Urban land-use mapping using a deep convolutional neural network with high spatial resolution multispectral remote sensing imagery

Huang, Bo; Zhao, Binghai; Song, Yimeng

doi:10.1016/j.rse.2018.04.050

Cited by 449 publications

(222 citation statements)

References 37 publications

Supporting

Mentioning

193

Contrasting

Unclassified

Order By: Relevance

“…Although the classification accuracy of the present study exceeded 80%, ours and the above studies are subject to limitations determined by the availability and resolution of remote imagery, as well as the presence of clouds and gaps, the difficulty in obtaining on-ground reference data, and time and cost constraints (Christiansen and Schöner 2004, Yan and Roy 2014, Zou et al 2017, Hao et al 2018, Hu and Hu, 2019. A variety of classification and analytical methods, including object-based systems (Arvor et al 2013), convolutional neural networks (Huang et al 2018), and reference time-series-based mapping (Hao et al 2018), have been devised to address these issues (Hansen and Loveland 2012). Difficulties in interpreting spectral signatures in Central Asia generally relate to the inability to distinguish cropland from shrubland, grassland, and/or bare areas (Hao et al 2018, Hu andHu 2019).…”

Section: Discussionmentioning

confidence: 61%

Changes in land use/land cover and net primary productivity in the transboundary Ili-Balkhash basin of Central Asia, 1995–2015

Shen

Pueppke

et al. 2019

Environ. Res. Commun.

View full text Add to dashboard Cite

Using remotely sensed data, we documented changes in land use/land cover (LULC) from 1995 to 2015 and net primary productivity (NPP) from 2000 to 2014 in Central Asia's 415,048 km 2 Ili-Balkhash Basin (IBB). This basin, which is shared by China and Kazakhstan, is in the midst of significant socioeconomic transformation due to the collapse of the Soviet Union, the emergence of Kazakhstan, and the economic rise of China. Grazing land covered 82.4% of the IBB in 1995; water bodies and bare land were the only other LULC categories occupying more than 3% of the basin's area. Changes in LULC were most evident on the Chinese side of the border, where crop production areas increased and grazing areas decreased between 1995 and 2015. The area of irrigated cropland in China grew by nearly 30%, primarily in the upper Tekes River valley and along the Ili River near the border with Kazakhstan. In contrast, the irrigated lands in Kazakhstan shifted geographically during this period, but the extent did not change. Expansion of wetlands and permanent water bodies, which occupied 2.1 to 2.9% and 4.6 to 4.7%, respectively, of the IBB in 1995 and 2015, was associated with accretion of Lake Balkhash and Kapchagai Reservoir in Kazakhstan and the construction of new reservoirs in China. NPP of the basin approached 700 g C m −2 /year in a few areas but was generally less than half this level and characterized by a declining trend except in highly productive irrigated areas of dense, stable vegetation. NPP decreases of more than −10 g C m −2 /year were apparent in mountainous and upland areas, as well as broad band of grassland and cropland in Kazakhstan. Areas surrounding Lake Balkhash were characterized by unstable to moderately stable, often sparse vegetation.

show abstract

Section: Discussionmentioning

confidence: 61%

Changes in land use/land cover and net primary productivity in the transboundary Ili-Balkhash basin of Central Asia, 1995–2015

Shen

Pueppke

et al. 2019

Environ. Res. Commun.

View full text Add to dashboard Cite

show abstract

“…The class accuracies are shown in Figure 8; averaged OA and AA values are given in Table 1. By comparing our multimodal model against the unimodal variants, we observe an increase of around 6% for OA and more than 7% for AA against the VGG16-based model trained on overhead imagery, while Specifically, AlexNet [49] that was used in [50] to perform landuse mapping with mutltispectral remote sensing images and ResNet50 [51] that was used in [52] to do large-scale land cover classification of satellite imagery. The results of these methods are presented in Table 2.…”

Section: Joint Cnn Trainingmentioning

confidence: 99%

Understanding urban landuse from the above and ground perspectives: A deep learning, multimodal solution

Srivastava

Muñoz

Tuia

2019

Remote Sensing of Environment

121

View full text Add to dashboard Cite

This is the pre-acceptance version, to read the final version published in the journal Remote Sensing of Environment, please go to: https:// doi.org/10.1016/j.rse.2019.04.014 Landuse characterization is important for urban planning. It is traditionally performed with field surveys or manual photo interpretation, two practices that are time-consuming and labor-intensive.Therefore, we aim to automate landuse mapping at the urban-object level with a deep learning approach based on data from multiple sources (or modalities).We consider two image modalities: overhead imagery from Google Maps and ensembles of ground-based pictures (side-views) per urban-object from Google Street View (GSV). These modalities bring complementary visual information pertaining to the urban-objects. We propose an end-to-end trainable model, which uses OpenStreetMap annotations as labels. The model can accommodate a variable number of GSV pictures for the ground-based branch and can also function in the absence of ground pictures at prediction time. We test the effectiveness of our model over the area ofÎle-de-France, France, and test its generalization abilities on a set of urban-objects from the city of Nantes, France.Our proposed multimodal Convolutional Neural Network achieves considerably higher accuracies than methods that use a single image modality, making it * First Author: SS, shivangi.srivastava@wur.nl.

show abstract

“…Thus, many supervised algorithms have been proposed to accurately discriminate land cover classes in remote sensing images, including support vector machines, random forests, and neural networks [1]- [6]. Recently, deep learning, particularly deep convolutional neural networks (CNNs), has achieved considerable progress in remote sensing image classification, including hyperspectral image classification, and has achieved state-of-the-art results [7]- [12].…”

Section: Introductionmentioning

confidence: 99%

Multitask Deep Learning With Spectral Knowledge for Hyperspectral Image Classification

Liu

Shi

2020

IEEE Geosci. Remote Sensing Lett.

View full text Add to dashboard Cite

In this letter, we propose a multitask deep learning method for classification of multiple hyperspectral data in a single training. Deep learning models have achieved promising results on hyperspectral image classification, but their performance highly rely on sufficient labeled samples, which are scarce on hyperspectral images. However, samples from multiple data sets might be sufficient to train one deep learning model, thereby improving its performance. To do so, we trained an identical feature extractor for all data, and the extracted features were fed into corresponding softmax classifiers. Spectral knowledge was introduced to ensure that the shared features were similar across domains. Four hyperspectral data sets were used in the experiments. We achieved higher classification accuracies on three data sets (Pavia University, Pavia Center, and Indian Pines) and competitive results on the Salinas Valley data compared with the baseline. Spectral knowledge was useful to prevent the deep network from overfitting when the data shared similar spectral response. The proposed method successfully utilized samples from multiple data sets to increase its performance.

show abstract

Urban land-use mapping using a deep convolutional neural network with high spatial resolution multispectral remote sensing imagery

Cited by 449 publications

References 37 publications

Changes in land use/land cover and net primary productivity in the transboundary Ili-Balkhash basin of Central Asia, 1995–2015

Changes in land use/land cover and net primary productivity in the transboundary Ili-Balkhash basin of Central Asia, 1995–2015

Understanding urban landuse from the above and ground perspectives: A deep learning, multimodal solution

Multitask Deep Learning With Spectral Knowledge for Hyperspectral Image Classification

Contact Info

Product

Resources

About