Small target and shielded target detection method suitable for remote sensing image

Abstract

The invention discloses a small target and shielded target detection method suitable for a remote sensing image, and the method comprises the steps: firstly collecting a high-resolution remote sensing image data set, and converting the high-resolution remote sensing image data set into a YOLO format; then, a remote sensing image detection model is constructed for target detection, the remote sensing image detection model takes a YOLOv10 model as a reference network, a C2f-CPCC module is used for replacing a C2f module in a backbone network and a neck network, detailed position information of a low-order feature map and rich semantic information of a high-order feature map are combined, the detection capability is improved, a DAPD module is used for replacing a convolution module, and the detection efficiency is improved. Dynamic processing is carried out according to the region of input features and context information, key features in a remote sensing image are extracted, and overall information of the region is reserved, so that the capturing capability of a model on global features is enhanced, a Repulsion-IoU loss function is introduced to solve the problem of mutual shielding among small targets in the detection process, and the detection accuracy is improved. And the recognition capability of the model on the shielded target is improved.

Description

Technical Field

[0001] The invention relates to the technical field of remote sensing image data processing, in particular to a small target and occluded target detection method suitable for remote sensing images. Background Art

[0002] With the rapid development of remote sensing technology, the application of remote sensing images in many fields such as geographic information, environmental monitoring, urban planning, and agricultural monitoring has become more and more extensive. Remote sensing images are usually obtained through satellites, drones, or other remote sensing equipment. These images can provide a lot of valuable information about the earth's surface and are of extremely high value. In the fields of large-scale earth surface change analysis, disaster monitoring, land use planning, urban expansion, and agricultural development, remote sensing images are undoubtedly an important basis for analysis and decision-making. However, the complexity of remote sensing images also brings considerable challenges to target detection, especially the large differences in target size, complex background, and poor image quality.

[0003] The application of target detection technology in remote sensing images is mainly used to extract, locate and classify ground features, such as buildings, roads, farmlands, water bodies, forests, etc. The accurate extraction of these ground features not only helps environmental monitoring, urban planning and land use management, but also plays an important role in precision agriculture, climate change monitoring and ecological protection. In recent years, the rapid progress of deep learning technology has brought significant breakthroughs in remote sensing image target detection. Traditional target detection methods rely on manual feature extraction, which is inefficient and difficult to adapt to complex changes in remote sensing images. Deep learning, especially convolutional neural networks (CNNs) and region extraction algorithms, can automatically learn and extract effective features in images, greatly improving the accuracy, speed and robustness of target detection.

[0004] Remote sensing image target detection algorithms can be roughly divided into two categories: two-stage detection algorithms and single-stage detection algorithms. Two-stage detection algorithms such as Faster R-CNN can usually provide higher detection accuracy, but due to their high computational complexity and slow processing speed, they are not suitable for real-time monitoring tasks. Single-stage detection algorithms, such as YOLO, SSD, RetinaNet, etc., have higher processing speeds and lower computational overhead, so they are more suitable for real-time detection and large-scale remote sensing image processing tasks. With the gradual popularization of computing resources and the continuous optimization of algorithms, single-stage algorithms have become a common choice for remote sensing image target detection.

[0005] However, target detection in remote sensing images still faces some challenges, mainly in the following aspects: (1) Small targets are difficult to detect: Remote sensing images are usually taken by satellite or drone, and the image resolution is low. Due to the scale differences of surface targets, the detection of some small ground objects such as small buildings, farmland and water bodies is often difficult due to the low image clarity. The low resolution may cause small targets to be blurred or distorted in the image, and traditional target detection methods have difficulty in accurately identifying and locating small targets from these low-quality images. (2) Occlusion problem: In remote sensing images, target objects are often occluded, especially in urban environments, where targets such as buildings, roads, and transportation facilities are often partially occluded or overlapped by other objects. In addition, targets in natural environments such as forests, mountains or agricultural areas may also be affected by trees, land, other ground objects or weather conditions, resulting in partial loss or invisibility of the target, which poses a huge challenge to target detection algorithms. (3) High computing requirements: Remote sensing images usually cover a large area and contain a large amount of ground object information. In order to ensure detection accuracy, target detection models must process and analyze a large amount of image data, which places high demands on computing resources. Especially in real-time detection applications, image processing speed and model computational efficiency have become a major bottleneck. Currently, many high-precision target detection models have large computational requirements, and require a large amount of computing resources and storage space when deployed and executed. They are difficult to deploy efficiently on resource-constrained devices, especially in environments with limited computing resources such as drones or edge devices, and are difficult to meet the needs of real-time processing. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide a small target and occluded target detection method suitable for remote sensing images. By optimizing the backbone network and neck network of the YOLOv10 model, efficient feature extraction and fusion are achieved, thereby significantly improving the performance of the model for small target and occluded target detection in remote sensing images, achieving efficient and accurate target detection effect.

[0007] The technical solution of the present invention is: A small target and occluded target detection method suitable for remote sensing images specifically includes the following steps: (1) Collect high-resolution remote sensing image datasets and convert them into YOLO format; (2) Construct a remote sensing image detection model. The remote sensing image detection model uses the YOLOv10 model as the benchmark network, including a backbone network, a neck network, and a head network. The C2f-CPCC module is used to replace the C2f module in the backbone network and the neck network. The C2f-CPCC module replaces the Bottleneck block in the C2f module with the CPCC module. The CPCC module performs maximum pooling, multi-scale convolution, and fusion operations on the input feature map. Through multi-scale convolution and feature fusion, the model's perception of targets of different scales is improved. The DAPD module is used to replace the Conv and SCDown modules in the backbone network and the neck network. The DAPD module performs grouped convolution, depthwise separable convolution, and average pooling operations on the input feature map, thereby retaining global information and avoiding over-reliance on local features. (3) The constructed remote sensing image detection model is iteratively trained. The training loss function is the Repulsion-IoU loss function, and its calculation formula is as follows (3): (3); In formula (3), Represents the Repulsion-IoU loss function, is the attraction item of the target real box to the target detection box, It is the exclusion item of the non-target real box to the target detection box. is the IoU loss function, and is a hyperparameter; (4) Use the trained remote sensing image detection model to detect the target to be detected in the remote sensing image and obtain the detection result.

[0008] The backbone network includes two Conv modules, two C2f modules, two C2f-CPCC modules, three DAPD modules, one SPPF module and one PSA 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 remote sensing image, , , Three feature maps of different scales extracted by the backbone network; The neck network includes two Upsample modules, one C2f module, three C2f-CPCC modules, one DAPD module, one Conv module and four Concat modules. The processing process of the neck network is shown in the following formula (2): (2); In formula (2), is the intermediate feature map of the neck network, , and Three feature maps output by the neck network; Three feature maps output by the neck network , and Input into the head network for detection and output the final detection result.

[0009] The processing process of the C2f-CPCC module specifically includes the following steps: S11, the input feature map is transformed through the first convolution layer to generate an intermediate feature map, and then the generated intermediate feature map is split into two parts by the Split function: one part is passed to multiple CPCC modules for further processing, and finally the multiple feature maps obtained by the multiple CPCC modules are spliced ​​with the other part after the split to form a fused feature map, and the fused feature map is then passed through the second convolution layer to output the feature map; S12. The processing process of each CPCC module is as follows: the input feature map is divided along the channel dimension by the Split function and then input into two branches respectively. The number of feature map channels input into each branch is half of the input feature map. The first branch first performs the maximum pooling operation, and then uses 1×1 convolution to fuse the channel features and outputs the feature map. The second branch uses convolution kernels of different sizes to extract multi-scale features to obtain multiple multi-scale feature maps. After the multiple multi-scale feature maps are spliced ​​and activated by the ReLU activation function, the feature map is output. Finally, the feature map output by the first branch and the feature map output by the second branch are spliced, and the features are further fused by 1×1 convolution to output the feature map.

[0010] The processing process of the DAPD module specifically includes the following steps: S21, the input feature map is divided along the channel dimension and input into two branches respectively. The number of feature map channels input into each branch is half of the input feature map. The first branch first performs group convolution, and then the output feature map is divided into two branches. After the two branches perform depthwise separable convolution and average pooling operations respectively, they are batch normalized by the BN layer and then output two feature maps. S22, the second branch performs depth-wise separable convolution and outputs a feature map; S23: Concatenate the two feature maps output by the first branch and the feature map output by the second branch, perform 1×1 convolution on the concatenated maps, fuse different features, and output a feature map.

[0011] The attraction item of the target real box to the target detection box The calculation process of is shown in the following formula (4): (4); In formula (4), is the prediction box, It is a real frame. is the smooth L1 loss function.

[0012] The non-target real frame excludes the target detection frame The calculation process of is shown in the following formula (5): (5); In formula (5), is the weight of the occluded background sample, which is used to measure the degree of rejection of the target detection frame by the non-target area. It is a real frame. is the prediction box.

[0013] The IoU loss function The calculation process of is shown in the following formula (6): (6); In formula (6), It is a real frame. is the prediction box.

[0014] Advantages of the present invention: (1) The present invention replaces some C2f modules in the backbone network and neck network of the existing YOLOv10 model with C2f-CPCC modules. In the backbone network, the C2f-CPCC module is mainly used to enhance the feature extraction capability, and improve the model's perception capability of targets of different scales through multi-scale convolution and feature fusion; in the neck network, the function of the C2f-CPCC module is to optimize feature fusion, and further improve the accuracy of target detection by fusing and enhancing the multi-scale features output by the backbone network, especially when processing complex and occluded scenes; that is, the C2f-CPCC module effectively reduces redundant calculations while extracting multi-scale information, thereby improving detection performance.

[0015] (2) The present invention replaces some Conv and SCDown modules in the backbone network and neck network of the existing YOLOv10 model with DAPD modules. The deep average pooling downsampling DAPD module has obvious advantages over the traditional convolution Conv and SCDown modules. First, the DAPD module can better retain global information. By calculating the average value in the region, it reduces the loss of local features, especially when processing targets with large scale differences. It is more stable. Second, the DAPD module has a strong inhibitory effect on noise, can smooth outliers, and avoid over-reliance on local features. Third, the DAPD module is relatively simple to calculate, especially when processing large-scale inputs, it can significantly reduce the amount of calculation. In contrast, the SCDown module may have a higher calculation cost when resources are limited. Finally, DAPD has better adaptability and generalization ability, can effectively process targets of different scales, and is not prone to losing important information. However, the traditional convolution Conv and SCDown modules may not perform well when extracting features from small-scale targets or complex backgrounds.

[0016] (3) The loss function for training in the present invention is the Repulsion-IoU loss function. On the basis of the traditional IoU loss function, an exclusion mechanism between targets is introduced, which effectively solves the detection problem of small target occlusion and dense target area in remote sensing images. While calculating the target frame overlap, the loss function adds an exclusion term to deal with the mutual interference between occluded targets, avoids false detection or missed detection caused by occlusion, and enhances the model's ability to distinguish between dense targets and small targets. In particular, when occlusion or overlap occurs between targets, the detection ability of the model can be better optimized. The model is optimized by measuring the overlap between the predicted frame and the true frame (IoU loss function), so that the model performs better in processing the precise positioning between large targets and target frames. 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 remote sensing image detection model of the present invention.

[0019] Figure 3 It is a network framework diagram of the C2f-CPCC module of the present invention.

[0020] Figure 4 It is a network framework diagram of the CPCC module part in the C2f-CPCC module of the present invention.

[0021] Figure 5 It is a network framework diagram of the DAPD module of the present invention. DETAILED DESCRIPTION

[0022] 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.

[0023] See Figure 1 , a small target and occluded target detection method suitable for remote sensing images, specifically including the following steps: (1) Collect high-resolution remote sensing image datasets and convert them into YOLO format; (2) See Figure 2 , build a remote sensing image detection model. The remote sensing image detection model uses the YOLOv10 model as the benchmark network, including the backbone network, neck network and head network; The backbone network includes two Conv (convolution) modules, two C2f modules, two C2f-CPCC modules, three DAPD modules, one SPPF module and one PSA 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 remote sensing image, , , Three feature maps of different scales extracted by the backbone network; The neck network includes two Upsample modules, one C2f module, three C2f-CPCC modules, one DAPD module, one Conv module and four Concat modules. The processing process of the neck network is shown in the following formula (2): (2); In formula (2), is the intermediate feature map of the neck network, , and Three feature maps output by the neck network; Three feature maps output by the neck network , and Input into the head network for detection and output the final detection result; See Figure 3 and Figure 4 The processing process of the C2f-CPCC module specifically includes the following steps: S11, the input feature map is transformed through the first convolution layer to generate an intermediate feature map, and then the generated intermediate feature map is split into two parts by the Split function: one part is passed to multiple CPCC modules for further processing, and finally the multiple feature maps obtained by the multiple CPCC modules are spliced ​​with the other part after the split to form a fused feature map, and the fused feature map is then passed through the second convolution layer to output the feature map; S12. The processing process of each CPCC module is as follows: the input feature map is divided along the channel dimension by the Split function and then input into two branches respectively. The number of feature map channels input into each branch is half of the input feature map. The first branch first performs the maximum pooling (Max-Pool) operation, and then uses 1×1 convolution to fuse the channel features and output the feature map. The second branch uses three convolution kernels of different sizes (1×1 convolution, 3×3 convolution, 5×5 convolution) to extract multi-scale features and obtain three multi-scale feature maps. The three multi-scale feature maps are spliced ​​and activated by the ReLU activation function to output the feature map. Finally, the feature map output by the first branch and the feature map output by the second branch are spliced, and the features are further fused by 1×1 convolution to output the feature map. See Figure 5 , the DAPD module is a deep average pooling downsampling module, and the processing process specifically includes the following steps: S21, the input feature map is split along the channel dimension by the Split function and then input into two branches respectively. The number of feature map channels input into each branch is half of the input feature map. The first branch first performs group convolution (Group-Conv), and then the output feature map is divided into two branches. These two branches perform depth-separable convolution (DWConv) and average pooling (Average-Pool) operations respectively, and then both are batch normalized by the BN layer to output two feature maps; S22, the second branch performs depthwise separable convolution (DWConv) and outputs the feature map; S23, concatenating the two feature maps output by the first branch and the feature map output by the second branch, performing 1×1 convolution after concatenation, fusing different features, and outputting a feature map; (3) The constructed remote sensing image detection model is iteratively trained. The training loss function is the Repulsion-IoU loss function, and its calculation formula is as follows (3): (3); In formula (3), Represents the Repulsion-IoU loss function, is the attraction item of the target real box to the target detection box, It is the exclusion item of the non-target real box to the target detection box. is the IoU loss function, and is a hyperparameter; Among them, the attraction item of the target real box to the target detection box The calculation process of is shown in the following formula (4): (4); In formula (4), is the prediction box, It is a real frame. is the smooth L1 loss function; The exclusion item of the non-target real box to the target detection box The calculation process of is shown in the following formula (5): (5); In formula (5), is the weight of the occluded background sample, which is used to measure the degree of rejection of the target detection frame by the non-target area. It can be determined in a variety of ways, including: ① IoU weighting, which is calculated based on the IoU between the target frame and the occluded area. The higher the IoU, the greater the weight; ② Area ratio weighting, the larger the area ratio of the occluded area, the higher the weight; ③ Confidence weighting, the lower the target confidence, the more serious the occlusion, and the weight increases accordingly; ④ Attention mechanism, using the channel or spatial attention module to dynamically adjust the weight. The above weight determination methods can be used alone or in combination to more accurately measure the impact of occlusion on target detection; It is a real frame; is the prediction box; IoU loss function The calculation process of is shown in the following formula (6): (6); In formula (6), It is a real frame. is the prediction box;

[0024] (4) Use the trained remote sensing image detection model to detect the target to be detected in the remote sensing image and obtain the detection result.

[0025] Performance Analysis: The remote sensing image detection model (Ours) of the present invention is compared with other existing detection models (Faster RCNN, DETR, Tood, DDQ, ATSS, DINO, YOLOv10) for performance testing experiments, and the experimental results are shown in Table 1 below. As can be seen from Table 1, the detection model (Ours) of the present invention leads other existing detection models with an accuracy of 0.817, indicating that it performs best in correctly identifying positive samples, and has the lowest Params (parameter quantity) indicator, and the detection performance is greatly improved.

[0026] Table 1

[0027] 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 method for detecting small targets and occluded targets in remote sensing images, characterized in that: The specific steps include: (1) Collect high-resolution remote sensing image datasets and convert them into YOLO format; (2) Construct a remote sensing image detection model. The remote sensing image detection model uses the YOLOv10 model as the benchmark network, including a backbone network, a neck network, and a head network. The C2f-CPCC module is used to replace the C2f module in the backbone network and the neck network. The C2f-CPCC module replaces the Bottleneck block in the C2f module with the CPCC module. The CPCC module performs maximum pooling, multi-scale convolution, and fusion operations on the input feature map. Through multi-scale convolution and feature fusion, the model's perception ability of targets of different scales is improved. The DAPD module is used to replace the Conv and SCDown modules in the backbone network and the neck network. The DAPD module performs grouped convolution, depthwise separable convolution, and average pooling operations on the input feature map. (3) The constructed remote sensing image detection model is iteratively trained. The training loss function is the Repulsion-IoU loss function, and its calculation formula is as follows (3): （3）； In formula (3), Represents the Repulsion-IoU loss function, is the attraction item of the target real box to the target detection box, It is the exclusion item of the non-target real box to the target detection box. is the IoU loss function, and is a hyperparameter; (4) Use the trained remote sensing image detection model to detect the target to be detected in the remote sensing image and obtain the detection result.

2. The method for detecting small targets and occluded targets in remote sensing images according to claim 1, characterized in that: The backbone network includes two Conv modules, two C2f modules, two C2f-CPCC modules, three DAPD modules, one SPPF module and one PSA 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 remote sensing image, , , Three feature maps of different scales extracted by the backbone network; The neck network includes two Upsample modules, one C2f module, three C2f-CPCC modules, one DAPD module, one Conv module and four Concat modules. The processing process of the neck network is shown in the following formula (2): （2）； In formula (2), is the intermediate feature map of the neck network, , and Three feature maps output by the neck network; Three feature maps output by the neck network , and Input into the head network for detection and output the final detection result.

3. The method for detecting small targets and occluded targets in remote sensing images according to claim 2, characterized in that: The processing process of the C2f-CPCC module specifically includes the following steps: S11, the input feature map is transformed through the first convolution layer to generate an intermediate feature map, and then the generated intermediate feature map is split into two parts by the Split function: one part is passed to multiple CPCC modules for further processing, and finally the multiple feature maps obtained by the multiple CPCC modules are spliced ​​with the other part after the split to form a fused feature map, and the fused feature map is then passed through the second convolution layer to output the feature map; S12. The processing process of each CPCC module is as follows: the input feature map is divided along the channel dimension by the Split function and then input into two branches respectively. The number of feature map channels input into each branch is half of the input feature map. The first branch first performs the maximum pooling operation, and then uses 1×1 convolution to fuse the channel features and outputs the feature map. The second branch uses convolution kernels of different sizes to extract multi-scale features to obtain multiple multi-scale feature maps. After the multiple multi-scale feature maps are spliced ​​and activated by the ReLU activation function, the feature map is output. Finally, the feature map output by the first branch and the feature map output by the second branch are spliced, and the features are further fused by 1×1 convolution to output the feature map.

4. The method for detecting small targets and occluded targets in remote sensing images according to claim 2, characterized in that: The processing process of the DAPD module specifically includes the following steps: S21, the input feature map is divided along the channel dimension and input into two branches respectively. The number of feature map channels input into each branch is half of the input feature map. The first branch first performs group convolution, and then the output feature map is divided into two branches. After the two branches perform depthwise separable convolution and average pooling operations respectively, they are batch normalized by the BN layer and then output two feature maps. S22, the second branch performs depth-wise separable convolution and outputs a feature map; S23: Concatenate the two feature maps output by the first branch and the feature map output by the second branch, perform 1×1 convolution on the concatenated maps, fuse different features, and output the feature map.

5. The small target and occluded target detection method applicable to remote sensing images according to claim 1, characterized in that: The attraction item of the target real box to the target detection box The calculation process of is shown in the following formula (4): （4）； In formula (4), is the prediction box, It is a real frame. is the smooth L1 loss function.

6. The small target and occluded target detection method applicable to remote sensing images according to claim 1, characterized in that: The non-target real frame excludes the target detection frame The calculation process of is shown in the following formula (5): （5）； In formula (5), is the weight of the occluded background sample, which is used to measure the degree of rejection of the target detection frame by the non-target area. It is a real frame. is the prediction box.

7. The small target and occluded target detection method applicable to remote sensing images according to claim 1, characterized in that: The IoU loss function The calculation process of is shown in the following formula (6): （6）； In formula (6), It is a real frame. is the prediction box.

Images