Indoor Place Recognition using Online Independent Support Vector Machines

Abstract

In the framework of indoor mobile robotics, place recognition is a challenging task, where it is crucial that self-localization be enforced precisely, notwithstanding the changing conditions of illumination, objects being shifted around and/or people affecting the appearance of the scene. In this scenario online learning seems the main way out, thanks to the possibility of adapting to changes in a smart and flexible way. Nevertheless, standard machine learning approaches usually suffer when confronted with massive amounts of data and when asked to work online. Online learning requires a high training and testing speed, all the more in place recognition, where a continuous flow of data comes from one or more cameras. In this paper we follow the Support Vector Machines-based approach of Pronobis et al. [26], proposing an improvement that we call Online Independent Support Vector Machines. This technique exploits linear independence in the image feature space to incrementally keep the size of the learning machine remarkably small while retaining the accuracy of a standard machine. Since the training and testing time crucially depend on the size of the machine, this solves the above stated problems. Our experimental results prove the effectiveness of the approach.

Towards robust place recognition for robot localization

“…Additionally, information about the confidence of the final decision could be utilized to further improve reliability [6]. Also, the possibility to categorize should be exploited for knowledge transfer, as proposed in [32], and for life long learning, as suggested in [7], [33]; we intend to explore both possibilities.…”

Semantic Modelling of Space

“…Approximate techniques [91,92,90] seem to be better suited for our problem because, at each incremental step, they discard non-informative training vectors, thus reducing the memory requirements. Other methods, such as [93,94], instead require storing in memory all the training data. The basic principle behind the memory-controlled method is to combine the fixed-partition incremental extension [92] with an algorithm for controlling the memory growth [95].…”

Towards generalization of semi-supervised place classification over generalized Voronoi graph

“…For this purpose, a few variations of the original SVM algorithm have been proposed. For example, algorithms like online independent-SVM (OISVM) and memory-controlled incremental SVM do not require storing all incoming data, and have selection mechanisms to guarantee a bounded memory growth [11,26,33]. These approaches focus more on the algorithmic efficiency and can be further improved by considering the spatial context.…”

“…This is compared with place recognition, which refers to the ability to recognize previously seen parts of the environment [9][10][11]. A commonly held outlook is that place classification is a more challenging problem due to the presence of higher intra-class variations which warrants the formation of a conceptual model of the place [12][13][14].…”

“…In general terms, the current place classification approaches face a variety of challenges including changes in appearances and illumination conditions [9,11], in-class variations due to discrepancies in shape, color and texture across individual samples [23], the influence of choosing training and test data on the overall performance of the system [9], and the interference of dynamic environments and human activities [9,11,26]. Many of these problems can be summarized as the reflections of the generalization ability of the system, which may be improved by incorporating contextual information and adopting semisupervised learning algorithms.…”