Multi-level Semantics Research Articles

Discovery of Semantic Relationships in PolSAR Images Using Latent Dirichlet Allocation

We propose a multilevel semantics discovery approach for bridging the semantic gap when mining high-resolution polarimetric synthetic aperture radar (PolSAR) remote sensing images. First, an Entropy/Anisotropy/Alpha-Wishart classifier is employed to discover low-level semantics as classes representing the physical scattering properties of targets (e.g., low-entropy/surface scattering/high anisotropy). Then, the images are tiled into patches and each patch is modeled as a bag-of-words, a histogram of the class labels. Next, latent Dirichlet allocation is applied to discover their higher level semantics as a set of topics. Our results demonstrate that topic semantics are close to human semantics used for basic land-cover types (e.g., grassland). Therefore, using the topic description (bag-of-topics) of PolSAR images leads to a narrower semantic gap in image mining. In addition, a visual exploration of the topic descriptions helps to find semantic relationships, which can be used for defining new semantic categories (e.g., mixed land-cover types) and designing rule-based categorization schemes.

Towards Multi-Level Semantics for Multi-Agent Systems

In multi-agent systems, changes in an agent's mental state often reflect changes at the social level as well. Furthermore, most work on agent languages use structural operational semantics to formalise multi-agent systems. As multi-agent systems are complex systems, it is difficult to formalise all relevant aspects of such system as a single transition system, and even transition systems for a single agent can easily become cumbersome. In this paper, we propose a particular style of semantic rules that make it visually clearer how changes at one level of a multi-agent system require simultaneous changes in other levels of the system (where each component of each level is modelled as a separate transition system).

A computational modeling language for complex laboratory experiments

BioScapeP, a probabilistic version of BioScape, is a new modeling language for a state-dependent stochastic simulation of parallel processes in three dimensional space. Normally a modeling language describes an initial concentration of entities after which all changes are driven by the simulation of reactions. We instead design conditional simulation commands which depend on a global state. Our contribution is a new command in the form whenRrunA1⋯An, which will cause A1⋯An entities to be added to the system when property R in the context of a global state is satisfied. Commands and reactions are further equipped with user-defined update functions to produce side effects on the global state. The global state for simulation is defined to include at least a simulation clock to enable time dependent computation.BioScapeP matches the realistic nature of experimentation, by defining uncertainty from two sources: stochastic movement generating reactions on proximity, and probabilistic choice, where an entity has the ability to be involved in more than one reaction.To capture the richer notion of state dependent conditional commands, which need to be evaluated periodically, we must define a system of multi-level semantics consisting of two layers: World Level Semantics and Individual Level Semantics. Both levels take turns to evaluate their respective domains. The World Level Semantics evaluates the aforementioned when-run conditional commands, while the Individual Level Semantics simulates reactions, entity movement, and maintains the simulation clock and timed entities.

Tile-Level Annotation of Satellite Images Using Multi-Level Max-Margin Discriminative Random Field

This paper proposes a multi-level max-margin discriminative analysis (M3DA) framework, which takes both coarse and fine semantics into consideration, for the annotation of high-resolution satellite images. In order to generate more discriminative topic-level features, the M3DA uses the maximum entropy discrimination latent Dirichlet Allocation (MedLDA) model. Moreover, for improving the spatial coherence of visual words neglected by M3DA, conditional random field (CRF) is employed to optimize the soft label field composed of multiple label posteriors. The framework of M3DA enables one to combine word-level features (generated by support vector machines) and topic-level features (generated by MedLDA) via the bag-of-words representation. The experimental results on high-resolution satellite images have demonstrated that, using the proposed method can not only obtain suitable semantic interpretation, but also improve the annotation performance by taking into account the multi-level semantics and the contextual information.

ASTER/Terra Imagery and a Multilevel Semantic Network for Semi-automated Classification of Landforms in a Subtropical Area

This research is committed to develop a semi-automated landforms classification method for a subtropical area located in the southeast of Brazil, using optical medium-resolution imagery from ASTER/Terra. A four-level semantic network driven by a set of spectral, textural, and geomorphometric variables was used. The textural and geomorpho-metric variables were extracted from an ASTER/Terra DEM. The semantic network was initially conceived to classify macro morphogenetic landforms and was then further detailed to allow a finer classification, which amounted to eleven classes of morphographic landforms. In order to assess the classification accuracy, statistical indices were derived from a contingency table obtained by means of a comparison between the classified scene and a reference map. The final agreement indices for the macro and detailed landforms classifications were 76 percent and 80 percent, respectively. The employed object-based image analysis has proved to be a suitable method for semiautomated procedures in the classification of landforms.