Search

The main challenge of new information technologies is to retrieve intelligible information from the large volume of digital data gathered every day. Among the variety of existing data sources, the satellites continuously observing the surface of the Earth are key to the monitoring of our environment. The new generation of satellite sensors are tremendously increasing the possibilities of applications but also increasing the need for efficient processing methodologies in order to extract information relevant to the users' needs in an automatic or semi-automatic way. This is where machine learning comes into play to transform complex data into simplified products such as maps of land-cover changes or classes by learning from data examples annotated by experts. These annotations, also called labels, may actually be difficult or costly to obtain since they are established on the basis of ground surveys. As an example, it is extremely difficult to access a region recently flooded or affected by wildfires. In these situations, the detection of changes has to be done with only annotations from unaffected regions. In a similar way, it is difficult to have information on all the land-cover classes present in an image while being interested in the detection of a single one of interest. These challenging situations are called novelty detection or one-class classification in machine learning. In these situations, the learning phase has to rely only on a very limited set of annotations, but can exploit the large set of unlabeled pixels available in the images. This setting, called semi-supervised learning, allows significantly improving the detection. In this Thesis we address the development of methods for novelty detection and one-class classification with few or no labeled information. The proposed methodologies build upon the kernel methods, which take place within a principled but flexible framework for learning with data showing potentially non-linear feature relations. The thesis is divided into two parts, each one having a different assumption on the data structure and both addressing unsupervised (automatic) and semi-supervised (semi-automatic) learning settings. The first part assumes the data to be formed by arbitrary-shaped and overlapping clusters and studies the use of kernel machines, such as Support Vector Machines or Gaussian Processes. An emphasis is put on the robustness to noise and outliers and on the automatic retrieval of parameters. Experiments on multi-temporal multispectral images for change detection are carried out using only information from unchanged regions or none at all. The second part assumes high-dimensional data to lie on multiple low dimensional structures, called manifolds. We propose a method seeking a sparse and low-rank representation of the data mapped in a non-linear feature space. This representation allows us to build a graph, which is cut into several groups using spectral clustering. For the semi-supervised case where few labels of one class of interest are available, we study several approaches incorporating the graph information. The class labels can either be propagated on the graph, constrain spectral clustering or used to train a one-class classifier regularized by the given graph. Experiments on the unsupervised and oneclass classification of hyperspectral images demonstrate the effectiveness of the proposed approaches.

Assessing pattern printability in new large layouts faces important challenges of runtime and false detection. Lithographic simulation tools and classification techniques do not scale well. We propose a fast pattern detection method by learning an overcomplete basis representing each reference pattern. A pattern from a new design is detected “novel” if its reconstruction error, when coded in the learned basis, is large. We show high speedup (1000x) compared to nearest neighbor search. A new boundary detection technique selects the minimal set of the novel patterns to predict problematic patterns; 14.93% of the novel patterns suffice to predict ORC hotspots, while 53.77% are needed using traditional methods.

Candidate layout patterns can be assessed using a sparse pattern dictionary of known design layout patterns by determining sparse coefficients for each candidate pattern, reconstructing the respective candidate pattern, and determining reconstruction error. Any pattern with reconstruction error over a threshold value can be flagged. Compressive sampling can be employed, such as by projecting each candidate pattern onto a random line or a random matrix. The dictionary can be built by determining sparse coefficients of known patterns and respective basis function sets using matching pursuit, variants of SVD, and/or other techniques.

The database of satellite images covering the earth is growing extremely quickly and represents an important amount of data. The extension of this huge amount of data with images from very different sources and time of acquisition can provide a wide range of applications. The climate change monitoring could be more easily achieved or Natural disasters (deforestation, floods, dry rivers, earthquake) characterized by Unmanned Aerial Vehicle (UAV), allowing a rapid intervention. This project aims to realize a robust registration of images from different modalities (view angle, resolution, sensors sensitivity) having temporal changes and possible "novelties". A novelty would be a change not due to normal season changes, like with new buildings or natural disaster. An introduction to the application of novelty detection on multi-modal and multi-temporal images is discussed based on recent research on novelty detection. Video recordings from UAV are provided by RUAG. Different flight conditions have been realized to get a wide range of images, in terms of content, view angle, field of view or stabilization. The images, taken as reference, are from satellite (SPOT 5) and airborne (Google Map) sources. A registration algorithm based on the phase-correlation allowing to retrieve rotation, scaling and translation between two images is proposed. It handles the multi-modal characteristics of the images by correcting the perspective deformation, exploiting the redundancy of video sequence and adaptively choose between the image edges or the extraction of large structure from the image depending on the image frequency content. The results are encouraging for the possibility of novelty detection. The algorithm is sensitive to the initial conditions but accurate registration with stabilized and unstabilized flights is achieved using Google Maps. Less precise registration is achieved with SPOT images coming from their lower resolution.

The analysis of multi-modal and multi-sensor images is nowadays of paramount importance for Earth Observation (EO) applications. There exist a variety of methods that aim at fusing the different sources of information to obtain a compact representation of such datasets. However, for change detection existing methods are often unable to deal with heterogeneous image sources and very few consider possible nonlinearities in the data. Additionally, the availability of labeled information is very limited in change detection applications. For these reasons, we present the use of a semi-supervised kernel-based feature extraction technique. It incorporates a manifold regularization accounting for the geometric distribution and jointly addressing the small sample problem. An exhaustive example using Landsat 5 data illustrates the potential of the method for multi-sensor change detection.