![]() The feature enhancement module is used to enhance the clarity and local details of the road. In MSPFE-Net, the multi-level strip pooling module is responsible for fully extracting long-range context information. Therefore, a multi-level strip pooling and feature enhancement network (MSPFE-Net) called MSPFE-Net is designed in this paper. The idea of strip pooling is well applied in remote sensing image road extraction scenes. It is fundamentally different from traditional spatial pooling. The network combined with strip pooling has the ability to obtain multiple types of contexts. Then, strip pooling maintains a narrow shape along a spatial dimension, which helps to capture the local feature of targets and can reduce the interference of irrelevant target information. Firstly, the strip pooling has a long and banded shape at a dimension, so it can capture long-range relationships of isolated regions. The strip pooling has several distinct characteristics. , this paper introduces and improves upon it. Inspired by the idea of strip pooling proposed by Hou, Feng et al. Other methods include atrous convolution and spatial pyramid pooling, which can expand the receptive field of convolutional neural network, but the strip target features extracted by the square window may be mixed with irrelevant target information. The attention mechanism is a method to improve the ability of global context modeling. In order to solve these problems, road extraction models need strong long-distance dependencies or global context information, and the road extraction algorithm usually uses attention mechanism or atrous convolution technology to obtain long-distance dependencies or global context information. Although the model based on deep learning has achieved good results in extracting road tasks, and many road extraction algorithms have problems, such as road breaking caused by occlusion, difficult extraction of the narrow road, and incorrect identification of roads and background. Scholars also began to use deep learning technology to complete remote sensing image road extraction. The experimental data showed that the model in this paper was better than the comparison models.ĭeep learning promotes the progress of computer vision, especially in object detection, semantic segmentation, image classification, and other aspects, and it has a good effect. We perform a series of experiments on the dataset, Massachusetts Roads Dataset, a public dataset. The other module is the feature enhancement module, which is used to enhance the clarity and local details of the road. One is a multi-level strip pooling module, which aggregates long-range dependencies of different levels to ensure the connectivity of the road. The overall architecture of MSPFE-Net is encoder-decoder, and this network has two main modules. This paper designs a network (MSPFE-Net) based on multi-level strip pooling and feature enhancement. ![]() ![]() In addition, the continuity and accuracy of road extraction are also affected by narrow roads and roads blocked by trees. However, many models using convolutional neural networks ignore the attributes of roads, and the shape of the road is banded and discrete. Road extraction is a hot task in the field of remote sensing, and it has been widely concerned and applied by researchers, especially using deep learning methods. ![]()
0 Comments
Leave a Reply. |