Semisupervised One-Class Support Vector Machines for Classification of Remote Sensing Data

Abstract

Abstract-This paper presents two semisupervised one-class support vector machine (OC-SVM) classifiers for remote sensing applications. In one-class image classification, one tries to detect pixels belonging to one of the classes in the image and reject the others. When few labeled pixels of only one class are available, obtaining a reliable classifier is a difficult task. In the particular case of SVM-based classifiers, this task is even harder because the free parameters of the model need to be finely adjusted, but no clear criterion can be adopted. In order to improve the OC-SVM classifier accuracy and alleviate the problem of free-parameter selection, the information provided by unlabeled samples present in the scene can be used. In this paper, we present two state-of-the-art algorithms for semisupervised one-class classification for remote sensing classification problems. The first proposed algorithm is based on modifying the OC-SVM kernel by modeling the data marginal distribution with the graph Laplacian built with both labeled and unlabeled samples.

Graph Embedded One-Class Classifiers for media data classification

“…Here we should also note that local geometric information expressed by kNN graphs has also been exploited in a semi-supervised one-class classification setting, as in [3] where the Laplacian One-Class SVM (LAP-OC-SVM) is proposed.…”

“…In our previous example, it is expected that the leading actor will appear multiple times during the entire movie, and he (or she) will be easier to be identified by an annotator. In order to efficiently model a class of interest in media classification tasks, we consider the use of One-Class Classification (OCC) methods [1,2,3,4,5]. Related OCC applications include hyperspectral image classification [3], video summarization [6,7], image segmentation [8].…”

“…In order to efficiently model a class of interest in media classification tasks, we consider the use of One-Class Classification (OCC) methods [1,2,3,4,5]. Related OCC applications include hyperspectral image classification [3], video summarization [6,7], image segmentation [8]. Other OCC use case scenarios include applications when only one class is well sampled and must be distinguished from every other possibility, e.g., medical diagnostic problems, faults and failure detection, video surveillance and mobile fraud detection [9].…”

“…In [19], kNN graphs are employed for manifold regularization, so that local geometric data relationships between labeled and unlabeled data are expressed in the deformed space. The corresponding semi-supervised OCC method has been proposed in [3].…”

“…Thus, employing the methods proposed in [10] and [20] does not model the subclass properties of the target class, sufficiently. To this end, one could employ kNN graphs in the OC-SVM optimization process as in [3] (for the supervised classification case). By employing kNN graphs, local geometric data relationships between the training data may be sufficiently modeled.…”