Remote Sensing Image Road Extraction Network Based on MSPFE-Net

Abstract

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. However, many models using convolutional neural networks ignore the attributes of roads, and the shape of the road is banded and discrete. In addition, the continuity and accuracy of road extraction are also affected by narrow roads and roads blocked by trees. This paper designs a network (MSPFE-Net) based on multi-level strip pooling and feature enhancement. The overall architecture of MSPFE-Net is encoder-decoder, and this network has two main modules. One is a multi-level strip pooling module, which aggregates long-range dependencies of different levels to ensure the connectivity of the road. The other module is the feature enhancement module, which is used to enhance the clarity and local details of the road. We perform a series of experiments on the dataset, Massachusetts Roads Dataset, a public dataset. The experimental data showed that the model in this paper was better than the comparison models.

Improvements in Forest Segmentation Accuracy Using a New Deep Learning Architecture and Data Augmentation Technique

“…Moreover, the robustness of the network was further improved after training with the proposed data augmentation method. Therefore, we believe that the proposed data augmentation algorithm could be applied to similar binary classification remote sensing image segmentation tasks, such as road extraction [56], water extraction [57], agricultural land mapping [58], and more. Furthermore, we also believe that the proposed network could also be applied to other remote sensing image segmentation tasks and even to non-remote sensing tasks that require small detail feature information extraction.…”

A Survey of Deep Learning Road Extraction Algorithms Using High-Resolution Remote Sensing Images

“…Inspired by dense convolution, Q. Wu et al [ 56 ] introduced the dense and global spatial pyramid pooling module (DGSPP) into the decoder and encoder to enhance the network’s perception and aggregation of contextual information. Wei and Zhang [ 57 ] integrated the multi-level strip pooling module (MSPM) into the skip connection layers to ensure road connectivity by aggregating long-range dependencies from different levels. LinkNet used ResNet-18 as the encoder backbone and improved segmentation efficiency by directly connecting the encoder and decoder.…”

“…To address the issue of sample imbalance, the focal loss function has been employed by some researchers [ 28 , 89 , 134 ]. Additionally Wei and Zhang [ 57 ] combined focal loss with the dice function. The focal loss function [ 135 ] differs from traditional cross entropy functions by focusing on resolving sample imbalances and confounding pixel categories.…”

Computational Intelligence in Remote Sensing

“…The second paper, written by Z. Wei and Z. Zhang, describes a network built on multi-level strip pooling and a feature enhancement module (MSPFE-Net). Here, deep learning is effectively applied to address the challenge of road extraction [8]. In the third paper, L. Zeng and Y. Huo et al develop the high-quality seed instance mining (HSIM) module, alongside the dynamic pseudo-instance label assignment (DPILA), to address the issue of weakly supervised detection in remote sensing images [9].…”