Contact network dropper state detection method based on improved YOLOv11s

Abstract

The invention relates to the technical field of catenary dropper detection, in particular to a catenary dropper state detection method based on improved YOLOv11s, and the method comprises the steps: constructing a small target data set based on a small target defect image of a catenary dropper; inputting the small target data set into a constructed improved YOLOv11s algorithm network for training to obtain a trained detection model; and recognizing a catenary dropper image collected on site through the trained detection model, obtaining a small target detection result, and completing state detection. According to the invention, multi-scale feature fusion is realized based on the improved YOLOv11s detection model, small targets can be detected more accurately, missed detection and false detection are reduced, and the robustness, reliability and integrity of detection in a complex environment are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of catenary dropper string detection, and in particular to a catenary dropper string state detection method based on improved YOLOv11s. Background Art

[0002] In high-speed railway systems, the catenary system is the core of train power transmission. As a key component, the catenary wires bear the weight, support the transmission lines, and buffer vibration and shock. Their condition is crucial to the safety and stability of railway operations. However, due to the long-term dual loads of vehicles and the environment, the wires are prone to loosening, deformation, and other faults, threatening power supply stability and train safety. Therefore, regular and accurate inspection is essential. In the past, manual inspection was inefficient and prone to errors. The introduction of 4C devices has partially solved these problems, but with the dramatic increase in image data, existing algorithms are experiencing missed detections and false detections, making them unable to meet demand. Summary of the Invention

[0003] In order to solve the problems of low manual detection efficiency and missed detection and false detection in existing algorithms in the existing catenary dropper status detection, the present invention provides a catenary dropper status detection method based on improved YOLOv11s.

[0004] The present invention provides a method for detecting the state of a catenary dropper based on an improved YOLOv11s, the method comprising: A small target dataset is constructed based on small target defect images of catenary suspension strings; Input the small target data set into the constructed improved YOLOv11s algorithm network for training to obtain a trained detection model; The trained detection model is used to identify the catenary string images collected on site, obtain small target detection results, and complete status detection. The construction process of the improved YOLOv11s algorithm network includes: Build a YOLOv11s basic network; the YOLOv11s basic network includes a Backbone network, a Neck network, and a Head network; A small target detection layer is constructed, added to the Neck network, and the SEAM attention module is introduced to obtain an improved Neck network; the improved Neck network is used to obtain feature images of four different channels from the Backbone network and output feature images of four different resolutions; The Head detection head in the Head network is combined with the DASFF module to be improved into a four-head DASFFHead detection head to obtain an improved YOLOv11s algorithm network; the DASFF module is used to fuse feature images of four different resolutions, so that the Head network outputs small target detection results.

[0005] According to a specific embodiment, in the above detection method, after the original image is input into the Backbone network, the first channel feature image is output through the first C3k2 module in the Backbone network, the second channel feature image is output through the second C3k2 module in the Backbone network, the third channel feature image is output through the third C3k2 module in the Backbone network, and the fourth channel feature image is output through the C2PSA module in the Backbone network.

[0006] According to a specific implementation, in the above detection method, the small target detection layer includes a first C3k2 module, a second C3k2 module, a first Concat module, a second Concat module, a first Conv module and a first Upsample module.

[0007] According to a specific embodiment, in the above detection method, the improved Neck network is used to upsample the fourth channel feature image, fuse the features with the third channel feature image and extract the features before upsampling, and then fuse the features with the second channel feature image, extract the features and upsample through the first Upsample module, fuse the features with the first channel feature image through the first Concat module, and then extract the features through the first C3k2 module to obtain the first resolution feature image, which is output to the DASFF module.

[0008] According to a specific embodiment, in the above detection method, the improved Neck network is also used to downsample the features extracted by the first C3k2 module through the first Conv module, fuse the features with the input of the first Upsample module through the second Concat module, extract the features through the second C3k2 module, input them into the SEAM attention module, obtain the second resolution feature image, and output it to the DASFF module.

[0009] According to a specific embodiment, in the above detection method, the SEAM attention module is used to extract small target features in the occluded area through depth-separable convolution with residual connection.

[0010] According to a specific embodiment, in the above detection method, the improved Neck network is also used to upsample the fourth channel feature image, fuse the features with the third channel feature image and extract the features, and then fuse the features with the output of the downsampled SEAM attention module to extract the features to obtain a third resolution feature image, and output it to the DASFF module; then downsample the third resolution feature image, fuse the features with the fourth channel feature image, extract the features to obtain a fourth resolution feature image, and output it to the DASFF module.

[0011] According to a specific embodiment, in the above detection method, the DASFFHead detection head adopts a DASFF module to receive inputs of four different scales, and fuses the first resolution feature image to the fourth resolution feature image respectively.

[0012] According to a specific embodiment, in the above detection method, a small target dataset is constructed based on the small target defect image of the catenary dropper string, specifically including: introducing random Poisson noise into the small target defect image, and performing data enhancement and data expansion to obtain a data set; The dataset is divided into a training set, a validation set, and a test set to generate a small target dataset.

[0013] Compared with the prior art, the present invention has the following beneficial effects: The detection model based on the improved YOLOv11s adds a small target detection layer to the neck network and combines it with the improved detection head network to achieve multi-scale feature fusion, which can detect small targets more accurately and reduce missed detections and false detections; the present invention embeds the SEAM attention mechanism and utilizes depthwise separable convolution to effectively extract small target features in occluded areas, thereby improving the robustness, reliability and integrity of detection in complex environments; although the average detection time is slightly increased, the small target detection capability is greatly improved without affecting the detection accuracy of other categories, and the comprehensive detection performance is better, providing reliable technical support for high-speed railway contact network dropper fault detection and ensuring the safe operation of high-speed railways. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 A schematic diagram of a flow chart of a method for detecting the state of a catenary dropper based on an improved YOLOv11s according to an embodiment of the present invention; Figure 2 A schematic diagram of the structure of the improved YOLOv11s algorithm network provided by an embodiment of the present invention; Figure 3 A schematic diagram of the structure of the SEAM attention module provided by an embodiment of the present invention; Figure 4 A schematic structural diagram of the DASFFHead detection head module provided in an embodiment of the present invention; Figure 5 A schematic diagram comparing the average detection accuracy of different algorithms provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0015] The present invention will be further described in detail below with reference to specific embodiments. However, this should not be construed as limiting the scope of the present invention to the following embodiments, as all technologies implemented based on the present invention fall within the scope of the present invention.

[0016] Unless otherwise specified, in the description of the specific embodiments of the present invention, the expressions "first", "second", "third", etc. are merely used to distinguish the descriptions of the same or similar components and should not be understood as emphasizing or implying the relative importance of specific components.

[0017] The purpose of this invention is to achieve accurate and efficient catenary dropper status recognition and small target detection, improving the accuracy and efficiency of catenary detection and ensuring the stable operation of high-speed railway power supply systems. Therefore, it is necessary to implement multi-scale feature fusion to adapt the model to detect dropper targets of different sizes, and to extract the features of small targets in occluded areas to overcome detection obstacles caused by occlusion.

[0018] The technical solution provided by the present invention is introduced and explained in detail below in conjunction with specific implementation methods.

[0019] Please refer to Figure 1 , which shows a flow chart of a method for detecting the state of a catenary dropper based on an improved YOLOv11s according to an embodiment of the present invention, the method comprising: Step 1: Construct a small target dataset based on the small target defect images of the catenary suspension strings.

[0020] Among them, building a small target data set specifically includes: introducing random Poisson noise into the small target defect image, and performing data enhancement and data expansion to obtain a data set; The dataset is divided into a training set, a validation set, and a test set to generate a small target dataset.

[0021] In one possible implementation, random Poisson noise is introduced to adapt the training images to low-light conditions, given the complex operating environment and the unique conditions for capturing images at night. Simultaneously, a variety of data augmentation methods, such as brightness enhancement, rotation, translation, cropping, mirroring, and occlusion, are used to augment the training images and generate a dataset.

[0022] Specifically, the dataset includes four defect types: normal dropper strings, bent dropper strings, current ring faults, and minor defects. For this example, 800 images are taken of normal dropper strings, 400 each of bent dropper strings, current ring faults, and minor defects, all at a resolution of 5120×5120 pixels, for a total of 2000 images. This 2000-image dataset is then divided into training, validation, and test sets in an 8:1:1 ratio, completing the construction of the small target dataset.

[0023] Step 2: Input the small target dataset into the constructed improved YOLOv11s algorithm network for training to obtain a trained detection model.

[0024] In one possible implementation, first build the YOLOv11s basic network. Then construct a small target detection layer, add the small target detection layer to the Neck network, and introduce the SEAM attention module to obtain an improved Neck network; the improved Neck network is used to obtain feature images of four different channels from the Backbone network and output feature images of four different resolutions. Finally, the Head detection head in the Head network is combined with the DASFF module to improve it into a four-head DASFFHead detection head to obtain an improved YOLOv11s algorithm network; the DASFF module is used to fuse feature images of four different resolutions so that the Head network outputs small target detection results. Please refer to Figure 2 , which shows a schematic structural diagram of the improved YOLOv11s algorithm network provided by an embodiment of the present invention.

[0025] Specifically, the YOLOv11s basic network includes a Backbone network, a Neck network, and a Head network. The Backbone network consists of two Conv modules, a C3k2 module, a Conv module, a C3k2 module, a Conv module, a C3k2 module, a Conv module, a C3k2 module, a SPPF module, and a C2PSA module connected in sequence. After the image is input into the Backbone network, the first C3k2 module in the Backbone network outputs a first channel feature image, the second C3k2 module in the Backbone network outputs a second channel feature image, the third C3k2 module in the Backbone network outputs a third channel feature image, and the C2PSA module in the Backbone network outputs a fourth channel feature image.

[0026] Furthermore, the small target detection layer includes a first C3k2 module, a second C3k2 module, a first Concat module, a second Concat module, a first Conv module and a first Upsample module. The improved Neck network is used to upsample the fourth channel feature image, fuse the features with the third channel feature image and extract the features before upsampling, and then fuse the features with the second channel feature image, extract the features and upsample through the first Upsample module, fuse the features with the first channel feature image through the first Concat module, and then extract the features through the first C3k2 module to obtain the first resolution feature image and output it to the DASFF module. On this basis, the improved Neck network is also used to downsample the features extracted by the first C3k2 module through the first Conv module, fuse the features with the input of the first Upsample module through the second Concat module, and then extract the features through the second C3k2 module, input it to the SEAM attention module, obtain the second resolution feature image, and output it to the DASFF module. In a possible implementation method, please refer to Figure 3 , which shows a schematic diagram of the structure of the SEAM attention module provided by an embodiment of the present invention. The SEAM attention module is used to extract small target features in occluded areas through depthwise separable convolution with residual connections.

[0027] Furthermore, the improved Neck network is also used to upsample the fourth channel feature image, fuse the features with the third channel feature image and extract the features, and then fuse the features with the output of the downsampled SEAM attention module to extract the features to obtain a third resolution feature image, and output it to the DASFF module; then downsample the third resolution feature image, fuse the features with the fourth channel feature image, extract the features to obtain a fourth resolution feature image, and output it to the DASFF module.

[0028] Furthermore, the DASFFHead detection head uses the DASFF module to receive inputs of four different scales and fuse the first resolution feature image to the fourth resolution feature image. Figure 4 , which shows a schematic structural diagram of the DASFFHead detection head provided by an embodiment of the present invention.

[0029] Understandably, DASFF enables the network to adaptively learn how to filter features from other levels in space, retaining only useful information for combination. At each spatial location, features from different levels are adaptively fused, which means that some features may be filtered out because they carry conflicting information, while others may dominate detection and provide more discriminative clues.

[0030] It's understandable that the head detection head in the YOLOv11s base network typically only uses head modules of three different scales. However, the present invention adds a higher-resolution feature map output to the small object detection layer, enabling the detection of small objects. Specifically, the DASFF module employs dynamic adaptive spatial feature fusion technology, achieving pixel-by-pixel feature fusion through a dynamic spatial weighting mechanism (1x1 convolution + Softmax). Combined with lightweight channel compression (8 / 16 dimensions) and a cross-layer adaptive sampling strategy (high-level convolutional pooling downsampling / low-level interpolation upsampling), this constructs a four-head detection architecture to enhance multi-scale feature interaction, significantly improving small object detection performance while maintaining computational efficiency. For example, the three head detection heads in the YOLOv11s base network have strides of 8, 16, and 32, respectively. Building on this, the present invention sets the strides of the four DASFF head detection heads to 4, 8, 16, and 32, respectively. The addition of an additional DASFF module with a stride of 4 extracts information from the early feature maps of the backbone network, preserving more spatial detail. Small objects may only occupy a few pixels in low-resolution feature maps, making it easy to lose details. By introducing a high-resolution detection layer, the model can capture more shallow features (such as edges and textures), thereby improving the detection ability of small objects.

[0031] In one possible implementation, an embodiment of the present invention provides a training example of an improved YOLOv11s algorithm network, wherein, in the training parameters, the batch size is 16, the initial learning rate is set to 0.01, the final learning rate is 0.01, the momentum parameter is set to 0.937, the weight decay is 0.0005, the input image is uniformly converted to a size of 640x640, and a stochastic gradient descent optimizer is used.

[0032] Step 3: Use the trained detection model to identify the catenary string image collected on site, obtain the small target detection result, and complete the status detection.

[0033] Specifically, please refer to Figure 5 , which shows a schematic diagram comparing the average detection accuracy of different algorithms provided by embodiments of the present invention. The improved YOLOv11s algorithm network provided by embodiments of the present invention improved the overall mAP@0.5 detection accuracy of dropper strings by 1.3%, with the AP value for detection of dropper strings with small defects increasing by 8.7%.

[0034] Based on the above technical solution, the present invention realizes multi-scale feature fusion by adding a small target detection layer to the neck network based on the improved YOLOv11s detection model, and combines it with the improved detection head network, which can detect small targets more accurately and reduce missed detections and false detections; the present invention embeds the SEAM attention mechanism and utilizes deep separable convolution to effectively extract small target features in occluded areas, thereby improving the robustness, reliability and integrity of detection in complex environments; although the average detection time is slightly increased, the small target detection capability is greatly improved without affecting the detection accuracy of other categories, and the comprehensive detection performance is better, providing reliable technical support for high-speed railway contact network dropper fault detection and ensuring the safe operation of high-speed railways.

[0035] The methods, steps, and logic diagrams disclosed in the embodiments of the present invention can be implemented or executed. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of the present invention can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software modules within the decoding processor. The software modules can be located in a storage medium well-established in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. The processor reads the information from the storage medium and, in conjunction with its hardware, completes the steps of the aforementioned methods.

[0036] Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present invention should not be interpreted as being more preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner to facilitate understanding.

[0037] 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 scope of protection of the present invention.

Claims

1. A method for detecting the state of a catenary dropper based on improved YOLOv11s, characterized in that: The method comprises: A small target dataset is constructed based on small target defect images of catenary suspension strings; Input the small target data set into the constructed improved YOLOv11s algorithm network for training to obtain a trained detection model; The trained detection model is used to identify the catenary string images collected on site, obtain small target detection results, and complete status detection. The construction process of the improved YOLOv11s algorithm network includes: Build a YOLOv11s basic network; the YOLOv11s basic network includes a Backbone network, a Neck network, and a Head network; A small target detection layer is constructed, added to the Neck network, and the SEAM attention module is introduced to obtain an improved Neck network; the improved Neck network is used to obtain feature images of four different channels from the Backbone network and output feature images of four different resolutions; The Head detection head in the Head network is combined with the DASFF module to be improved into a four-head DASFFHead detection head to obtain an improved YOLOv11s algorithm network; the DASFF module is used to fuse feature images of four different resolutions, so that the Head network outputs small target detection results.

2. The method for detecting the state of a catenary dropper based on improved YOLOv11s according to claim 1, characterized in that: After the original image is input into the Backbone network, the first channel feature image is output through the first C3k2 module in the Backbone network, the second channel feature image is output through the second C3k2 module in the Backbone network, the third channel feature image is output through the third C3k2 module in the Backbone network, and the fourth channel feature image is output through the C2PSA module in the Backbone network.

3. The method for detecting the state of a catenary dropper based on improved YOLOv11s according to claim 2, characterized in that: The small target detection layer includes a first C3k2 module, a second C3k2 module, a first Concat module, a second Concat module, a first Conv module and a first Upsample module.

4. The method for detecting the state of a catenary dropper based on improved YOLOv11s according to claim 3, characterized in that: The improved Neck network is used to upsample the fourth channel feature image, fuse features with the third channel feature image, extract features, and then upsample. The improved Neck network is used to fuse features with the second channel feature image, extract features, and upsample through the first Upsample module. The improved Neck network is used to fuse features with the first channel feature image through the first Concat module, and extract features through the first C3k2 module to obtain a first-resolution feature image, which is output to the DASFF module.

5. The method for detecting the state of a catenary dropper based on improved YOLOv11s according to claim 4, characterized in that: The improved Neck network is also used to downsample the features extracted by the first C3k2 module through the first Conv module, fuse the features with the input of the first Upsample module through the second Concat module, extract the features through the second C3k2 module, input them into the SEAM attention module, obtain the second resolution feature image, and output it to the DASFF module.

6. The method for detecting the state of a catenary dropper based on improved YOLOv11s according to claim 5, characterized in that: The SEAM attention module is used to extract small object features in occluded areas through depthwise separable convolution with residual connections.

7. The method for detecting the state of a catenary dropper based on improved YOLOv11s according to claim 5, characterized in that: The improved Neck network is also used to upsample the fourth channel feature image, fuse the features with the third channel feature image and extract the features, and then fuse the features with the output of the downsampled SEAM attention module to extract the features to obtain a third resolution feature image, and output it to the DASFFHEAD detection head; then downsample the third resolution feature image, fuse the features with the fourth channel feature image, extract the features to obtain a fourth resolution feature image, and output it to the DASFF module.

8. The method for detecting the state of a catenary dropper based on improved YOLOv11s according to claim 7, characterized in that: The DASFFHead detection head adopts the DASFF module to receive inputs of four different scales and fuses the first resolution feature image to the fourth resolution feature image respectively.

9. The method for detecting the state of a catenary dropper based on improved YOLOv11s according to claim 1, characterized in that: A small target dataset is constructed based on the small target defect images of the catenary dropper strings, including: introducing random Poisson noise into the small target defect image, and performing data enhancement and data expansion to obtain a data set; The dataset is divided into a training set, a validation set, and a test set to generate a small target dataset.

Images