Ratko Obradovic

A Smart House environment - the System of Systems approach to Model Driven Simulation of Building (house) Attributes

The System-of-Systems research area, with its opened challenging approach to engineering practice transformation, has a significant impact to the closely related problem domains cooperation. The Architectural Design, Urban Planning, and Construction Engineering Domains, even when treated separately, are extremely demanding targets for sophisticated software tools development. While managing the complex urban blocks interactions in the contemporary city configuration, it is particularly important that these three domains cooperate closely. The application of a smart house paradigm is dominantly focused on the internal facility parameters monitoring and control. For the subset of these internal parameters, like for example: daylight illumination; privacy; and natural ventilation, the facility (building) environment impacts analysis is unavoidable. In this article, based on the System-of-Systems paradigm, we present the model of a smart house environment impact analysis framework suitable for Model Driven Simulations of Building (house) Attributes. The proposed framework architecture is designed to support the cooperation/collaboration of software tools. The verification has shown that it may be effectively used for the impact analysis of arbitrary external (urban block specific) attributes on their internal state and the behavior (specific for a single building).

Epicardial fat registration by local adaptive morphology-thresholding based 2D segmentation

3D heart registration has become an important issue in cardio vascular diagnosis and treatment. This is mainly due to advanced medical imaging technologies that provide significant amount of data with high precision. One of the important features of the heart that has recently drawn attention is epicardial fat (surrounds the heart), which according to some preliminary studies can be correlated well with risk prediction of various cardiovascular diseases. Consequently, automatic detection and registration of epicardial fat is considered as important task for medical doctors to include as additional feature within the already existing software for medical imaging and visualization. In this paper, we analyze heart images obtained by 4D CT technology and propose a segmentation scheme that automatically extracts epcardial fat in each 2D slice in order to perform 3D epicardial fat registration and visualization. The segmentation algorithm first enhances input image after which it performs patch based labeling and clustering of the selected features. The experimental results indicate good epicardial fat registration performance in comparison to manual segmentation obtained by the medical doctors.

Teaching 3D Character Animation Through Four Related Tasks

This paper presents the content of the course Fundamentals of Engineering Animation (FEA). The course FEA deals with manlike character animation through four tasks. Before taking FEA the students have mastered 3D character modeling, mapping, lighting and rendering. Also, they are familiar with rigid body animation, as well as with simple animal animation requiring rigging, like fish or animals with simple skeleton. The above mentioned tasks are increasingly more complex and this paper presents the topics of these tasks, details about them and their sequence as well as the evaluation of students’ work. The students’ project requires creating a 3D character mesh as well as rigging and making controls for two cases: Inverse Kinematik (IK) and Forward Kinematik (FK). The first project task (FPT) represents the creation of a simple movement in a sport with several phases: Hold, Anticipation, Target Pose, Overshoot and Hold again. The second project task (SPT) deals with character walk cycle which should be displayed in minimum five walking steps with some personal characteristics. The last task is Final project (FP) which shows the combined movement of the character in which the animation cycle is connected with linear animation. All movements must be shown as happening in the same sequence without interruptions. The aim of the FEA course is that through the four interrelated tasks the students pass through all basic variations of manlike character animation so that they gain the necessary experience before entering the professional production work.

The Eclipses Abacus, the mechanical predictor of the solar and lunar eclipses

This paper exposes and explains the synthesis of a mechanism, the Eclipses Abacus, by which the date of solar and lunar eclipses can be predicted and determined sufficiently accurately. Moreover, a three-dimensional (3D) model of this mechanical device is created by using Solid Works application and the simulation of its operation is accomplished by which the accuracy of its predictions can be demonstrated and inspected. This work is significant mainly for education in the field of 3D modeling, mechanism synthesis, and simulation. Moreover, the Eclipses Abacus is interesting for teaching courses in astronomy since it can demonstrate solar and lunar eclipses simply and clearly.

Semiautomatic Epicardial Fat Segmentation Based on Fuzzy c-Means Clustering and Geometric Ellipse Fitting

Automatic segmentation of particular heart parts plays an important role in recognition tasks, which is utilized for diagnosis and treatment. One particularly important application is segmentation of epicardial fat (surrounds the heart), which is shown by various studies to indicate risk level for developing various cardiovascular diseases as well as to predict progression of certain diseases. Quantification of epicardial fat from CT images requires advance image segmentation methods. The problem of the state-of-the-art methods for epicardial fat segmentation is their high dependency on user interaction, resulting in low reproducibility of studies and time-consuming analysis. We propose in this paper a novel semiautomatic approach for segmentation and quantification of epicardial fat from 3D CT images. Our method is a semisupervised slice-by-slice segmentation approach based on local adaptive morphology and fuzzy c-means clustering. Additionally, we use a geometric ellipse prior to filter out undesired parts of the target cluster. The validation of the proposed methodology shows good correspondence between the segmentation results and the manual segmentation performed by physicians.

Geometrical and mechanical characteristics of deformed balance spring obtained by simulation study

This paper analyzes the geometrical and mechanical characteristics of balance springs, which have the most important influence on the oscillation rate of the balance wheel and, consequently, the most decisive effects on the timekeeper’s accuracy. These characteristics are obtained by a set of SolidWorks simulations of static structural studies based on the finite element model, which are performed and accomplished in a specific and original way. Our study demonstrates that the position of the balance spring center of mass and its momentum of inertia are not constant but depend on the angular displacement of the spring collet end. It also shows that the torque spring rate (the spring stiffness) varies as a function of angular displacement during the balanced spring twist. All results are presented numerically and graphically by a set of diagrams and can contribute to the better understanding of the balance spring geometrical and mechanical features and, therefore, be applied for improving a timepiece’s accuracy. Moreover, the method of simulation study disclosed and explained in this paper provides general suggestions and guidance for similar studies. The accuracy of the presented results, as well as the validity of the complete simulation-based method, is proven by the experimentally determined relationship between balance spring restoring torque, stiffness, and spring angular displacement.

GMMs similarity measure based on LPP-like projection of the parameter space

We propose the novel more efficient similarity measure between GMMs.It is done by projecting GMMs from high dimensional to a lower dimensional space.GMMs distance is reduced to the distance between lower dimensional euclidian vectors.Greater discriminativity and lower computational cost is obtained.We confirm our results on artificial and real experimental data. The need for a comparison between two Gaussian Mixture Models (GMMs) plays a crucial role in various pattern recognition tasks and is involved as a key components in many expert and artificial intelligence (AI) systems dealing with real-life problems. As those system often operate on large data-sets and use high dimensional features, it is crucial for their recognition component to be computationally efficient in addition to its good recognition accuracy. In this work we deliver the novel similarity measure between GMMs, by LPP-like projecting the components of a particular GMM, from the high dimensional original parameter space, to a much lower dimensional space. Thus, finding the distance between two GMMs in the original space is reduced to finding the distance between sets of lower dimensional Euclidian vectors, pondered by corresponding weights. By doing so, we manage to obtain much better trade-off between the recognition accuracy and the computational complexity, in comparison to the measures between GMMs utilizing distances between Gaussian components evaluated in the original parameter space. Thus, the GMM measure that we propose is suitable for applications in AI systems that use GMMs in their recognition tasks and operate on large data sets, as the required number of overall Gaussian components involved in such systems is always large. We evaluate the proposed GMM measure on artificial, as well as real-world experimental data obtaining a much better trade-off between recognition accuracy and the computational complexity, in comparison to all baseline GMM similarity measures tested.

3D epicardial fat registration optimization based on structural prior knowledge and subjective-objective correspondence

3D heart registration has become an important issue in cardio-vascular diagnosis and treatment. This is mainly due to more accessible medical imaging technologies that can nowadays provide high precision imaging data at relatively lower cost. One of the important features of the heart that has recently drawn attention is epicardial fat (surrounds the heart), which according to some preliminary studies can indicate risk level of various cardiovascular diseases. As such, 2D/3D registration of epicardial fat, through automatic or semi-automatic detection/segmentation, is considered as valuable task for medical doctors (MDs) to include as additional feature within the already existing software for medical imaging and visualization. Although MDs can visually detect regions of epicardial fat from the image slices manually, i.e., subjectively, it is usually time consuming and error prone task. Moreover, due to considerable amount of parameters used for image pre-processing, which can strongly influence visibility of certain features in the image by MD, it often happens that some important features are missed. Consequently, the most preferable solution is the one that combines objective and subjective (by MD) description of particular image feature (in this example epicardial fat) and then subsequently employs semi-automatic segmentation approach, where in execution stage MD would only roughly indicate particular region of interest (ROI), based on which designed algorithm would process the whole heart volume and compute the 3D volume of the heart and epicardial fat. In this paper, we aim at optimizing and enhancing previously developed algorithm for 2D fat segmentation based on (i) pre-knowledge about epicardial structure (provided by the MDs) and (ii) subjective and objective metric correspondence. Based on the 2D segmentation method we compute the 3D volume in order to perform 3D registration. This new optimized approach is shown to considerably improve the accuracy of the epicardial fat registration using CT images.