Optimized Sample Selection in SVM Classification by Combining with DMSP-OLS, Landsat NDVI and GlobeLand30 Products for Extracting Urban Built-Up Areas

Ma, Xiaolong; Tong, Xiaohua; Luo, Xin; Xie, Huan; Li, Chengming

doi:10.3390/rs9030236

Cited by 56 publications

(33 citation statements)

References 51 publications

(52 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Their methodology combines nighttime-light data and Landsat 8 images and overcomes the lack of extensive ground reference data. Ma et al [58] presented a sample-optimized approach for classifying urban area data in several cities of western China using a combination of the Defense Meteorological Satellite Program (DMSP)-Operational Linescan System (OLS) for nighttime-light data, Landsat images, and GlobeLand30. Goldblatt et al [59] applied a classification and regression tree, SVM, and RFs to extract urban areas in India based on a single pixel using the GEE platform.…”

Section: Built-up Area Extraction From Medium-resolution Imagesmentioning

confidence: 99%

A Comprehensive Evaluation of Approaches for Built-Up Area Extraction from Landsat OLI Images Using Massive Samples

Zhang

Tang

2018

Remote Sensing

View full text Add to dashboard Cite

Detailed information about built-up areas is valuable for mapping complex urban environments. Although a large number of classification algorithms for such areas have been developed, they are rarely tested from the perspective of feature engineering and feature learning. Therefore, we launched a unique investigation to provide a full test of the Operational Land Imager (OLI) imagery for 15-m resolution built-up area classification in 2015, in Beijing, China. Training a classifier requires many sample points, and we proposed a method based on the European Space Agency's (ESA) 38-m global built-up area data of 2014, OpenStreetMap, and MOD13Q1-NDVI to achieve the rapid and automatic generation of a large number of sample points. Our aim was to examine the influence of a single pixel and image patch under traditional feature engineering and modern feature learning strategies. In feature engineering, we consider spectra, shape, and texture as the input features, and support vector machine (SVM), random forest (RF), and AdaBoost as the classification algorithms. In feature learning, the convolutional neural network (CNN) is used as the classification algorithm. In total, 26 built-up land cover maps were produced. The experimental results show the following: (1) The approaches based on feature learning are generally better than those based on feature engineering in terms of classification accuracy, and the performance of ensemble classifiers (e.g., RF) are comparable to that of CNN. Two-dimensional CNN and the 7-neighborhood RF have the highest classification accuracies at nearly 91%; (2) Overall, the classification effect and accuracy based on image patches are better than those based on single pixels. The features that can highlight the information of the target category (e.g., PanTex (texture-derived built-up presence index) and enhanced morphological building index (EMBI)) can help improve classification accuracy. The code and experimental results are available at https://github.com/zhangtao151820/CompareMethod. challenge for remote sensing due to the high spatial frequency and heterogeneity of surface features. Various algorithms have been applied to extract built-up areas, including supervised classification, unsupervised clustering, and reinforcement learning [9][10][11]. In the process of built-up area extraction, the most important considerations are how to design or learn better features to characterize the buildings and how to choose a more appropriate classification strategy. Li et al. [12] tested two unsupervised and 13 supervised classification algorithms to distinguish urban land in Guangzhou City and then assessed all the algorithms in a per-pixel classification decision experiment and all the supervised algorithms in a segment-based experiment. Momeni et al. [13] compared the influence of spatial resolution, spectral band set, and the classification approach for mapping detailed urban land cover in Nottingham, UK. A WorldView-2 image provides the basis for a set of 12 images with variable spatial and s...

show abstract

Section: Built-up Area Extraction From Medium-resolution Imagesmentioning

confidence: 99%

A Comprehensive Evaluation of Approaches for Built-Up Area Extraction from Landsat OLI Images Using Massive Samples

Zhang

Tang

2018

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…LULC Map Classification. LULC status are categorized into five classes using the support vector machine (SVM) classifier [10] according to the Classification Criteria for Land Use Status/GB-T21010-2015 and GlobeLand30 standard products [11]. The GlobelLand30 is an open-access 30 m resolution global land cover data product which comprises ten land cover types, including water bodies, wetlands, artificial surfaces, cultivated lands, forests, shrub lands, grasslands, and barren lands [12].…”

Section: Methodsmentioning

confidence: 99%

Landscape Pattern and Ecological Security Assessment and Prediction Using Remote Sensing Approach

et al. 2018

View full text Add to dashboard Cite

In this work, we present a processing chain for landscape pattern and ecological security status assessment and prediction based on cellular automata Markov (CA-Markov) and pressure status response pattern (PSRP) models using remotely sensed data (RSD) captured in 1986, 1996, 2006, 2016, and RSD simulated in 2026 over Zhengzhou city, Henan province, China. Three major findings can be withdrawn through the experiments. First, there is a significant changing of landscape type area, especially for building land. The area of building land is up to more than 5%, from 1986 to 2016. Secondly, the heterogeneity of landscape is increasing, and the diversity of landscape is becoming more and more diversifying and complex. Third, the changing trend of ecological security of Zhengzhou city shaped as decreasing and increasing gradually during the last 40 years. While the ecological security status, nowadays, appeared to a good trend by contrast of the previous stages. The predicted results with CAMarkov model show that the level of ecological security is still in moderate and has a trend of moving toward to the center in 2026.

show abstract

“…Its operational linescan system sensors operate at night to detect low-intensity lights from city lights, small-scale settlements, and traffic, making the lights distinct from the dark countryside background [29,30]. The DMSP/OLS nighttime light data in this study was obtained from The National Oceanic and Atmospheric Administration's National Geophysical Data Center (NOAA/NGDC) website with 1 km image resolution [31].…”

Section: Data Sourcesmentioning

confidence: 99%

Mapping Development Pattern in Beijing-Tianjin-Hebei Urban Agglomeration Using DMSP/OLS Nighttime Light Data

Peng

Liu

et al. 2017

Remote Sensing

View full text Add to dashboard Cite

Spatial inequality of urban development may cause problems like inequality of living conditions and the lack of sustainability, drawing increasing academic interests and societal concerns. Previous studies based on statistical data can hardly reveal the interior mechanism of spatial inequality due to the limitation of statistical units, while the application of remote sensing data, such as nighttime light (NTL) data, provides an effective solution. In this study, based on the DMSP/OLS NTL data, the urbanization type of all towns in the Beijing-Tianjin-Hebei urban agglomeration was analyzed from the aspects of development level and speed. Meanwhile, spatial cluster analysis of development level by local Moran's I was used to explore spatial inequality, and the trend was discussed by comparing the development characteristics on both sides of the transition line of different development levels (inequality boundary). The results showed that the development level of the whole region increased dramatically as the mean DN value increased by 65.99%, and 83.72% of the towns showed a positive development during 2000-2012. The spatial distribution of urbanization types showed that Beijing and Tianjin were at a high urbanization level with rapid speed of development, with the southern region having a medium development level and the northwestern region lagging behind. The spatial cluster analysis also revealed a gradually intensifying trend of inequality as the number of towns with balanced development reduced by 319 during 2000-2012, while the towns in the high-high areas increased by 99 and those in the low-low areas increased by 229. Moreover, the development speed inside the inequality boundary was obviously higher than that outside, indicating an increasingly serious situation for spatial inequality of urban development in the whole region.

show abstract

Optimized Sample Selection in SVM Classification by Combining with DMSP-OLS, Landsat NDVI and GlobeLand30 Products for Extracting Urban Built-Up Areas

Cited by 56 publications

References 51 publications

A Comprehensive Evaluation of Approaches for Built-Up Area Extraction from Landsat OLI Images Using Massive Samples

A Comprehensive Evaluation of Approaches for Built-Up Area Extraction from Landsat OLI Images Using Massive Samples

Landscape Pattern and Ecological Security Assessment and Prediction Using Remote Sensing Approach

Mapping Development Pattern in Beijing-Tianjin-Hebei Urban Agglomeration Using DMSP/OLS Nighttime Light Data

Contact Info

Product

Resources

About