New semi-automated mapping of asbestos cement roofs using rule-based object-based image analysis and Taguchi optimization technique from WorldView-2 images

The use of asbestos cement (AC) roofing materials is a significant concern because of their deleterious effects on human health and the environment. The main objective of this study was to map AC roofs from WorldView-2 (WV-2) images using object-based image analysis (OBIA). A robust Taguchi optimization technique was used to optimize segmentation parameters for WV-2 images in heterogeneous urban areas. In this research, two subsets of WV-2 satellite image sets were utilized to map AC roofs. Rule-based OBIA framework was developed on the first study area. Different supervised OBIA classifiers, such as Bayes, k-nearest neighbour (k-NN), support vector machine (SVM), and random forest (RF), were tested on the first image of the study areas to evaluate the performance of a rule-based classifier. Results of the supervised classifiers showed confusion between AC roof class and some urban features, with overall accuracies of 72.21%, 77%, 81.75%, and 82.02% for Bayes, k-NN, SVM, and RF, respectively. To assess the transferability of the proposed method, the adopted classification framework was applied to larger subsets of WV-2 of the second study area. The results of the proposed approach showed outstanding performance, with overall accuracies of 93.10% and 90.74% for the first and second classified images, respectively. The McNemar test emphasized the statistical reliability of rule-based result (in the first site) compared with supervised classification results. Therefore, the proposed framework of using rule-based classification and Taguchi optimization technique provide an efficient and expeditious approach to mapping and monitoring the presence of AC roofs and help local authorities in their decision-making strategies and policies.     

Detailed intra-urban mapping through transferable OBIA rule sets using WorldView-2 very-high-resolution satellite images

A detailed knowledge of the types and coverage of intra-urban features is helpful for different applications, such as roof run-off approximation and urban micro-climate studies. Previous studies have applied object-based image analysis (OBIA) to explore the detailed urban characterization on a single image of satellite sensors with very-high- resolution. The automated and transferable detection of intra-urban features is challenging because of variations of the spatial and spectral characteristics. This study utilizes the rule-based structure of OBIA to investigate the transferability of the OBIA rule sets on three subsets of a WorldView-2 (WV-2) image. Spatial, spectral, and textural features as well as several spectral indices are incorporated in these rule sets. The rule sets are developed on the first study site and reused in the second and third images. This OBIA framework provides a transferable process of detecting the intra-urban features without manually adjusting the rule set parameters and thresholds. Overall accuracies of 88%, 88%, and 86% are obtained for the first, second, and third images, respectively. The rule sets used in this study can be applied to other study areas or temporal WV-2 images for accurate detection of the intra-urban land-cover classes.     

Improving detailed rule-based feature extraction of urban areas from WorldView-2 image and lidar data

Urbanization is commonly accepted as an important contributor to the growth of man-made structures and as a rapid convertor of natural environments to impervious surfaces. Roofs are one class of impervious surface whose materials can highly influence the quality of urban surface water. In this study, two data sources, WorldView-2 (WV-2) imagery and a combination of WV-2 and lidar data, were utilized to map intra-urban targets, with 13 classes. Images were classified using object-based image analysis. Pixel-based classifications using the support vector machine (SVM) and maximum likelihood (ML) methods were also tested for their abilities to use both lidar data and WV-2 imagery. ML and SVM classifications yielded overall accuracies of 72.46% and 75.69%, respectively. The results of these classifiers exhibited mixed pixels and salt-and-pepper effects. Spectral, spatial, and textural attributes as well as various spectral indices were employed in the object-based classification of WV-2 imagery. Feature classification of WV-2 imagery resulted in 85% overall accuracy. Lidar data were added to WV-2 imagery to assist in the spatial and spectral diversities of urban infrastructures. Classified image made from WV-2 imagery and lidar data achieved 92.84% overall accuracy. Rule-sets of these fused datasets effectively reduced the spectral variation and spatial heterogeneities of intra-urban classes, causing finer boundaries among land-cover classes. Therefore, object-based classification of WV-2 imagery and lidar data efficiently improved detailed characterization of roof types and other urban surface materials.     

Land-cover classification of partly missing data using support vector machines

Land-cover classification based on multi-temporal satellite images for scenarios where parts of the data are missing due to, for example, clouds, snow or sensor failure has received little attention in the remote-sensing literature. The goal of this article is to introduce support vector machine (SVM) methods capable of handling missing data in land-cover classification. The novelty of this article consists of combining the powerful SVM regularization framework with a recent statistical theory of missing data, resulting in a new method where an SVM is trained for each missing data pattern, and a given incomplete test vector is classified by selecting the corresponding SVM model. The SVM classifiers are evaluated on Landsat Enhanced Thematic Mapper Plus (ETM?+?) images covering a scene of Norwegian mountain vegetation. The results show that the proposed SVM-based classifier improves the classification accuracy by 5–10% compared with single image classification. The proposed SVM classifier also outperforms recent non-parametric k-nearest neighbours (k-NN) and Parzen window density-based classifiers for incomplete data by about 3%. Moreover, since the resulting SVM classifier may easily be implemented using existing SVM libraries, we consider the new method to be an attractive choice for classification of incomplete data in remote sensing.     

A multiple-point spatially weighted k-NN classifier for remote sensing

A novel classification method based on multiple-point statistics (MPS) is proposed in this article. The method is a modified version of the spatially weighted k-nearest neighbour (k-NN) classifier, which accounts for spatial correlation through weights applied to neighbouring pixels. The MPS characterizes the spatial correlation between multiple points of land-cover classes by learning local patterns in a training image. This rich spatial information is then converted to multiple-point probabilities and incorporated into the k-NN classifier. Experiments were conducted in two study areas, in which the proposed method for classification was tested on a WorldView-2 sub-scene of the Sichuan mountainous area and an IKONOS image of the Beijing urban area. The multiple-point weighted k-NN method (MPk-NN) was compared to several alternatives; including the traditional k-NN and two previously published spatially weighted k-NN schemes; the inverse distance weighted k-NN, and the geostatistically weighted k-NN. The classifiers using the Bayesian and Support Vector Machine (SVM) methods, and these classifiers weighted with spatial context using the Markov random field (MRF) model, were also introduced to provide a benchmark comparison with the MPk-NN method. The proposed approach increased classification accuracy significantly relative to the alternatives, and it is, thus, recommended for the identification of land-cover types with complex and diverse spatial distributions.     

Image mining by spectral features: A case study of scenery image classification

Spectral features of images, such as Gabor filters and wavelet transform can be used for texture image classification. That is, a classifier is trained based on some labeled texture features as the training set to classify unlabeled texture features of images into some pre-defined classes. The aim of this paper is twofold. First, it investigates the classification performance of using Gabor filters, wavelet transform, and their combination respectively, as the texture feature representation of scenery images (such as mountain, castle, etc.). A k-nearest neighbor (k-NN) classifier and support vector machine (SVM) are also compared. Second, three k-NN classifiers and three SVMs are combined respectively, in which each of the combined three classifiers uses one of the above three texture feature representations respectively, to see whether combining multiple classifiers can outperform the single classifier in terms of scenery image classification. The result shows that a single SVM using Gabor filters provides the highest classification accuracy than the other two spectral features and the combined three k-NN classifiers and three SVMs.     

Genetic classifiers for remotely sensed images: Comparison with standard methods

In this article the effectiveness of some recently developed genetic algorithm-based pattern classifiers was investigated in the domain of satellite imagery which usually have complex and overlapping class boundaries. Landsat data, SPOT image and IRS image are considered as input. The superiority of these classifiers over k-NN rule, Bayes' maximum likelihood classifier and multilayer perceptron (MLP) for partitioning different landcover types is established. Results based on producer's accuracy (percentage recognition score), user's accuracy and kappa values are provided. Incorporation of the concept of variable length chromosomes and chromosome discrimination led to superior performance in terms of automatic evolution of the number of hyperplanes for modelling the class boundaries, and the convergence time. This non-parametric classifier requires very little a priori information, unlike k-NN rule and MLP (where the performance depends heavily on the value of k and the architecture, respectively), and Bayes' maximum likelihood classifier (where assumptions regarding the class distribution functions need to be made).     

The effects of combining classifiers with the same training statistics using Bayesian decision rules

Multiple classifier systems (MCS) are attracting increasing interest in the field of pattern recognition and machine learning. Recently, MCS are also being introduced in the remote sensing field where the importance of classifier diversity for image classification problems has not been examined. In this article, Satellite Pour l'Observation de la Terre (SPOT) IV panchromatic and multispectral satellite images are classified into six land cover classes using five base classifiers: contextual classifier, k-nearest neighbour classifier, Mahalanobis classifier, maximum likelihood classifier and minimum distance classifier. The five base classifiers are trained with the same feature sets throughout the experiments and a posteriori probability, derived from the confusion matrix of these base classifiers, is applied to five Bayesian decision rules (product rule, sum rule, maximum rule, minimum rule and median rule) for constructing different combinations of classifier ensembles. The performance of these classifier ensembles is evaluated for overall accuracy and kappa statistics. Three statistical tests, the McNemar's test, the Cochran's Q test and the Looney's F-test, are used to examine the diversity of the classification results of the base classifiers compared to the results of the classifier ensembles. The experimental comparison reveals that (a) significant diversity amongst the base classifiers cannot enhance the performance of classifier ensembles; (b) accuracy improvement of classifier ensembles can only be found by using base classifiers with similar and low accuracy; (c) increasing the number of base classifiers cannot improve the overall accuracy of the MCS and (d) none of the Bayesian decision rules outperforms the others.     

A combination of CSP-based method with soft margin SVM classifier and generalized RBF kernel for imagery-based brain computer interface applications

Several methods utilizing common spatial pattern (CSP) algorithm have been presented for improving the identification of imagery movement patterns for brain computer interface applications. The present study focuses on improving a CSP-based algorithm for detecting the motor imagery movement patterns. A discriminative filter bank of CSP method using a discriminative sensitive learning vector quantization (DFBCSP-DSLVQ) system is implemented. Four algorithms are then combined to form three methods for improving the efficiency of the DFBCSP-DSLVQ method, namely the kernel linear discriminant analysis (KLDA), the kernel principal component analysis (KPCA), the soft margin support vector machine (SSVM) classifier and the generalized radial bases functions (GRBF) kernel. The GRBF is used as a kernel for the KLDA, the KPCA feature selection algorithms and the SSVM classifier. In addition, three types of classifiers, namely K-nearest neighbor (K-NN), neural network (NN) and traditional support vector machine (SVM), are employed to evaluate the efficiency of the classifiers. Results show that the best algorithm is the combination of the DFBCSP-DSLVQ method using the SSVM classifier with GRBF kernel (SSVM-GRBF), in which the best average accuracy, attained are 92.70% and 83.21%, respectively. Results of the Repeated Measures ANOVA shows the statistically significant dominance of this method at p <?0.05. The presented algorithms are then compared with the base algorithm of this study i.e. the DFBCSP-DSLVQ with the SVM-RBF classifier. It is concluded that the algorithms, which are based on the SSVM-GRBF classifier and the KLDA with the SSVM-GRBF classifiers give sufficient accuracy and reliable results.

     

k-Top Scoring Pair Algorithm for feature selection in SVM with applications to microarray data classification

Top Scoring Pair (TSP) and its ensemble counterpart, k-Top Scoring Pair (k-TSP), were recently introduced as competitive options for solving classification problems of microarray data. However, support vector machine (SVM) which was compared with these approaches is not equipped with feature or variable selection mechanism while TSP itself is a kind of variable selection algorithm. Moreover, an ensemble of SVMs should also be considered as a possible competitor to k-TSP. In this work, we conducted a fair comparison between TSP and SVM-recursive feature elimination (SVM-RFE) as the feature selection method for SVM. We also compared k-TSP with two ensemble methods using SVM as their base classifier. Results on ten public domain microarray data indicated that TSP family classifiers serve as good feature selection schemes which may be combined effectively with other classification methods.     

Classification of impervious land-use features using object-based image analysis and data fusion

The proportion of impervious area within a watershed is a key indicator of the impacts of urbanization on water quality and stream health. Research has shown that object-based image analysis (OBIA) techniques are more effective for urban land-cover classification than pixel-based classifiers and are better suited to the increased complexity of high-resolution imagery. Focusing on five 2-km² study areas within the Black Creek sub-watershed of the Humber River, this research uses eCognition® software to develop a rule-based OBIA workflow for semi-automatic classification of impervious land-use features (e.g., roads, buildings, Parking Lots, driveways). The overall classification accuracy ranges from 88.7 to 94.3%, indicating the effectiveness of using an OBIA approach and developing a sequential system for data fusion and automated impervious feature extraction. Similar accuracy results between the calibrating and validating sites demonstrates the strong potential for the transferability of the rule-set from pilot study sites to a larger area.     

Identifying potential areas of Cannabis sativa plantations using object-based image analysis of SPOT-5 satellite data

The rapid and efficient detection of illicit drug cultivation, such as that of Cannabis sativa, is important in reducing consumption. The objective of this study was to identify potential sites of illicit C. sativa plantations located in the semi-arid, southern part of Pernambuco State, Brazil. The study was conducted using an object-based image analysis (OBIA) of Système Pour l'Observation de la Terre high-resolution geometric (SPOT-5 HRG) images (overpass: 31 May, 2007). OBIA considers the target's contextual and geometrical attributes to overcome the difficulties inherent in detecting illicit crops associated with the grower's strategies to conceal their fields and optimizes the spectral information extracted to generate land-cover maps. The capabilities of the SPOT-5 near-infrared and shortwave infrared bands to discriminate herbaceous vegetation with high water content, and employment of the support vector machine classifier, contributed to accomplishing this task. Image classification included multiresolution segmentation with an algorithm available in the eCognition Developer software package. In addition to a SPOT-5 HRG multispectral image with 10 m spatial resolution and a panchromatic image with 2.5 m spatial resolution, first-order indices such as the normalized difference vegetation index and ancillary data including land-cover classes, anthropogenic areas, slope, and distance to water sources were also employed in the OBIA. The classification of segments (objects) related to illegal cultivation employed fuzzy logic and fixed-threshold membership functions to describe the following spectral, geometrical, and contextual properties of targets: vegetation density, topography, neighbourhood, and presence of water supplies for irrigation. The results of OBIA were verified from a weight of evidence analysis. Among 15 previously known C. sativa sites identified during police operations conducted on 5–17 June 2007, eight sites were classified as maximum-alert areas (total area of 22.54 km² within a total area of object-oriented image classification of ～1800 km²). The approach proposed in this study is feasible for reducing the area to be searched for illicit cannabis cultivation in semi-arid regions.     

Fuzzy classifier fusion: an application to reservoir facies identification

An application of classifier fusion technique is presented to improve the performance of automated reservoir facies identification system. The algorithm presented in this study uses three well-known classifiers, namely Bayesian, k-nearest neighbor (kNN), and support vector machine (SVM) to automatically identify four defined facies of Asmari Formation from log-derived amplitude versus offset (AVO) attributes. Fuzzy Sugeno integral (FSI) method is then employed to combine the outputs of three investigated classifiers and increase the consistency of reservoir facies identification process. The experimental results obtained from applying the presented algorithm on data related to three wells drilled in Asmari Formation provide evidence of the effectiveness of the proposed algorithm regarding true positive (TP), false positive (FP), and classification accuracy criteria.

     

Guidance on and comparison of machine learning classifiers for Landsat-based land cover and land use mapping

Remote sensing scientists are increasingly adopting machine learning classifiers for land cover and land use (LCLU) mapping, but model selection, a critical step of the machine learning classification, has usually been ignored in the past research. In this paper, step-by-step guidance (for classifier training, model selection, and map production) with supervised learning model selection is first provided. Then, model selection is exhaustively applied to different machine learning (e.g. Artificial Neural Network (ANN), Decision Tree (DT), Support Vector Machine (SVM), and Random Forest (RF)) classifiers to identify optimal polynomial degree of input features (d) and hyperparameters with Landsat imagery of a study region in China and Ghana. We evaluated the map accuracy and computing time associated with different versions of machine learning classification software (i.e. ArcMap, ENVI, TerrSet, and R).

The optimal classifiers and their associated polynomial degree of input features and hyperparameters vary for the two image datasets that were tested. The optimum combination of d and hyperparameters for each type of classifier was used across software packages, but some classifiers (i.e. ENVI and TerrSet ANN) were customized due to the constraints of software packages. The LCLU map derived from ENVI SVM has the highest overall accuracy (72.6%) for the Ghana dataset, while the LCLU map derived from R DT has the highest overall accuracy (48.0%) for the FNNR dataset. All LCLU maps for the Ghana dataset are more accurate compared to those from the China dataset, likely due to more limited and uncertain training data for the China (FNNR) dataset. For the Ghana dataset, LCLU maps derived from tree-based classifiers (ArcMap RF, TerrSet DT, and R RF) routines are accurate, but these maps have artefacts resulting from model overfitting problems.     

Modified nearest neighbour classifier for hyperspectral data classification

A modified k-nearest neighbour (k-NN) classifier is proposed for supervised remote sensing classification of hyperspectral data. To compare its performance in terms of classification accuracy and computational cost, k-NN and a back-propagation neural network classifier were used. A classification accuracy of 91.2% was achieved by the proposed classifier with the data set used. Results from this study suggest that the accuracy achieved with this classifier is significantly better than the k-NN and comparable to a back-propagation neural network. Comparison in terms of computational cost also suggests the effectiveness of modified k-NN classifier for hyperspectral data classification. A fuzzy entropy-based filter approach was used for feature selection to compare the performance of modified and k-NN classifiers with a reduced data set. The results suggest a significant increase in classification accuracy by the modified k-NN classifier in comparison with k-NN classifier with selected features.     

Handwritten digit recognition: benchmarking of state-of-the-art techniques

This paper presents the results of handwritten digit recognition on well-known image databases using state-of-the-art feature extraction and classification techniques. The tested databases are CENPARMI, CEDAR, and MNIST. On the test data set of each database, 80 recognition accuracies are given by combining eight classifiers with ten feature vectors. The features include chaincode feature, gradient feature, profile structure feature, and peripheral direction contributivity. The gradient feature is extracted from either binary image or gray-scale image. The classifiers include the k-nearest neighbor classifier, three neural classifiers, a learning vector quantization classifier, a discriminative learning quadratic discriminant function (DLQDF) classifier, and two support vector classifiers (SVCs). All the classifiers and feature vectors give high recognition accuracies. Relatively, the chaincode feature and the gradient feature show advantage over other features, and the profile structure feature shows efficiency as a complementary feature. The SVC with RBF kernel (SVC-rbf) gives the highest accuracy in most cases but is extremely expensive in storage and computation. Among the non-SV classifiers, the polynomial classifier and DLQDF give the highest accuracies. The results of non-SV classifiers are competitive to the best ones previously reported on the same databases.     

Class-specific multiple classifiers scheme to recognize emotions from speech signals

Automatic emotion recognition from speech signals is one of the important research areas, which adds value to machine intelligence. Pitch, duration, energy and Mel-frequency cepstral coefficients (MFCC) are the widely used features in the field of speech emotion recognition. A single classifier or a combination of classifiers is used to recognize emotions from the input features. The present work investigates the performance of the features of Autoregressive (AR) parameters, which include gain and reflection coefficients, in addition to the traditional linear prediction coefficients (LPC), to recognize emotions from speech signals. The classification performance of the features of AR parameters is studied using discriminant, k-nearest neighbor (KNN), Gaussian mixture model (GMM), back propagation artificial neural network (ANN) and support vector machine (SVM) classifiers and we find that the features of reflection coefficients recognize emotions better than the LPC. To improve the emotion recognition accuracy, we propose a class-specific multiple classifiers scheme, which is designed by multiple parallel classifiers, each of which is optimized to a class. Each classifier for an emotional class is built by a feature identified from a pool of features and a classifier identified from a pool of classifiers that optimize the recognition of the particular emotion. The outputs of the classifiers are combined by a decision level fusion technique. The experimental results show that the proposed scheme improves the emotion recognition accuracy. Further improvement in recognition accuracy is obtained when the scheme is built by including MFCC features in the pool of features.     

Towards a comprehensive evaluation of V-I-S sub-pixel fractions and land surface temperature for urban land-use classification in the USA

Remote-sensing image classification based on the vegetation–impervious surface–soil (V-I-S) model and land-surface temperature (LST) has proved to be more efficient in characterizing the urban landscape than conventional spectral-based classification. However, current literature emphasizes discussion of the classifier's accuracy improvement achieved by the input of V-I-S fractions and LST over conventional spectral-based classification while ignoring the stability evaluation. Hence, this study proposes an evaluation framework for exploring the superiority of the input features and the stability of classifiers by integrating statistical randomization techniques and a kappa-error diagram. The evaluation framework was applied to case studies for demonstrating the superiority of V-I-S fractions and LST in the context of urban land-use classification with five different types of classifiers, including the maximum likelihood classifier (MLC), the tree classifier, the Bagging classifier, the random forest (RF) and the support vector machine (SVM). It followed that the use of V-I-S fractions and LST (1) could alleviate the ‘salt and pepper’ effect; (2) is preferred by tree and tree-based ensembles for branch splitting; (3) could produce classification trees with less complexity; (4) could benefit the stability of classifiers in addition to the accuracy improvement; and (5) could allow histograms following nearly normal distribution in its feature space, which boosts the performance of MLC. It is shown that MLC becomes comparable with modern classifiers when trained with V-I-S fractions and LST combination. Because of its adequacy and simplicity, MLC is recommended for urban land-use classification when V-I-S fractions and LST are used as the only input features. However, replacing them with, or including, the band reflectance might degrade MLC. A direct use of spectral band reflectance is not recommended for any of the classification approaches being considered in this study, except for SVM, which is the most robust classifier as it has a consistently high performance for all the input feature combinations. We recommend using tree-based ensemble classifiers or SVM when V-I-S fractions and LST as well as the band reflectance are all used in the classification. The proposed evaluation framework can also be applied to the assessment of input features and classifiers in other remote-sensing classification endeavours.     

Commitment and typicality measures for the Self-Organizing Map

Soft classification using Kohonen's Self-Organizing Map (SOM) has not been explored as thoroughly as the Multi-Layer-Perceptron (MLP) neural network. In this paper, we propose two non-parametric algorithms for the SOM to provide soft classification outputs. These algorithms, which are labelling-frequency-based, are called SOM Commitment (SOM-C) and SOM Typicality (SOM-T), expressing in the first case the degree of commitment the classifier has for each class for a specific pixel and in the second case, how typical that pixel's reflectances are of those upon which the classifier was trained. To evaluate the two proposed algorithms, soft classifications of a Satellite Pour l'Observation de la Terre (SPOT) High Resolution Visible (HRV) image and an Airborne Visible Infrared Imaging Spectrometer (AVIRIS) image were undertaken. Both traditional soft classifiers, i.e. Bayesian posterior probability and Mahalanobis typicality classifier, and the most frequently used non-parametric neural network model, i.e. MLP, were used as a comparison. Principal-components analysis (PCA) was used to explore the relationship between these measures. Results indicate that great similarities exist between the SOM-C, MLP and the Bayesian posterior probability classifiers, while the SOM-T corresponds closely with Mahalanobis typicality probabilities. However, as implemented, they have the advantage of being non-parametric. The proposed measures significantly outperformed Bayesian and Mahalanobis classifiers when using the hyperspectral AVIRIS image.