Pieter Kempeneers

A band selection technique for spectral classification

In hyperspectral remote sensing, sensors acquire reflectance values at many different wavelength bands, to cover a complete spectral interval. These measurements are strongly correlated, and no new information might be added when increasing the spectral resolution. Moreover, the higher number of spectral bands increases the complexity of a classification task. Therefore, feature reduction is a crucial step. An alternative would be to choose the required sensor bands settings a priori. In this letter, we introduce a statistical procedure to provide band settings for a specific classification task. The proposed procedure selects wavelength band settings which optimize the separation between the different spectral classes. The method is applicable as a band reduction technique, but it can as well serve the purpose of data interpretation or be an aid in sensor design. Results on a vegetation classification task show an improvement in classification performance over feature selection and other band selection techniques.

Data Fusion of Different Spatial Resolution Remote Sensing Images Applied to Forest-Type Mapping

A data fusion method for land cover (LC) classification is proposed that combines remote sensing data at a fine and a coarse spatial resolution. It is a two-step approach, based on the assumption that some of the LC classes can be merged into a more generalized LC class. Step one creates a generalized LC map, using only the information available at the fine spatial resolution. In the second step, a new classifier refines the generalized LC classes to create distinct subclasses of its parent class, using the generalized LC map as a mask. This classifier uses all image information (bands) available at both fine and coarse spatial resolutions. We followed a simple data fusion technique by stacking the individual image bands into a multidimensional vector. The advantage of the proposed approach is that the spatial detail of the generalized LC classes is retained in the final LC map. The method has been designed for operational LC mapping over large areas. Within this paper, it is shown that the proposed data fusion approach increased the robustness of forest-type mapping within Europe. Robustness is particularly important when creating continental LC maps at fine spatial resolution. These maps become more popular now that remote sensing data at fine resolution are easier to access.

Comprehensive Monitoring of Wildfires in Europe: The European Forest Fire Information System (EFFIS)

Fires are an integral component of ecosystem dynamics in European landscapes. However, uncontrolled fires cause large environmental and economic damages, especially in the Mediterranean region. On average, about 65000 fires occur in Europe every year, burning approximately half a million ha of wildland and forest areas; most of the burnt area, over 85%, is in the European Mediterranean region. Trends in number of fires and burnt areas in the Mediterranean region are presented in Fig. 1.

Generic wavelet-based hyperspectral classification applied to vegetation stress detection

This communication studies the detection of vegetation stress in hyperspectral data. Compared to traditional vegetation stress indices, the proposed approach uses the complete reflectance spectrum and its wavelet representation. The detection strategy is formulated as a classification problem. Experiments are conducted on fruit tree stress detection. The experiments show the superior performance of the proposed strategy and demonstrate its generic nature.

Model inversion for chlorophyll estimation in open canopies from hyperspectral imagery

This paper presents the results of estimation of leaf chlorophyll concentration through model inversion, from hyperspectral imagery of artificially treated orchard crops. The objectives were to examine model inversion robustness under changing viewing conditions, and the potential of multi‐angle hyperspectral data to improve accuracy of chlorophyll estimation. The results were compared with leaf chlorophyll measurements from laboratory analysis and field spectroscopy. Two state‐of‐the‐art canopy models were compared. The first is a turbid medium canopy reflectance model (MCRM) and the second is a 3D model (FLIGHT). Both were linked to the PROSPECT leaf model. A linear regression using a single band was also performed as a reference. The different techniques were able to detect nutrient deficiencies that caused stress from the hyperspectral data obtained from the airborne AHS sensor. However, quantitative chlorophyll retrieval was found largely dependent on viewing conditions for regression and the turbid medium model inversion. In contrast, the 3D model was successful for all observations. It offers a robust technique to extract chlorophyll quantitatively from airborne hyperspectral data. When multi‐angular data were combined, the results for both the turbid medium and 3D model increased. Final RMSE values of 5.8 µg cm−2 (MCRM) and 4.7 µg cm−2 (FLIGHT) were obtained for chlorophyll retrieval on canopy level.

Superresolution Enhancement of Hyperspectral CHRIS/Proba Images With a Thin-Plate Spline Nonrigid Transform Model

Given the hyperspectral-oriented waveband configuration of multiangular CHRIS/Proba imagery, the scope of its application could widen if the present 18-m resolution would be improved. The multiangular images of CHRIS could be used as input for superresolution (SR) image reconstruction. A critical procedure in SR is an accurate registration of the low-resolution images. Conventional methods based on affine transformation may not be effective given the local geometric distortion in high off-nadir angular images. This paper examines the use of a nonrigid transform to improve the result of a nonuniform interpolation and deconvolution SR method. A scale-invariant feature transform is used to collect control points (CPs). To ensure the quality of CPs, a rigorous screening procedure is designed: 1) an ambiguity test; 2) the m-estimator sample consensus method; and 3) an iterative method using statistical characteristics of the distribution of random errors. A thin-plate spline (TPS) nonrigid transform is then used for the registration. The proposed registration method is examined with a Delaunay triangulation-based nonuniform interpolation and reconstruction SR method. Our results show that the TPS nonrigid transform allows accurate registration of angular images. SR results obtained from simulated LR images are evaluated using three quantitative measures, namely, relative mean-square error, structural similarity, and edge stability. Compared to the SR methods that use an affine transform, our proposed method performs better with all three evaluation measures. With a higher level of spatial detail, SR-enhanced CHRIS images might be more effective than the original data in various applications.

Increasing Robustness of Postclassification Change Detection Using Time Series of Land Cover Maps

The monitoring of land cover requires that stable land cover classes be distinguished from changes over time. Within this paper, a postclassification method is presented that provides land cover change information, based on a time series of land cover maps. The method applies a kernel filter to sequential land cover maps. Under some basic assumptions, it shows robustness against classification errors. Despite seasonality, land cover changes often occur at a low temporal frequency (e.g., maximum once every 5-10 years). If land cover maps are available more frequently, some of the information will become redundant (oversampling). The proposed method uses this redundancy for tolerating (nonsystematic) misclassifications. In order to demonstrate the benefits and limitations of the proposed method, analytical expressions have been derived. When compared to a simple postclassification comparison, one of the key strengths of the proposed approach is that it is able to improve both the overall and users accuracy of change, while also maintaining the same level of producers accuracy. As a case study, MODerate Resolution Imaging Spectroradiometer remote sensing data from 2006-2010 were classified into forest (F)/nonforest (NF) at pan-European scale. Promising results were obtained for detecting forest loss due to natural disasters. Quality was assessed using burnt area maps in southern Europe and a forest damage report after a windstorm in France. Results indicated a considerable reduction of change detection errors, confirming the theoretical results.

Increasing Spatial Detail of Burned Scar Maps Using IRS‑AWiFS Data for Mediterranean Europe

Abstract: A two stage burned scar detection approach is applied to produce a burned scar map for Mediterranean Europe using IRS-AWiFS imagery acquired at the end of the 2009 fire season. The first stage identified burned scar seeds based on a learning algorithm (Artificial Neural Network) coupled with a bootstrap aggregation process. The second stage implemented a region growing process to extend the area of the burned scars. Several ancillary datasets were used for the accuracy assessment and a final visual check was performed to refine the burned scar product. Training data for the learning algorithm were obtained from MODIS-based polygons, which were generated by the Rapid Damage Assessment module of the European Forest Fire Information System. The map produced from this research is the first attempt to increase the spatial detail of current burned scar maps for the Mediterranean region. The map has been analyzed and compared to existing burned area polygons from the European Forest Fire Information System. The comparison showed that the IRS-AWiFS-based burned scar map improved the delineation of burn scars; in addition the process identified a number of small burned scars that were not detected on lower resolution sensor data. Nonetheless, the results do not clearly support the improved capability for the detection of smaller burned scars. A number of reasons can be provided for the under-detection of burned scars, these include: the lack of a full coverage and cloud free imagery, the time lag between forest fires and image acquisition date and the occurrence of fires after the image acquisition dates. On the other hand, the limited

An Evaluation of Ecotope Classification using Superresolution Images Derived from Chris/Proba Data

This paper discusses the application of superresolution (SR) image reconstruction on multi-angle Chris/Proba images. The goal is to increase the spatial resolution of Chris/Proba images, with 18 bands from 0.4-1.0 mum in the hope to obtain a better ecotope classification. The SR approach chosen for this study is Total Variation, an iterative method which models the relationship between the desired high resolution image and the low resolution images, with the following components: a subsampling factor, a point spread function, an estimated rotation and shift, and a regularization term. This regularization approach is fast in implementation and flexible in handling noise. Efficient gradient descent methods can be used to find the desired high resolution image. The spatial resolution of the original image is improved from 25 m to 12 m using Total Variation. Subjective assessment through visual interpretation shows substantial improvement in detail. A tree-based ensemble classifier Random Forest is used for the classification of 18 ecotopes. Overall accuracy shows a 10% increase with the SR derived Chris/Proba images, compared with a classification based on the original imagery. Our results demonstrate that SR methods can improve spatial detail of multi-angle images, and subsequently classification accuracy.

Wavelet-based feature extraction for hyperspectral vegetation monitoring

The high spectral and high spatial resolution, intrinsic to hyperspectral remote sensing, result in huge quantities of data, which slows down the data processing and can result in a poor performance of classifiers. To improve the classification performance, efficient feature extraction methods are needed. This paper introduces a set of features based on the discrete wavelet transform (DWT). Wavelet coefficients, wavelet energies and wavelet detail histogram features are employed as new features for classification. As a feature reduction procedure, we propose a sequential floating search method. Selection is performed using a cost function based on the estimated probability of error, using the Fisher criterion. This procedure selects the best combination of features. To demonstrate the proposed wavelet features and selection procedure, we apply it to vegetation stress detection. For this application, it is shown that wavelet coefficients outperform spectral reflectance and that the proposed selection procedure outperforms combining the best single features.