Multi-Resolution Transformer Network for Building and Road Segmentation of Remote Sensing Image
نویسندگان
چکیده
Extracting buildings and roads from remote sensing images is very important in the area of land cover monitoring, which great help to urban planning. Currently, a deep learning method used by majority building road extraction algorithms. However, for existing semantic segmentation, it has limitation on receptive field high-resolution images, means that can not show long-distance scene well during pixel classification, image features compressed down-sampling, meaning detailed information lost. In order address these issues, Hybrid Multi-resolution Transformer Network (HMRT) proposed this paper, global each be provided, small convolutional neural networks (CNN) overcome, ability understanding enhanced well. Firstly, we blend branches different resolutions keep multi-resolution down-sampling fully retain feature information. Secondly, introduce sequence network use encoding decoding realize field. The recall, F1, OA MIoU HMPR obtain 85.32%, 84.88%, 85.99% 74.19%, respectively, main experiment reach 91.29%, 90.41%, 91.32% 84.00%, generalization experiment, prove better than methods.
منابع مشابه
Multi-agent Remote Sensing Image Segmentation Algorithm
Due to fractal network evolution algorithm (FNEA) in the treatment of the high spatial resolution remote sensing image (HSRI) using a parallel global control strategies which limited when the objects in each cycle by traversal of and not good use the continuity of homogenous area on the space and lead to problems such as bad image segmentation, therefore puts forward the remote sensing image se...
متن کاملMulti-scale and Multi-feature Segmentation of High Resolution Remote Sensing Image
With the development of the remote sensing technology, high resolution remote sensing images widely penetrates into the common people’s life. Traditional medium or low resolution image processing method based on pixel doesn't meet people’s requirement any more. In view of it, this paper puts forward a high resolution remote sensing image segmentation method based on the traditional watershed al...
متن کاملMulti-spectral Texture Characterisation for Remote Sensing Image Segmentation
A multi-spectral texture characterisation model is proposed, the Multi-spectral Local Differences Texem – MLDT, as an affordable approach to be used in multi-spectral images that may contain large number of bands. The MLDT is based on the Texem model. Using an inter-scale post-fusion strategy for image segmentation, framed in a multi-resolution approach, we produce unsupervised multi-spectral i...
متن کاملRoad Extraction from Remote Sensing Image Based on Multi-resolution Analysis
In the paper, we present an approach of multi-resolution analysis to extract road information from high resolution imagery. Firstly, high resolution imagery is divided into a series of different resolution images. According to the representation characteristics of roads at different resolution, we analyze roads’ geometrical, radiometric, and topological attributes and carry out a variety of pro...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: ISPRS international journal of geo-information
سال: 2022
ISSN: ['2220-9964']
DOI: https://doi.org/10.3390/ijgi11030165