The application of drill core hyperspectral data in exploration campaigns is receiving great interest to obtain a general overview of a mineral deposit. However, the main approach to the investigation of such data is by visual interpretation, which is subjective and time consuming. To address this issue, recently, the use of machine learning techniques is proposed for the analysis of this data. For drill core samples that for which only very little prior knowledge is often available, applying classification algorithms which are supervised learning methods is very challenging. In this paper, we suggest to use clustering (unsupervised) methods for mineral mapping, which are similar to classification but no predefined class labels are needed. To handle mapping of the very highly mixed pixels in drill core hyperspectral data, we propose to use advance subspace clustering methods, in which pixels are assumed to lie in a union of low-dimensional subspaces. We conduct a comparative study and evaluate the performance of two well-known subspace clustering methods, i.e., sparse subspace clustering (SSC) and low rank representation (LRR). For the experiments, we acquired VNIR-SWIR hyperspectral data and applied scanning electron microscopy based Mineral Liberation Analysis (MLA) for two drill core samples. MLA is a high resolution imaging technique that allows detailed mineral charactrisition. We use the high-resolution MLA image as a reference to analyse the clustering results. Qualitative analysis of the obtained clustering maps indicate that the subspace clustering methods can accurately map the available minerals in the drill core hyperspectral data, especially in comparison to the traditional k-means clustering method.
The amount of radar sounder data, which are used to analyze the subsurface of icy environments (e.g., Poles of Earth and Mars), is dramatically increasing from both airborne campaigns at the ice sheets and satellite missions on other planetary bodies. However, the main approach to the investigation of such data is by visual interpretation, which is subjective and time consuming. Moreover, the few available automatic techniques have been developed for analyzing highly reflective subsurface targets, e.g., ice layers, basal interface. Besides the high reflective targets, glaciologists have also shown great interest in the analysis of non-reflective targets, such as the echo-free zone in ice sheets, and the reflective free zone in the subsurface of the South Pole of Mars. However, in the literature, there is no dedicated automatic technique for the analysis of non-reflective targets. To address this limitation, we propose an automatic classification technique for the identification of non-reflective targets in radar sounder data. The method is made up of two steps, i.e., i) feature extraction, which is the core of the method, and ii) automatic classification of subsurface targets. We initially prove that the commonly employed features for the analysis of the radar signal (e.g., statistical and texture based features) are ineffective for the identification of non-reflective targets. Thus, for feature extraction, we propose to exploit structural information based on the morphological closing profile. We show the effectiveness of such features in discriminating of non-reflective target from the other ice subsurface targets. In the second step, a random forest classifier is used to perform the automatic classification. Our experimental results, conducted using two data sets from Central Antarctica and South Pole of Mars, point out the effectiveness of the proposed technique for the accurate identification of non-reflective targets.
Subglacial lakes decouple the ice sheet from the underlying bedrock, thus facilitating the sliding of the ice masses towards the borders of the continents, consequently raising the sea level. This motivated increasing attention in the detection of subglacial lakes. So far, about 70% of the total number of subglacial lakes in Antarctica have been detected by analysing radargrams acquired by radar sounder (RS) instruments. Although the amount of radargrams is expected to drastically increase, from both airborne and possible future Earth observation RS missions, currently the main approach to the detection of subglacial lakes in radargrams is by visual interpretation. This approach is subjective and extremely time consuming, thus difficult to apply to a large amount of radargrams. In order to address the limitations of the visual interpretation and to assist glaciologists in better understanding the relationship between the subglacial environment and the climate system, in this paper, we propose a technique for the automatic detection of subglacial lakes. The main contribution of the proposed technique is the extraction of features for discriminating between lake and non-lake basal interfaces. In particular, we propose the extraction of features that locally capture the topography of the basal interface, the shape and the correlation of the basal waveforms. Then, the extracted features are given as input to a supervised binary classifier based on Support Vector Machine to perform the automatic subglacial lake detection. The effectiveness of the proposed method is proven both quantitatively and qualitatively by applying it to a large dataset acquired in East Antarctica by the MultiChannel Coherent Radar Depth Sounder.