Assessing the performance of multiple spectral–spatial features of a hyperspectral image for classification of urban land cover classes using support vector machines and artificial neural network

Abstract

Accurate and spatially detailed mapping of complex urban environments is essential for land managers. Classifying high spectral and spatial resolution hyperspectral images is a challenging task because of its data abundance and computational complexity. Land-use classification based on the consideration of spectral information, i.e., without any spatial organization, has limited potential in the final classification map. Consequently, approaches with a combination of spectral and spatial information in a single classification framework have attracted special attention because of their potential to improve the classification accuracy. In this study, we extracted multiple features from spectral and spatial domains of hyperspectral image and evaluated with two supervised classification algorithms such as support vector machines (SVM and artificial neural network (ANN). The spatial features considered in this study are produced by Gray Level Co-occurrence Matrix (GLCM) and extended multiattribute profiles (EMAP). All of these features were stacked and the most informative features selected using a genetic algorithm-based support vector machine (GA-SVM). After selecting the most informative features, the classification model was integrated with a segmentation map. The segmentation procedure was performed on the hyperspectral image using the Hidden Markov Random Field (HMRF). We tested the proposed method on a real application of hyperspectral image acquired from AisaFENIX and also on widely used hyperspectral images (ROSIS and AVIRIS). From the results, it can be concluded that the proposed framework significantly improves the results with different spectral and spatial resolutions over different instrumentation.

Optimization of Extraction Process Based on Neural Network

“…In the field of construction engineering, artificial neural networks are used to predict concrete strength and find the nonlinear input-output relationship between concrete strength and its influencing factors [9,10]. In addition, artificial neural networks are used in the field of plant diseases control [11][12][13], process control and optimization [14][15][16], troubleshooting [17][18][19], intelligent control of industrial product assembly line [20][21][22], robotic surgery [23][24][25], intelligent driving [26][27][28], chemical product development [29][30][31], signal processing [32][33][34], and so on.…”

Minimax Bridgeness-Based Clustering for Hyperspectral Data

“…The Pavia University image was acquired over Pavia, Italy (Lat-Long coordinates of scene centre: 45 • The Massey University scene [53] was captured in Palmerston North, New Zealand (Lat-Long coordinates of scene centre: 40 • 23 17 S, 175 • 37 07 E), with an airborne AisaFENIX hyperspectral sensor covering visible to short-wave infrared (380 to 2500 nm). It has 339 bands (after removal of water absorption and noisy bands) and 9564 pixels with corresponding ground truth.…”

Hybrid passive polarimetric imager and lidar combination for material classification

“…We propose using a support vector machine (SVM) for classifying material type due to the proven success of this classifier in similar applications such as hyperspectral imaging for land cover classification and target detection. 39,40 However, we believe advanced classifiers could be designed, based on the proposed technique (i.e., hybrid sensing with known viewing orientation), that optimize performance for a specific application. The SVM presented in this paper demonstrates the general application of material classification.…”