Margarita Bachiller

On the correspondence between objects and events for the diagnosis of situations in visual surveillance tasks

A key problem in visual surveillance systems (VSS) is to find an effective procedure for linking the geometric descriptions of a scene at the object level with the corresponding descriptions of the agents intervening in this scene at the activity level. In this work, we explore a constructivist approach based on using the usual Artificial Intelligence (AI) techniques and methods to establish correspondences between the entities and relations of the ontologies in these two levels. The proposal is exemplified using a real interior scenario that uses images from just one fixed camera and where the purpose of the surveillance is to do a preventive diagnosis of the activity of abandoning a potentially dangerous object in a sensitive area. The work stresses: (1) anchoring the object-level labels in the result of analytical processes on blobs, (2) specifying contextual knowledge that has to be injected to link the activities, as described by a human surveillance expert, with the objects, as they are labelled by the same expert from geometric descriptions. The work is set within the context of the 50th anniversary of AI and the leading theories on human visual perception.

On the effect of feedback in multilevel representation spaces for visual surveillance tasks

In this work we propose a general top-down feedback scheme between adjacent description levels to interpret video sequences. This scheme distinguishes two types of feedback: repair-oriented feedback and focus-oriented feedback. With the first it is possible to improve the systems performance and produce more reliable and consistent information, and with the second it is possible to adjust the computational load to match the aims. Finally, the general feedback scheme is used in different examples for a visual surveillance application which improved the final result of each description level by using the information in the higher adjacent level.

Knowledge modeling for the image understanding task as a design task

This paper formally analyzes the image understanding (IU) task at the knowledge level and in the observer domain. The analysis is done at three levels: task, method and domain knowledge, distinguishing the generic components in most of the IU tasks, thereby enabling the components to be reused. We model the IU problem as a design task and define a generic problem solving method (PSM) that allows us to tackle the task in a hierarchical and recursive way. The main advantage of this generic PSM is the possibility of instantiating specific PSMs through parameter space configuration, which enables the structure to be reused both across the task decomposition at different hierarchical levels and across different application domains. This work has been done following the well-established foundations of knowledge engineering that prescribe the maintenance of the conceptual structure from the modeling stage at the knowledge level down to the implementation. Finally, we apply the proposed framework to the problem of identifying the papilla in eye fundus images in order to exemplify the successive stages in the modeling process and system design, and accordingly justify the frameworks validity.

Segmentation of Moving Objects with Information Feedback Between Description Levels

In real sequences, one of the factors that most negatively affects the segmentation process result is the existence of scene noise. This impairs object segmentation which has to be corrected if we wish to have some minimum guarantees of success in the following tracking or classification stages. In this work we propose a generic knowledge-based model to improve the segmentation process. Specifically, the model uses a decomposition strategy in description levels to enable the feedback of information between adjacent levels. Finally, two case studies are proposed that instantiate the model proposed for detecting humans.

A Modular Scheme for Controller Design and Performance Evaluation in 3D Visual Servoing

In this paper we present a modular scheme for designing and evaluating different control systems for position based dynamic look and move visual servoing systems. This scheme is particularly applied to a 6 DOF industrial manipulator equipped with a camera mounted on its end effector. The manipulator with its actuators and its current feedback loops can be modeled as a Cartesian device commanded through a serial line. In this case the manipulator can be considered as a decoupled system with 6 independent loops. The use of computer vision as feedback transducer strongly affects the closed loop dynamics of the overall system, so that a visual controller is required for achieving fast response and high control accuracy. Due to the long delay in generating the control signal, it is necessary to carefully select the visual controller. In this paper we present a framework that allows the study of some conventional and new techniques to design this visual controller. Besides an experimental setup has been built and used to evaluate and compare the performance of the position based dynamic look and move system with different controllers. Some criterions for selecting the best strategy for each task are established. Quite a lot of results relative to different trajectory tracking control strategies are presented, showing both simulation and real platform responses.

An Automatic System for the Location of the Optic Nerve Head from 2D Images

In this paper, we present a vision system in order to improve the diagnosis process of patients with glaucoma. The more accurate mean of assessing is to study the optic nerve head directly from ocular fundus images. Due to the complexity of the problem, our approach decomposes it into simpler subtasks until the primitive level is reached. An example of operationalization of each primitive is shown. Besides a number of experiments were performed in order to detect the papilla contour on real medical images that confirm the proposed operationalization. The main advantages of the system designed are the elimination of the subjectivity that exists in the process of identifying the objects that are present in the ocular fundus image and the full automatization of the process.

A Block-Based Human Model for Visual Surveillance

This paper presents BB6-HM, a block-based human model for real-time monitoring of a large number of visual events and states related to human activity analysis, which can be used as components of a library to describe more complex activities in such important areas as surveillance. BB6-HM is inspired by the proportionality rules commonly used in Visual Arts, i.e., for dividing the human silhouette into six rectangles of the same height. The major advantage of this proposal is that analysis of the human can be easily broken down into parts, which allows us to introduce more specific domain knowledge and to reduce the computational load. It embraces both frontal and lateral views, is a fast and scale-invariant method and a large amount of task-focused information can be extracted from it.

Semiautomatic segmentation of the medial temporal lobe anatomical structures

Medial temporal lobe (MTL) is a region of the brain related with processing and declarative memory consolidation. Structural changes in this region are directly related with Alzheimers disease and other dementias. Manual delimitation of these structures is very time consuming and error prone. Automatic methods are needed in order to solve these problems and make it available in the clinical practice. Unfortunately, automatic methods are not robust enough yet. The use of semiautomatic methods provides an intermediate solution with the advantages of automatic methods under the supervision of the expert. This paper propose two semiautomatic methods oriented to make the delineation of the MTL structures easy, robust and fast.

Image understanding analysis at the knowledge level as a design task

It is broadly assumed that the incorporation of intelligence in computer-based devices should increase their capabilities. We lack an operational understanding of the elements that should be implemented in an artificial system to deserve the attribute of intelligent. Meanwhile, the devices are rather characterized by showing an interface able to convince naive users but lacking real intelligence. It could be worth to analyze the physiological properties of natural intelligence to draw conclusions on the operational properties that could be effectively implemented into an artificial system. It is stated that behaviors based just in programs, whether in natural or artificial systems, do not contain signs of intelligence. The emergence of intelligence is dissected and shown that at least in basic conditions this function can be explained in terms of associative learning, leveraged by some properties that result specifically apparent in humans: unlimited exploratory activity, apparent absence of genetically defined aims, transgenerational transference of information, and generalization and symbolic manipulation capabilities. It is proposed that the interest of the field should move towards the definition of rules to instruct an associative learning machine on adaptiveness.

Tridimensional part modelling based on shape global features. Applications to recognition of parts measure by range finder sensors

Abstract A new tridimensional modelling using a set of shape global features and a method for solving the 3D part recognition problem is described in this paper. The method uses a new set of global features as discriminant parameters. They are extracted from a particularly defined spherical representation model of the object. The proposed approach avoids to apply local features based techniques, which usually involve much more computational effort at matching stage. By the contrary, a very simple type of information is associated to each object in such a way that it can be easily distinguished from the others. The method has been successfully applied to a medium database corresponding to real parts sensed in a platform with a Gray Range Finder sensor. Polyhedral and curved objects have been considered. Experimental results and future research work are presented.