Alain Boucher

Towards ontology-based cognitive vision

Abstract.This paper details a visual-concept-ontology-driven knowledge acquisition methodology. We propose to use a visual concept ontology to guide experts in the visual description of the objects of their domain (e.g., pollen grain). The proposed knowledge acquisition process results in a knowledge base enabling semantic image interpretation. An important benefit of our approach is that the knowledge acquisition process guided by the ontology leads to a knowledge base close to low-level vision. A visual concept ontology and a dedicated knowledge acquisition tool have been developed and are presented. We propose a generic methodology that is not linked to any application domain. An example shows how the knowledge acquisition model can be applied to the description of pollen grain images.

GAMA: An Environment for Implementing and Running Spatially Explicit Multi-agent Simulations

In this paper, we introduce the GAMA (Gis & Agent-based Modelling Architecture) simulation platform, which aims at providing field experts, modellers, and computer scientists with a complete modelling and simulation development environment for building spatially explicit multi-agent simulations. The most important requirements of spatially explicit multi-agent simulations that our platform fulfils are: (1) the ability to transparently use complex Geographical Information System (GIS) data as an environment for the agents; (2) the ability to handle a vast number of (heterogeneous) agents (3); the ability to offer a platform for automated controlled experiments (by automatically varying parameters, recording statistics, etc.); (4) the possibility to let non-computer scientists design models and interact with the agents during simulations. While still in its implementation phase, the platform is currently used for two main applications. One is about the modelling of the spread of avian influenza in a province of North Vietnam in collaboration with CIRAD (French Agricultural Research Centre working for International Development). Its goal is to simulate the poultry value chain of a whole province using geolocalised data, and to use this to optimise a monitoring network. A second application conducted with the Institute for Marine Geology and Geophysics (VAST, Hanoi) is about using an interactive simulation for supporting decision-making during urban post-disaster situations. This application relies on geolocalised data as well, and requires facilities of interaction between users and the simulation.

Towards ontology based cognitive vision

Building knowledge bases for knowledge-based vision systems is a difficult task. This paper aims at showing how an ontology composed of visual concepts can be used as a guide for describing objects from a specific domain of interest. One of the most important benefits of our approach is that the knowledge acquisition process guided by the ontology leads to a knowledge base closer to low-level vision. A visual concept ontology and a dedicated knowledge acquisition tool have been developed and are also presented. We propose a generic methodology that is not linked to any application domain. Nevertheless, an example shows how the knowledge acquisition model can be applied to the description of pollen grain images. The use of an ontology for image description is the first step towards a complete cognitive vision system that will involve a learning layer.

Development of a semi-automatic system for pollen recognition

A semi-automatic system for pollen recognitionis studied for the european project ASTHMA. The goal of such a system is to provideaccurate pollen concentration measurements. This information can be used as well by thepalynologists, the clinicians or a forecastsystem to predict pollen dispersion. At first,our emphasis has been put on Cupressaceae, Olea, Poaceae and Urticaceae pollen types. The system is composed of two modules: pollengrain extraction and pollen grain recognition. In the first module, the pollen grains areobserved in light microscopy and are extractedautomatically from a pollen slide coloured withfuchsin and digitized in 3D. In the secondmodule, the pollen grain is analyzed forrecognition. To accomplish the recognition, itis necessary to work on 3D images and to usedetailed palynological knowledge. Thisknowledge describes the pollen types accordingto their main visible characteristerics and tothose which are important for recognition. Somepollen structures are identified like the porewith annulus in Poaceae, the reticulum in Oleaand similar pollen types or the cytoplasm inCupressaceae. The preliminary results show therecognition of some pollen types, likeUrticaceae or Poaceae or some groups of pollentypes, like reticulate group.

A novel approach for text detection in images using structural features

We propose a novel approach for finding text in images by using ridges at several scales. A text string is modelled by a ridge at a coarse scale representing its center line and numerous short ridges at a smaller scale representing the skeletons of characters. Skeleton ridges have to satisfy geometrical and spatial constraints such as the perpendicularity or non-parallelism to the central ridge. In this way, we obtain a hierarchical description of text strings, which can provide direct input to an OCR or a text analysis system. The proposed method does not depend on a particular alphabet, it works with a wide variety in size of characters and does not depend on orientation of text string. The experimental results show a good detection.

A society of goal-oriented agents for the analysis of living cells

This paper presents a new model for the segmentation and analysis of living cells. A multi-agent model has been developed for this application. It is based on a generic agent model, which is composed of different behaviors: perception, interaction and reproduction. The agent is further specialized to accomplish a specific goal. Different goals are defined from the different components of the cell images. The specialization specifies the parameters of the behaviors for the achievement of the agents goal. From these goal-oriented agents, a society is defined, and it evolves dynamically as the agents are created and deleted. An internal manager is integrated in the agent to control the behaviors execution. It makes use of an event-driven scheme to manage the behavior priorities. The present design is mainly oriented toward image segmentation, however, it includes some features on tracking and motion analysis.

A query language combining object features and semantic events for surveillance video retrieval

In this paper, we propose a novel query language for video indexing and retrieval that (1) enables to make queries both at the image level and at the semantic level (2) enables the users to define their own scenarios based on semantic events and (3) retrieves videos with both exact matching and similarity matching. For a query language, four main issues must be addressed: data modeling, query formulation, query parsing and query matching. In this paper we focus and give contributions on data modeling, query formulation and query matching. We are currently using color histograms and SIFT features at the image level and 10 types of events at the semantic level. We have tested the proposed query language for the retrieval of surveillance videos of a metro station. In our experiments the database contains more than 200 indexed physical objects and 48 semantic events. The results using different types of queries are promising.

Subtrajectory-based video indexing and retrieval

This paper proposes an approach for retrieving videos based on object trajectories and subtrajectories. First, trajectories are segmented into subtrajectories according to the characteristics of the movement. Efficient trajectory segmentation relies on a symbolic representation and uses selected control points along the trajectory. The selected control points with high curvature capture the trajectory various geometrical and syntactic features. This symbolic representation, beyond the initial numeric representation, does not suffer from scaling, translation or rotation. Then, in order to compare trajectories based on their subtrajectories, several matching strategies are possible, according to the retrieval goal from the user. Moreover, trajectories can be represented at the numeric, symbolic or the semantic level, with the possibility to go easily from one representation to another. This approach for indexing and retrieval has been tested with a database containing 2500 trajectories, with promising results.

A framework for surveillance video indexing and retrieval

We propose a framework for surveillance video indexing and retrieval. In this paper, we focus on the following features: (1) combine recognized video contents (output from a video analysis module) with visual words (computed over all the raw video frames) to enrich the video indexation in a complimentary way; using this scheme user can make queries about objects of interest even when the video analysis output is not available; (2) support an interactive feature generation (currently color histogram and trajectory) that gives a facility for users to make queries at different levels according to the a priori available information and the expected results from retrieval; (3) develop a relevance feedback module adapted to the proposed indexing scheme and the specific properties of surveillance videos for the video surveillance context. Results emphasizing these three aspects prove a good integration of video analysis for video surveillance and interactive indexing and retrieval.

Cell Migration Analysis After In Vitro WoundingInjury with a Multi-Agent Approach

This paper presents a multiagent system for studying in vitro cell motion. A typical application on the wound closure process is presented to illustrate the possibilities of the system, where different image sequences will be treated. The motion issue involves three aspects: image segmentation, object tracking and motion analysis. The current system version focuses mainly on the image segmentation aspect. A general agent model has been designed, which will be further expanded to include tracking and motion analysis behaviors as well. The agents integrate three basic behaviors: perception, interaction and reproduction. The perception evaluates pixels upon static and motion-based criteria. The interaction behavior allows two agents to merge or to negotiate parts of regions. The negotiation can be seen as a segmentation refinement process done by the agents. Finally, the reproduction behavior defines an exploration strategy of the images. Agents can start other agents around them, or they can duplicate themselves in the next frame. The frames are processed in pipeline, where previous information is used to treat the current frame. One unique agent model exist. Agents are specialized on execution time according to their goals. The results, coming from an existing prototype, show different types of cell behavior during cell migration, based on cell nuclei analysis.