Comparing Deep Neural Networks, Ensemble Classifiers, and Support Vector Machine Algorithms for Object-Based Urban Land Use/Land Cover Classification

Abstract

With the advent of high-spatial resolution (HSR) satellite imagery, urban land use/land cover (LULC) mapping has become one of the most popular applications in remote sensing. Due to the importance of context information (e.g., size/shape/texture) for classifying urban LULC features, Geographic Object-Based Image Analysis (GEOBIA) techniques are commonly employed for mapping urban areas. Regardless of adopting a pixel- or object-based framework, the selection of a suitable classifier is of critical importance for urban mapping. The popularity of deep learning (DL) (or deep neural networks (DNNs)) for image classification has recently skyrocketed, but it is still arguable if, or to what extent, DL methods can outperform other state-of-the art ensemble and/or Support Vector Machines (SVM) algorithms in the context of urban LULC classification using GEOBIA. In this study, we carried out an experimental comparison among different architectures of DNNs (i.e., regular deep multilayer perceptron (MLP), regular autoencoder (RAE), sparse, autoencoder (SAE), variational autoencoder (AE), convolutional neural networks (CNN)), common ensemble algorithms (Random Forests (RF), Bagging Trees (BT), Gradient Boosting Trees (GB), and Extreme Gradient Boosting (XGB)), and SVM to investigate their potential for urban mapping using a GEOBIA approach. We tested the classifiers on two RS images (with spatial resolutions of 30 cm and 50 cm). Based on our experiments, we drew three main conclusions: First, we found that the MLP model was the most accurate classifier. Second, unsupervised pretraining with the use of autoencoders led to no improvement in the classification result. In addition, the small difference in the classification accuracies of MLP from those of other models like SVM, GB, and XGB classifiers demonstrated that other state-of-the-art machine learning classifiers are still versatile enough to handle mapping of complex landscapes. Finally, the experiments showed that the integration of CNN and GEOBIA could not lead to more accurate results than the other classifiers applied.

Mapping Annual Land Use and Land Cover Changes in the Yangtze Estuary Region Using an Object-Based Classification Framework and Landsat Time Series Data

“…First, compared to the traditional supervised classification algorithms [17], the proposed framework can make better use of spatial information such as shape and context, which can improve classification accuracy for complex LULC mapping. Second, compared to popular machine learning and deep learning algorithms, the proposed method does not need training samples, which is difficult to select training samples for these popular algorithms [54]. Third, object-based classification framework can largely reduce the errors caused by spatial misregistration, which is a common phenomenon and difficult to completely eliminate when using pix-based methods [55].…”

Space-Based Earth Observations for Disaster Risk Management

“…Usually, the main elements retrieved from Earth observations include the distribution of human population, buildings and strategic infrastructure, and roads, but the mapping of crops and natural ecosystems can be considered as well (De Bono and Mora 2014; Van Westen 2013). The spatial distribution and amount of most of these features can be readily extracted from moderate to very high spatial resolution satellite image through supervised classification algorithms, using pixel or object-based approaches (Jozdani et al 2019). Challenges remain in the classification of different types of elements based on their material (e.g.…”

Predicting Land Cover Change in the Mamminasata Area, Indonesia, to Evaluate the Spatial Plan

“…A support vector machine (SVM) was adopted as a land cover classifier because it has been evaluated as a high-performance machine-learning algorithm and has been investigated in a number of studies [26,30,42,43]. In this study, the pixel-based classification was applied using six bands of three composite scenes.…”