Different Spectral Domain Transformation for Land Cover Classification Using Convolutional Neural Networks with Multi-Temporal Satellite Imagery

This study compares some different types of spectral domain transformations for convolutional neural network (CNN)-based land cover classification. A novel approach was proposed, which transforms one-dimensional (1-D) spectral vectors into two-dimensional (2-D) features: Polygon graph images (CNN-Polygon) and 2-D matrices (CNN-Matrix). The motivations of this study are that (1) the shape of the converted 2-D images is more intuitive for human eyes to interpret when compared to 1-D spectral input; and (2) CNNs are highly specialized and may be able to similarly utilize this information for land cover classification. Four seasonal Landsat 8 images over three study areas—Lake Tapps, Washington, Concord, New Hampshire, USA, and Gwangju, Korea—were used to evaluate the proposed approach for nine land cover classes compared to several other methods: Random forest (RF), support vector machine (SVM), 1-D CNN, and patch-based CNN. Oversampling and undersampling approaches were conducted to examine the effect of the sample size on the model performance. The CNN-Polygon had better performance than the other methods, with overall accuracies of about 93%–95 % for both Concord and Lake Tapps and 80%–84% for Gwangju. The CNN-Polygon particularly performed well when the training sample size was small, less than 200 per class, while the CNN-Matrix resulted in similar or higher performance as sample sizes became larger. The contributing input variables to the models were carefully analyzed through sensitivity analysis based on occlusion maps and accuracy decreases. Our result showed that a more visually intuitive representation of input features for CNN-based classification models yielded higher performance, especially when the training sample size was small. This implies that the proposed graph-based CNNs would be useful for land cover classification where reference data are limited.

Download Full-text

Impact of Training Set Size on Object-Based Land Cover Classification

International Journal of Applied Geospatial Research ◽

10.4018/ijagr.2014070104 ◽

2014 ◽

Vol 5 (3) ◽

pp. 49-67 ◽

Cited By ~ 10

Author(s):

Gerhard Myburgh ◽

Adriaan van Niekerk

Keyword(s):

Land Cover ◽

Classification Tree ◽

Land Cover Classification ◽

Training Sample ◽

Support Vector ◽

Training Set ◽

Training Sample Size ◽

Set Size ◽

Object Based ◽

Training Samples

Supervised classifiers are commonly employed in remote sensing to extract land cover information, but various factors affect their accuracy. The number of available training samples, in particular, is known to have a significant impact on classification accuracies. Obtaining a sufficient number of samples is, however, not always practical. The support vector machine (SVM) is a supervised classifier known to perform well with limited training samples and has been compared favourably to other classifiers for various problems in pixel-based land cover classification. Very little research on training-sample size and classifier performance has been done in a geographical object-based image analysis (GEOBIA) environment. This paper compares the performance of SVM, nearest neighbour (NN) and maximum likelihood (ML) classifiers in a GEOBIA environment, with a focus on the influence of training-set size. Training-set sizes ranging from 4-20 per land cover class were tested. Classification tree analysis (CTA) was used for feature selection. The results indicate that the performance of all the classifiers improved significantly as the size of the training set increased. The ML classifier performed poorly when few (<10 per class) training samples were used and the NN classifier performed poorly compared to SVM throughout the experiment. SVM was the superior classifier for all training-set sizes although ML achieved competitive results for sets of 12 or more training areas per class.

Download Full-text

Automatic Classification of Major Urban Land Covers Based on Novel Spectral Indices

ISPRS International Journal of Geo-Information ◽

10.3390/ijgi7120453 ◽

2018 ◽

Vol 7 (12) ◽

pp. 453 ◽

Cited By ~ 5

Author(s):

Mst Ilme Faridatul ◽

Bo Wu

Keyword(s):

Land Cover ◽

Land Cover Classification ◽

Urban Land ◽

Support Vector ◽

Landsat 8 ◽

Spectral Indices ◽

Urban Land Cover ◽

Svm Algorithm ◽

Bare Land ◽

Land Covers

Urban land cover classification and mapping is an important and ongoing research field in monitoring and managing urban sprawl and terrestrial ecosystems. The changes in land cover largely affect the terrestrial ecosystem, thus information on land cover is important for understanding the ecological environment. Quantification of land cover in urban areas is challenging due to their diversified activities and large spatial and temporal variations. To improve urban land cover classification and mapping, this study presents three new spectral indices and an automated approach to classifying four major urban land types: impervious, bare land, vegetation, and water. A modified normalized difference bare-land index (MNDBI) is proposed to enhance the separation of impervious and bare land. A tasseled cap water and vegetation index (TCWVI) is proposed to enhance the detection of vegetation and water areas. A shadow index (ShDI) is proposed to further improve water detection by separating water from shadows. An approach for optimizing the thresholds of the new indices is also developed. Finally, the optimized thresholds are used to classify land covers using a decision tree algorithm. Using Landsat-8 Operational Land Imager (OLI) data from two study sites (Hong Kong and Dhaka City, Bangladesh) with different urban characteristics, the proposed approach is systematically evaluated. Spectral separability analysis of the new indices is performed and compared with other common indices. The urban land cover classifications achieved by the proposed approach are compared with those of the classic support vector machine (SVM) algorithm. The proposed approach achieves an overall classification accuracy of 94-96%, which is superior to the accuracy of the SVM algorithm.

Download Full-text

Decision Tree Algorithms for Developing Rulesets for Object-Based Land Cover Classification

ISPRS International Journal of Geo-Information ◽

10.3390/ijgi9050329 ◽

2020 ◽

Vol 9 (5) ◽

pp. 329 ◽

Cited By ~ 2

Author(s):

Darius Phiri ◽

Matamyo Simwanda ◽

Vincent Nyirenda ◽

Yuji Murayama ◽

Manjula Ranagalage

Keyword(s):

Land Cover ◽

Decision Tree ◽

Land Cover Classification ◽

Support Vector ◽

Landsat 8 ◽

Object Based Image Analysis ◽

Object Based ◽

Minimum Number ◽

Tree Algorithms ◽

And Performance

Decision tree (DT) algorithms are important non-parametric tools used for land cover classification. While different DTs have been applied to Landsat land cover classification, their individual classification accuracies and performance have not been compared, especially on their effectiveness to produce accurate thresholds for developing rulesets for object-based land cover classification. Here, the focus was on comparing the performance of five DT algorithms: Tree, C5.0, Rpart, Ipred, and Party. These DT algorithms were used to classify ten land cover classes using Landsat 8 images on the Copperbelt Province of Zambia. Classification was done using object-based image analysis (OBIA) through the development of rulesets with thresholds defined by the DTs. The performance of the DT algorithms was assessed based on: (1) DT accuracy through cross-validation; (2) land cover classification accuracy of thematic maps; and (3) other structure properties such as the sizes of the tree diagrams and variable selection abilities. The results indicate that only the rulesets developed from DT algorithms with simple structures and a minimum number of variables produced high land cover classification accuracies (overall accuracy > 88%). Thus, algorithms such as Tree and Rpart produced higher classification results as compared to C5.0 and Party DT algorithms, which involve many variables in classification. This high accuracy has been attributed to the ability to minimize overfitting and the capacity to handle noise in the data during training by the Tree and Rpart DTs. The study produced new insights on the formal selection of DT algorithms for OBIA ruleset development. Therefore, the Tree and Rpart algorithms could be used for developing rulesets because they produce high land cover classification accuracies and have simple structures. As an avenue of future studies, the performance of DT algorithms can be compared with contemporary machine-learning classifiers (e.g., Random Forest and Support Vector Machine).

Download Full-text

Land use/land cover classification using machine learning models

International Journal of Electrical and Computer Engineering (IJECE) ◽

10.11591/ijece.v12i2.pp2040-2046 ◽

2022 ◽

Vol 12 (2) ◽

pp. 2040

Author(s):

Subhra Swetanisha ◽

Amiya Ranjan Panda ◽

Dayal Kumar Behera

Keyword(s):

Machine Learning ◽

Land Use ◽

Land Cover ◽

Land Cover Classification ◽

Support Vector ◽

Landsat 8 ◽

Electromagnetic Spectrum ◽

Learning Models ◽

Extreme Gradient Boosting ◽

Machine Learning Models

<p>An ensemble model has been proposed in this work by combining the extreme gradient boosting classification (XGBoost) model with support vector machine (SVM) for land use and land cover classification (LULCC). We have used the multispectral Landsat-8 operational land imager sensor (OLI) data with six spectral bands in the electromagnetic spectrum (EM). The area of study is the administrative boundary of the twin cities of Odisha. Data collected in 2020 is classified into seven land use classes/labels: river, canal, pond, forest, urban, agricultural land, and sand. Comparative assessments of the results of ten machine learning models are accomplished by computing the overall accuracy, kappa coefficient, producer accuracy and user accuracy. An ensemble classifier model makes the classification more precise than the other state-of-the-art machine learning classifiers.</p>

Download Full-text

Land-Cover Classification of Coastal Wetlands Using the RF Algorithm for Worldview-2 and Landsat 8 Images

Remote Sensing ◽

10.3390/rs11161927 ◽

2019 ◽

Vol 11 (16) ◽

pp. 1927 ◽

Cited By ~ 17

Author(s):

Xiaoxue Wang ◽

Xiangwei Gao ◽

Yuanzhi Zhang ◽

Xianyun Fei ◽

Zhou Chen ◽

...

Keyword(s):

Remote Sensing ◽

Land Cover ◽

Image Classification ◽

Classification Accuracy ◽

Coastal Wetlands ◽

Land Cover Classification ◽

Support Vector ◽

Small Scale ◽

Landsat 8

Wetlands are one of the world’s most important ecosystems, playing an important role in regulating climate and protecting the environment. However, human activities have changed the land cover of wetlands, leading to direct destruction of the environment. If wetlands are to be protected, their land cover must be classified and changes to it monitored using remote sensing technology. The random forest (RF) machine learning algorithm, which offers clear advantages (e.g., processing feature data without feature selection and preferable classification result) for high spatial image classification, has been used in many study areas. In this research, to verify the effectiveness of this algorithm for remote sensing image classification of coastal wetlands, two types of spatial resolution images of the Linhong Estuary wetland in Lianyungang—Worldview-2 and Landsat-8 images—were used for land cover classification using the RF method. To demonstrate the preferable classification accuracy of the RF algorithm, the support vector machine (SVM) and k-nearest neighbor (k-NN) methods were also used to classify the same area of land cover for comparison with the results of RF classification. The study results showed that (1) the overall accuracy of the RF method reached 91.86%, higher than the SVM and k-NN methods by 4.68% and 4.72%, respectively, for Worldview-2 images; (2) at the same time, the classification accuracies of RF, SVM, and k-NN were 86.61%, 79.96%, and 77.23%, respectively, for Landsat-8 images; (3) for some land cover types having only a small number of samples, the RF algorithm also achieved better classification results using Worldview-2 and Landsat-8 images, and (4) the addition texture features could improve the classification accuracy of the RF method when using Worldview-2 images. Research indicated that high-resolution remote sensing images are more suitable for small-scale land cover classification image and that the RF algorithm can provide better classification accuracy and is more suitable for coastal wetland classification than the SVM and k-NN algorithms are.

Download Full-text

A Patch-Based Light Convolutional Neural Network for Land-Cover Mapping Using Landsat-8 Images

Remote Sensing ◽

10.3390/rs11020114 ◽

2019 ◽

Vol 11 (2) ◽

pp. 114 ◽

Cited By ~ 3

Author(s):

Hunsoo Song ◽

Yonghyun Kim ◽

Yongil Kim

Keyword(s):

Neural Network ◽

Land Cover ◽

Convolutional Neural Network ◽

Sample Size ◽

Patch Size ◽

Hyperspectral Image ◽

Support Vector ◽

Landsat 8 ◽

Sample Sizes ◽

Wide Range

This study proposes a light convolutional neural network (LCNN) well-fitted for medium-resolution (30-m) land-cover classification. The LCNN attains high accuracy without overfitting, even with a small number of training samples, and has lower computational costs due to its much lighter design compared to typical convolutional neural networks for high-resolution or hyperspectral image classification tasks. The performance of the LCNN was compared to that of a deep convolutional neural network, support vector machine (SVM), k-nearest neighbors (KNN), and random forest (RF). SVM, KNN, and RF were tested with both patch-based and pixel-based systems. Three 30 km × 30 km test sites of the Level II National Land Cover Database were used for reference maps to embrace a wide range of land-cover types, and a single-date Landsat-8 image was used for each test site. To evaluate the performance of the LCNN according to the sample sizes, we varied the sample size to include 20, 40, 80, 160, and 320 samples per class. The proposed LCNN achieved the highest accuracy in 13 out of 15 cases (i.e., at three test sites with five different sample sizes), and the LCNN with a patch size of three produced the highest overall accuracy of 61.94% from 10 repetitions, followed by SVM (61.51%) and RF (61.15%) with a patch size of three. Also, the statistical significance of the differences between LCNN and the other classifiers was reported. Moreover, by introducing the heterogeneity value (from 0 to 8) representing the complexity of the map, we demonstrated the advantage of patch-based LCNN over pixel-based classifiers, particularly at moderately heterogeneous pixels (from 1 to 4), with respect to accuracy (LCNN is 5.5% and 6.3% more accurate for a training sample size of 20 and 320 samples per class, respectively). Finally, the computation times of the classifiers were calculated, and the LCNN was confirmed to have an advantage in large-area mapping.

Download Full-text