Lieven Verbeke

WordGen: A tool for word selection and nonword generation in Dutch, English, German, and French

WordGen is an easy-to-use program that uses the CELEX and Lexique lexical databases for word selection and nonword generation in Dutch, English, German, and French. Items can be generated in these four languages, specifying any combination of seven linguistic constraints: number of letters, neighborhood size, frequency, summated position-nonspecific bigram frequency, minimum position-nonspecific bigram frequency, position-specific frequency of the initial and final bigram, and orthographic relatedness. The program also has a module to calculate the respective values of these variables for items that have already been constructed, either with the program or taken from earlier studies. Stimulus queries can be entered through WordGen’s graphical user interface or by means of batch files. WordGen is especially useful for (1) Dutch and German item generation, because no such stimulus-selection tool exists for these languages, (2) the generation of nonwords for all four languages, because our program has some important advantages over previous nonword generation approaches, and (3) psycholinguistic experiments on bilingualism, because the possibility of using the same tool for different languages increases the cross-linguistic comparability of the generated item lists. WordGen is free and available athttp://expsy.ugent.be/wordgen.htm.

Using genetic algorithms in sub-pixel mapping

In remotely sensed images, mixed pixels will always be present. Soft classification defines the membership degree of these pixels for the different land cover classes. Sub-pixel mapping is a technique designed to use the information contained in these mixed pixels to obtain a sharpened image. Pixels are divided into sub-pixels, representing the land cover class fractions. Genetic algorithms combined with the assumption of spatial dependence assign a location to every sub-pixel. The algorithm was tested on synthetic and degraded real imagery. Obtained accuracy measures were higher compared with conventional hard classifications.

A sub‐pixel mapping algorithm based on sub‐pixel/pixel spatial attraction models

Soft classification techniques avoid the loss of information characteristic to hard classification techniques when handling mixed pixels. Sub‐pixel mapping is a method incorporating benefits of both hard and soft classification techniques. In this paper an algorithm is developed based on sub‐pixel/pixel attractions. The design of the algorithm is accomplished using artificial imagery but testing is done on artificial as well as real synthetic imagery. The algorithm is evaluated both visually and quantitatively using established classification accuracy indices. The resulting images show increased accuracy when compared to hardened soft classifications.

Influence of different topographic correction strategies on mountain vegetation classification accuracy in the Lancang Watershed, China

Mapping of vegetation using remote sensing in mountainous areas is considerably hampered by topographic effects on the spectral response pattern. A variety of topographic normalization techniques have been proposed to correct these illumination effects due to topography. The purpose of this study was to compare six different topographic normalization methods (Cosine correction, Minnaert correction, C-correction, Sun-canopy-sensor correction, two-stage topographic normalization, and slope matching technique) for their effectiveness in enhancing vegetation classification in mountainous environments. Since most of the vegetation classes in the rugged terrain of the Lancang Watershed (China) did not feature a normal distribution, artificial neural networks (ANNs) were employed as a classifier. Comparing the ANN classifications, none of the topographic correction methods could significantly improve ETM+ image classification overall accuracy. Nevertheless, at the class level, the accuracy of pine forest could be increased by using topographically corrected images. On the contrary, oak forest and mixed forest accuracies were significantly decreased by using corrected images. The results also showed that none of the topographic normalization strategies was satisfactorily able to correct for the topographic effects in severely shadowed areas.

Mapping dominant vegetation communities at Meili Snow Mountain, Yunnan Province, China using satellite imagery and plant community data

Mapping dominant vegetation communities is important work for vegetation scientists. It is very difficult to map dominant vegetation communities using multispectral remote sensing data only, especially in mountain areas. However plant community data contain useful information about the relationships between plant communities and their environment. In this paper, plant community data are linked with remote sensing to map vegetation communities. The Bayesian soft classifier was used to produce posterior probability images for each class. These images were used to calculate the prior probabilities. One hundred and eighty plant plots at Meili Snow Mountain, Yunnan Province, China were used to characterize the vegetation distribution for each class along altitude gradients. Then, the frequencies were used to modify the prior probabilities of each class. After stratification in a vegetation part and a non-vegetation part, a maximum-likelihood classification with equal prior probabilities was conducted, yielding an overall accuracy of 82.1% and a kappa accuracy of 0.797. Maximum-likelihood classification with modified prior probabilities in the vegetation part, conducted with a conventional maximum-likelihood classification for the non-vegetation part, yielded an overall accuracy of 87.7%, and a kappa accuracy of 0.861.

Semi-automated forest stand delineation using wavelet based segmentation of very high resolution optical imagery

Stand delineation is one of the cornerstones of forest inventory mapping and a key element to spatial aspects in forest management decision making. Stands are forest management units with similarity in attributes such as species composition, density, closure, height and age. Stand boundaries are traditionally estimated through subjective visual air photo interpretation. In this paper, an automatic stand delineation method is presented integrating wavelet analysis into the image segmentation process. The new method was developed using simulated forest stands and was subsequently applied to real imagery: scanned aerial photographs of a forest site in Belgium and ADS40 aerial digital data of an olive grove site in Les Beaux de Provence, France. The presented method was qualitatively and quantitatively compared with traditional spectral based segmentation, by assessing its ability to support the creation of pure forest stands and to improve classification performance. A parcel/stand purity index was developed to evaluate stand purity and the expected mapping accuracy was estimated by defining a potential mapping accuracy measure. Results showed that wavelet based image segmentation outperformed traditional segmentation. Multi-level wavelet analysis proved to be a valuable tool for characterizing local variability in image texture and therefore allowed for the discrimination between stands. In addition, the proposed evaluation measures were found appropriate as segmentation evaluation criteria.

EPSILON: an eQTL prioritization framework using similarity measures derived from local networks

MOTIVATION When genomic data are associated with gene expression data, the resulting expression quantitative trait loci (eQTL) will likely span multiple genes. eQTL prioritization techniques can be used to select the most likely causal gene affecting the expression of a target gene from a list of candidates. As an input, these techniques use physical interaction networks that often contain highly connected genes and unreliable or irrelevant interactions that can interfere with the prioritization process. We present EPSILON, an extendable framework for eQTL prioritization, which mitigates the effect of highly connected genes and unreliable interactions by constructing a local network before a network-based similarity measure is applied to select the true causal gene. RESULTS We tested the new method on three eQTL datasets derived from yeast data using three different association techniques. A physical interaction network was constructed, and each eQTL in each dataset was prioritized using the EPSILON approach: first, a local network was constructed using a k-trials shortest path algorithm, followed by the calculation of a network-based similarity measure. Three similarity measures were evaluated: random walks, the Laplacian Exponential Diffusion kernel and the Regularized Commute-Time kernel. The aim was to predict knockout interactions from a yeast knockout compendium. EPSILON outperformed two reference prioritization methods, random assignment and shortest path prioritization. Next, we found that using a local network significantly increased prioritization performance in terms of predicted knockout pairs when compared with using exactly the same network similarity measures on the global network, with an average increase in prioritization performance of 8 percentage points (P < 10(-5)). AVAILABILITY The physical interaction network and the source code (Matlab/C++) of our implementation can be downloaded from http://bioinformatics.intec.ugent.be/epsilon. CONTACT lieven.verbeke@intec.ugent.be, kamar@psb.ugent.be, jan.fostier@intec.ugent.be SUPPLEMENTARY INFORMATION Supplementary data are available at Bioinformatics online.

Previously trained neural networks as ensemble members: knowledge extraction and transfer

The use of Artificial Neural Networks (ANNs) for the classification of remotely sensed imagery offers several advantages over more conventional methods. Yet their training still requires a set of pixels with known land cover. To increase ANN classification accuracy when few training data are available, an algorithm was applied that allows experience gained in previous classifications to be reused. The proposed method was evaluated by classifying a tropical savannah region in northern Togo using Landsat Thematic Mapper (TM) imagery. The presented approach reached a mean kappa coefficient that was significantly larger (at the 95% level) than that obtained after training networks with randomly initialized weights. Also, the observed variances on the obtained accuracies were significantly lower when compared to networks that were randomly initialized. Finally, Bhattacharyya (BH) distances were used to explain why some land cover classes benefit more from knowledge transfer than others.

Training neural networks on artificially generated data: a novel approach to SAR speckle removal

A neural network-based method for speckle removal in synthetic aperture radar (SAR) images is introduced. The method rests on the idea that a neural network learning machine, trained on artificially generated input–target couples, can be used to efficiently process real SAR data. The explicit plus-point of the method is that it is trained with artificially generated data, reducing the demands put on real input data such as data quality, availability and cost price. The artificial data can be generated in such a way that they fit the particular characteristics of the images to be denoised, yielding case-specific, high-performing despeckling filters. A comparative study with three classical denoising techniques (Enhanced Frost (EF), Enhanced Lee (EL) and Gamma MAP (GM)) and a wavelet filter demonstrated a superior speckle removal performance of the proposed method in terms of quantitative performance measures. Moreover, qualitative evaluation of the despeckled results was in favour of the proposed method, confirming its speckle removal efficiency.

Wavelet-based texture measures for semicontinuous stand density estimation from very high resolution optical imagery

Stand density, expressed as the number of trees per unit area, is an important forest management parameter. It is used by foresters to evaluate regeneration, to assess the effect of forest management measures, or as an indicator variable for other stand parameters like age, basal area, and volume. In this work, a new density estimation procedure is proposed based on wavelet analysis of very high resolution optical imagery. Wavelet coefficients are related to reference densities on a per segment basis, using an artificial neural network. The method was evaluated on artificial imagery and two very high resolution datasets covering forests in Heverlee, Belgium and Les Beaux de Provence, France. Whenever possible, the method was compared with the well-known local maximum filter. Results show good correspondence between predicted and true stand densities. The average absolute error and the correlation between predicted and true density was 149 trees/ha and 0.91 for the artificial dataset, 100 trees/ha and 0.85 for the Heverlee site, and 49 trees/ha and 0.78 for the Les Beaux de Provence site. The local maximum filter consistently yielded lower accuracies, as it is essentially a tree localization tool, rather than a density estimator.