output the final detection result.

[0011] The processing process of the MSFLB module specifically includes the following steps: S21, the input feature map is first adjusted by 1×1 convolution for channel adjustment, and the output feature map is segmented by Split function, and the features are split into three branches. Each branch is sequentially subjected to 3×3 convolution to extract local spatial features, and then activated by ReLU activation function, and then the receptive field is expanded by using dilated convolution with different dilation rates to capture contextual information of different scales. The outputs of the three branches are then residually connected with the corresponding branch inputs; S22: The outputs of the three branch residual connections are concatenated in the channel dimension. Finally, the concatenated output is further convolved with a 1×1 convolution to further fuse the channel information, reduce the dimension of the output feature map, and obtain a multi-scale fused feature map.

[0012] The expansion rates of the dilated convolutions in the three branches are 1, 2, and 3, respectively, and the original convolution kernel sizes of the dilated convolutions in the three branches are all 3×3.

[0013] The aerial image detection model is iteratively trained, and the loss function of the iterative training is the FPIoU loss function, and the calculation formula is shown in the following formula (4): (4); In formula (4), represents the IoU loss function, and Represent the width and height of the real frame respectively, represents the distance between corner points, The calculation formula is shown in the following formula (5): (5); In formula (5), , Respectively represent the horizontal and vertical coordinates of a corner point of the prediction box, , They represent the horizontal and vertical coordinates of a corner point of the real box respectively.

[0014] Advantages of the present invention: (1) The present invention replaces the Conv module in the backbone network of the YOLOv11 model with a DBRDown module and adopts a lightweight downsampling structure to reduce the computational complexity of the model while maintaining stable detection accuracy, so as to better retain the target information characteristics.

[0015] (2) The present invention adds the MSFLB module to the neck network of the YOLOv11 model to address the problem of insufficient feature information extraction of small targets, enhance the feature extraction and fusion capabilities of the aerial image detection model for small targets, and improve the detection accuracy of the aerial image detection model for small targets.

[0016] (3) The loss function for iterative training of the aerial image detection model of the present invention is the FPIoU loss function. By combining the IoU loss function and the normalized vertex distance penalty, it not only measures the overlap between the predicted box and the true box, but also accurately optimizes the position and shape of the box. The logarithmic function smoothes the distance penalty, reduces extreme errors, and improves the robustness to objects of different sizes through normalization, effectively making up for the lack of position sensitivity of the IoU loss function and significantly improving the regression quality of the bounding box. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flow chart of the present invention.

[0018] Figure 2 It is a network framework diagram of the aerial image detection model of the present invention.

[0019] Figure 3 It is a network framework diagram of the DBRDown module of the present invention.

[0020] Figure 4 It is a network framework diagram of the MSFLB module of the present invention. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] See Figure 1 , a lightweight aerial image target detection method based on improved YOLOv11 specifically includes the following steps: (1) Collect aerial image data sets and convert the format of the aerial image data sets into YOLO format; (2) Build an aerial image detection model (see Figure 2 ), the aerial image detection model uses the YOLOv11 model as the benchmark network, including the backbone network, neck network and head network; The backbone network of the aerial image detection model is used to extract features of multiple scales from the input aerial images. It includes a Conv module, four DBRDown modules, four C3K2 modules, a SPPF module and a C2PSA module. The processing process of the backbone network is shown in the following formula (1): (1); In formula (1), As the input of the backbone network, that is, the collected aerial images, , , Three feature maps of different scales extracted by the backbone network; The neck network of the aerial image detection model is responsible for multi-scale feature fusion. It enhances the feature representation capability by fusing feature maps from different stages of the backbone network. It includes two Upsample modules, two MSFLB modules, four C3K2 modules, two Conv modules, and four Concat modules. The processing process of the neck network is shown in the following formula (3): (3); In formula (3), is the intermediate feature map of the neck network, , and Three feature maps output by the neck network, , , Three feature maps of different scales extracted by the backbone network; Three feature maps output by the neck network , and The input is sent to the three detection heads of the head network for detection, and the final detection result is output, that is, a bounding box is generated and the target in the bounding box is classified; See Figure 3 The DBRDown module is a lightweight downsampling convolution module that reduces the amount of calculation and parameters while improving the detection accuracy. The processing process of the DBRDown module specifically includes the following steps: S11. The input feature map is firstly processed by the depthwise separable convolution DWConv 3×3 module to extract local features. Then the feature map output by the DWConv 3×3 module is processed by the GeLU activation function and batch normalization BatchNorm to enhance the nonlinear expression ability and feature stability. The output feature map is then residually connected with the input feature map to alleviate the gradient vanishing problem and improve the optimization efficiency. S12. After the feature map after residual connection is segmented by the Split function, the features are split into two branches and processed through two paths respectively: one path extracts fine-grained local features through 3×3 convolution (Conv 3×3), and the other path successively passes through the maximum pooling layer (MaxPool2d) and 1×1 convolution (Conv 1×1) to achieve downsampling and global feature fusion. Finally, the outputs of the two paths are concatenated to integrate multi-scale information and output the concatenated feature map; The parameter calculation formula of the ordinary downsampling convolution Conv module is as follows (6): (6); The parameter calculation formula of the DBRDown module is as follows (2): (2); In formula (2) and formula (6), is the convolution kernel size, Represents the number of channels of the input feature map, Represents the number of channels of the output feature map; when , , ; Then the ratio of the parameters of the Conv module and the DBRDown module is: ,It can be seen that the number of parameters of the DBRDown module is ,greatly reduced; See Figure 4 The MSFLB module is a multi-scale feature learning module. It realizes the extraction and fusion of multi-scale features through multi-branch structure and dilated convolution with different dilation rates. The processing process of the MSFLB module specifically includes the following steps: S21. The input feature map is first adjusted through a 1×1 convolution (Conv 1×1) to enhance the feature expression capability. The representation space of the feature map is expanded by increasing the number of channels, so that the network can capture richer contour information. Then, the output feature map is segmented by the Split function, and the features are split into three branches. Each branch is sequentially subjected to a 3×3 convolution (Conv 3×3) to extract local spatial features. Then, it is activated by the ReLU activation function to have better nonlinear expression capability. Then, the dilated convolution with different dilation rates (DConv 3×3) is used to expand the receptive field and capture contextual information of different scales. The dilation rates d of the dilated convolutions in the three branches are 1, 2, and 3, respectively. The outputs of the three branches are then residually connected with the corresponding branch inputs to effectively retain the detailed features and enhance the feature learning capability of the network. S22, concatenate the outputs of the three branch residual connections in the channel dimension, so that the multi-level information from different dimensions is integrated to form a rich semantic expression. Finally, the concatenated output is further fused with channel information by 1×1 convolution (Conv 1×1), reducing the dimension of the output feature map and obtaining a multi-scale fused feature map. (3) Iteratively train the constructed aerial image detection model and set the model training hyperparameters, that is, use the SGD optimizer to optimize the aerial image detection model, with an initial learning rate of 0.01, a final learning rate of 0.0001, a momentum of 0.937, a batch normalization size of 16, and 200 iterations; iteratively train the aerial image detection model using the training set, and only save the weight information of the optimal model and the last round model during the training process; The loss function of iterative training is the FPIoU loss function, and the calculation formula is shown in the following formula (4): (4); In formula (4), represents the IoU loss function, and Represent the width and height of the real frame respectively, represents the distance between corner points, The calculation formula is shown in the following formula (5): (5); In formula (5), , Respectively represent the horizontal and vertical coordinates of a corner point of the prediction box, , They represent the horizontal and vertical coordinates of a corner point of the real box respectively.

[0023] Then, the trained aerial image detection model is used to detect the target to be detected in the aerial image to obtain the detection result.

[0024] Performance Analysis: The aerial image detection model of the present invention was subjected to a performance test experiment with the existing YOLO11n model (YOLOv11 training model), and the experimental results are shown in the table below. As can be seen from Table 1 below, the aerial image detection model of the present invention has improved by 3.6%, 1.2%, 2.1% and 4.3% in terms of Precision, Recall, mAP@0.5 and mAP@0.5:0.95 compared to the YOLO11n model, respectively. At the same time, the Params (parameter quantity) indicator of the aerial image detection model has been reduced by 16.2%. These experimental results show that the aerial image detection model of the present invention has achieved significant performance improvement while taking into account both model scale and detection accuracy.

[0025] Table 1

[0026] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A lightweight aerial image target detection method based on improved YOLOv11, characterized by: The specific steps include: (1) Collect aerial image data sets and convert the format of the aerial image data sets into YOLO format; (2) Construct an aerial image detection model. The aerial image detection model uses the YOLOv11 model as the benchmark network, including a backbone network, a neck network, and a head network. In the backbone network of the aerial image detection model, the Conv module in the backbone network of the YOLOv11 model is replaced with a DBRDown module. The neck network of the aerial image detection model is the neck network of the YOLOv11 model with the MSFLB module added. The DBRDown module is a lightweight downsampling convolution module that reduces the amount of calculation and parameters while improving detection accuracy; The MSFLB module is a multi-scale feature learning module, which realizes the extraction and fusion of multi-scale features through a multi-branch structure and dilated convolutions with different dilation rates; (3) Iteratively train the constructed aerial image detection model to obtain a trained aerial image detection model, and use the trained aerial image detection model to detect the target to be detected in the aerial image to obtain the detection result.

2. The lightweight aerial image target detection method based on improved YOLOv11 according to claim 1, characterized in that: The backbone network of the aerial image detection model includes a Conv module, four DBRDown modules, four C3K2 modules, a SPPF module and a C2PSA module. The processing process of the backbone network is shown in the following formula (1): （1）； In formula (1), As the input of the backbone network, that is, the collected aerial images, , , Feature maps of three different scales extracted by the backbone network.

3. The lightweight aerial image target detection method based on improved YOLOv11 according to claim 2 is characterized in that: The processing process of the DBRDown module specifically includes the following steps: S11, the input feature map first extracts local features through the depthwise separable convolution DWConv module, and then the feature map output by the DWConv module is processed by the GeLU activation function and batch normalization, and the output feature map is then residually connected with the input feature map; S12. After the feature map after residual connection is segmented by the Split function, the features are split into two branches and processed through two paths respectively: one path extracts fine-grained local features through 3×3 convolution, and the other path successively realizes downsampling and global feature fusion through the maximum pooling layer and 1×1 convolution. Finally, the outputs of the two paths are concatenated and Concat processed to integrate multi-scale information and output the spliced ​​feature map.

4. The lightweight aerial image target detection method based on improved YOLOv11 according to claim 3 is characterized in that: The parameter calculation formula of the DBRDown module is as follows (2): （2）； In formula (2), is the convolution kernel size, Represents the number of channels of the input feature map, Represents the number of channels of the output feature map.

5. The lightweight aerial image target detection method based on improved YOLOv11 according to claim 1, characterized in that: The neck network of the aerial image detection model includes two Upsample modules, two MSFLB modules, four C3K2 modules, two Conv modules and four Concat modules. The processing process of the neck network is shown in the following formula (3): （3）； In formula (3), is the intermediate feature map of the neck network, , and Three feature maps output by the neck network, , , Three feature maps of different scales extracted by the backbone network; Three feature maps output by the neck network , and Input into the head network for detection and output the final detection result.

6. The lightweight aerial image target detection method based on improved YOLOv11 according to claim 5 is characterized in that: The processing process of the MSFLB module specifically includes the following steps: S21, the input feature map is first adjusted by 1×1 convolution for channel adjustment, and the output feature map is segmented by Split function, and the features are split into three branches. Each branch is sequentially subjected to 3×3 convolution to extract local spatial features, and then activated by ReLU activation function, and then the receptive field is expanded by using dilated convolution with different dilation rates to capture contextual information of different scales. The outputs of the three branches are then residually connected with the corresponding branch inputs; S22: The outputs of the three branch residual connections are concatenated in the channel dimension. Finally, the concatenated output is further convolved with a 1×1 convolution to further fuse the channel information, reduce the dimension of the output feature map, and obtain a multi-scale fused feature map.

7. The lightweight aerial image target detection method based on improved YOLOv11 according to claim 6, characterized in that: The expansion rates of the dilated convolutions in the three branches are 1, 2, and 3, respectively, and the original convolution kernel sizes of the dilated convolutions in the three branches are all 3×3.

8. The lightweight aerial image target detection method based on improved YOLOv11 according to claim 1, characterized in that: The aerial image detection model is iteratively trained, and the loss function of the iterative training is the FPIoU loss function, and the calculation formula is shown in the following formula (4): （4）； In formula (4), represents the IoU loss function, and Represent the width and height of the real frame respectively, represents the distance between corner points, The calculation formula is shown in the following formula (5): （5）； In formula (5), , Respectively represent the horizontal and vertical coordinates of a corner point of the prediction box, , They represent the horizontal and vertical coordinates of a corner point of the real box respectively.

Images