A manifold learning approach to urban land cover classification with optical and radar data

The usage of Unmanned Aerial Vehicle (UAV) has grown rapidly in various fields, such as urban planning, search and rescue, and surveillance. Capturing images from UAV has many advantages compared with satellite imagery. For instance, higher spatial resolution and less impact from atmospheric variations can be obtained. However, there are difficulties in classifying urban features, due to the complexity of the urban land covers. The usage of Maximum Likelihood Classification (MLC) has limitations since it is based on the assumption of the normal distribution of pixel values, where, in fact, urban features are not normally distributed. There are advantages in using the Markov Random Field (MRF) for urban land cover classification as it assumes that neighboring pixels have a higher probability to be classified in the same class rather than a different class. This research aimed to determine the impact of the smoothness (λ) and the updating temperature (Tupd) on the accuracy result (κ) in MRF. We used a UAV VHIR sized 587 square meters, with six-centimetre resolution, taken in Bogor Regency, Indonesia. The result showed that the kappa value (κ) increases proportionally with the smoothness (λ) until it reaches the maximum (κ), then the value drops. The usage of higher (Tupd) has resulted in better (κ) although it also led to a higher Standard Deviations (SD). Using the most optimal parameter, MRF resulted in slightly higher (κ) compared with MLC.

Download Full-text

Improving Urban Land Cover Classification Using Fuzzy Image Segmentation

Transactions on Computational Science VI - Lecture Notes in Computer Science ◽

10.1007/978-3-642-10649-1_3 ◽

2009 ◽

pp. 41-56 ◽

Cited By ~ 2

Author(s):

Ivan Lizarazo ◽

Paul Elsner

Keyword(s):

Image Segmentation ◽

Land Cover ◽

Land Cover Classification ◽

Urban Land ◽

Urban Land Cover ◽

Fuzzy Image

Download Full-text

Urban Land Cover Classification of High-Resolution Aerial Imagery Using a Relation-Enhanced Multiscale Convolutional Network

Remote Sensing ◽

10.3390/rs12020311 ◽

2020 ◽

Vol 12 (2) ◽

pp. 311 ◽

Cited By ~ 6

Author(s):

Chun Liu ◽

Doudou Zeng ◽

Hangbin Wu ◽

Yin Wang ◽

Shoujun Jia ◽

...

Keyword(s):

Remote Sensing ◽

High Resolution ◽

Land Cover ◽

Land Cover Classification ◽

Urban Land ◽

Small Sample ◽

Connectivity Pattern ◽

Feature Maps ◽

Convolutional Network ◽

Urban Land Cover

Urban land cover classification for high-resolution images is a fundamental yet challenging task in remote sensing image analysis. Recently, deep learning techniques have achieved outstanding performance in high-resolution image classification, especially the methods based on deep convolutional neural networks (DCNNs). However, the traditional CNNs using convolution operations with local receptive fields are not sufficient to model global contextual relations between objects. In addition, multiscale objects and the relatively small sample size in remote sensing have also limited classification accuracy. In this paper, a relation-enhanced multiscale convolutional network (REMSNet) method is proposed to overcome these weaknesses. A dense connectivity pattern and parallel multi-kernel convolution are combined to build a lightweight and varied receptive field sizes model. Then, the spatial relation-enhanced block and the channel relation-enhanced block are introduced into the network. They can adaptively learn global contextual relations between any two positions or feature maps to enhance feature representations. Moreover, we design a parallel multi-kernel deconvolution module and spatial path to further aggregate different scales information. The proposed network is used for urban land cover classification against two datasets: the ISPRS 2D semantic labelling contest of Vaihingen and an area of Shanghai of about 143 km2. The results demonstrate that the proposed method can effectively capture long-range dependencies and improve the accuracy of land cover classification. Our model obtains an overall accuracy (OA) of 90.46% and a mean intersection-over-union (mIoU) of 0.8073 for Vaihingen and an OA of 88.55% and a mIoU of 0.7394 for Shanghai.

Download Full-text