Costas Panagiotakis

Equivalent Key Frames Selection Based on Iso-Content Principles

We present a key frames selection algorithm based on three iso-content principles (iso-content distance, iso-content error and iso-content distortion), so that the selected key frames are equidistant in video content according to the used principle. Two automatic approaches for defining the most appropriate number of key frames are proposed by exploiting supervised and unsupervised content criteria. Experimental results and the comparisons with existing methods from literature on large dataset of real-life video sequences illustrate the high performance of the proposed schemata.

Segmentation and Sampling of Moving Object Trajectories Based on Representativeness

Moving Object Databases (MOD), although ubiquitous, still call for methods that will be able to understand, search, analyze, and browse their spatiotemporal content. In this paper, we propose a method for trajectory segmentation and sampling based on the representativeness of the (sub)trajectories in the MOD. In order to find the most representative subtrajectories, the following methodology is proposed. First, a novel global voting algorithm is performed, based on local density and trajectory similarity information. This method is applied for each segment of the trajectory, forming a local trajectory descriptor that represents line segment representativeness. The sequence of this descriptor over a trajectory gives the voting signal of the trajectory, where high values correspond to the most representative parts. Then, a novel segmentation algorithm is applied on this signal that automatically estimates the number of partitions and the partition borders, identifying homogenous partitions concerning their representativeness. Finally, a sampling method over the resulting segments yields the most representative subtrajectories in the MOD. Our experimental results in synthetic and real MOD verify the effectiveness of the proposed scheme, also in comparison with other sampling techniques.

Interactive Image Segmentation Based on Synthetic Graph Coordinates

In this paper, we propose a framework for interactive image segmentation. The goal of interactive image segmentation is to classify the image pixels into foreground and background classes, when some foreground and background markers are given. The proposed method minimizes a min-max Bayesian criterion that has been successfully used on image segmentation problem and it consists of several steps in order to take into account visual information as well as the given markers, without any requirement of training. First, we partition the image into contiguous and perceptually similar regions (superpixels). Then, we construct a weighted graph that represents the superpixels and the connections between them. An efficient algorithm for graph clustering based on synthetic coordinates is used yielding an initial map of classified pixels. This method reduces the problem of graph clustering to the simpler problem of point clustering, instead of solving the problem on the graph data structure, as most of the known algorithms from literature do. Finally, having available the data modeling and the initial map of classified pixels, we use a Markov Random Field (MRF) model or a flooding algorithm to get the image segmentation by minimizing a min-max Bayesian criterion. Experimental results and comparisons with other methods from the literature are presented on LHI, Gulshan and Zhao datasets, demonstrating the high performance and accuracy of the proposed scheme.

Unsupervised trajectory sampling

A novel methodology for efficiently sampling Trajectory Databases (TD) for mobility data mining purposes is presented. In particular, a three-step unsupervised trajectory sampling methodology is proposed, that initially adopts a symbolic vector representation of a trajectory which, using a similarity-based voting technique, is transformed to a continuous function that describes the representativeness of the trajectory in the TD. This vector representation is then relaxed by a merging algorithm, which identifies the maximal representative portions of each trajectory, at the same time preserving the space-time mobility pattern of the trajectory. Finally, a novel sampling algorithm operating on the previous representation is proposed, allowing us to select a subset of a TD in an unsupervised way encapsulating the behavior (in terms of mobility patterns) of the original TD. An experimental evaluation over synthetic and real TD demonstrates the efficiency and effectiveness of our approach.

Automatic

In this paper, we propose a method for the automatic identification of P-phase arrival based on the distribution of local maxima (LM) in earthquake seismograms. The method efficiently combines energy and frequency characteristics of the LM distribution (LMD). The detection is mainly based on the energy of a seismic event in the case the earthquake has higher amplitude than seismic background noise. Otherwise, it is based on the frequency of LM. Thus, the method provides robust detection of P-phase arrival in any quality type of seismic data. Moreover, it uses two sequential sliding signal windows yielding very high accuracy on the P-phase estimation. A hierarchical P-phase detection algorithm dramatically reduces the computational cost, making possible a real-time implementation. Experimental results from a large database of more than 80 low, medium, and high signal-to-noise ratio seismic events and comparison with existing methods in the literature indicate the reliable performance of the proposed scheme.

P

In this paper, we propose an efficient clustering algorithm that has been applied to the microaggregation problem. The goal is to partition N given records into clusters, each of them grouping at least K records, so that the sum of the within-partition squared error (SSE) is minimized. We propose a successive Group Selection algorithm that approximately solves the microaggregation problem in O(N2 log N) time, based on sequential Minimization of SSE. Experimental results and comparisons to existing methods with similar computation cost on real and synthetic data sets demonstrate the high performance and robustness of the proposed scheme.

-Phase Picking Based on Local-Maxima Distribution

We present a method to solve the human silhouette tracking problem using 18 major human points. We used: a simple 2D model for the human silhouette, a linear prediction technique for initializing major points search, geometry anthropometric constraints for determining the search area and color measures for matching human body parts. In addition, we propose a method to solve the problem of human members recognition and 18 major human points detection using silhouette. This result can be used to initialize a human tracking algorithm for real time applications. Our main purpose is to develop a low computation cost algorithm, which can be used independently of camera motion. The output of the tracking algorithm is the position of 18 major human points and a 2D human body extraction. In cases of low quality imaging conditions or low background contrast, the result may be worst. For these cases we defined an appropriate criterion concerning tracking ability.

Successive Group Selection for Microaggregation

An automatic human shape-motion analysis method based on a fusion architecture is proposed for human action recognition in videos. Robust shape-motion features are extracted from human points detection and tracking. The features are combined within the Transferable Belief Model (TBM) framework for action recognition. The TBM-based modelling and fusion process allows to take into account imprecision, uncertainty and conflict inherent to the features. Action recognition is performed by a multilevel analysis. The sequencing is exploited for feedback information extraction in order to improve tracking results. The system is tested on real videos of athletics meetings to recognize four types of jumps: high jump, pole vault, triple jump and long jump

Recognition and Tracking of the Members of a Moving Human Body

We present new mathematical and geological models to assist civil protection authorities in the mitigation of potential oil spill accidents in the Eastern Mediterranean Sea. Oil spill simulations for 19 existing offshore wells were carried out based on novel and high resolution bathymetric, meteorological, oceanographic, and geomorphological data. The simulations show a trend for east and northeast movement of oil spills into the Levantine Basin, affecting the coastal areas of Israel, Lebanon and Syria. Oil slicks will reach the coast in 1 to 20 days, driven by the action of the winds, currents and waves. By applying a qualitative analysis, seabed morphology is for the first time related to the direction of the oil slick expansion, as it is able to alter the movement of sea currents. Specifically, the direction of the major axis of the oil spills, in most of the cases examined, is oriented according to the prevailing azimuth of bathymetric features. This work suggests that oil spills in the Eastern Mediterranean Sea should be mitigated in the very few hours after their onset, and before wind and currents disperse them. We explain that protocols should be prioritized between neighboring countries to mitigate any oil spills.

Shape-Motion based athlete tracking for multilevel action recognition

In this paper, we propose a method for automatic enhancement and identification of partial curvilinear structures. The accurate identification of line structures in geophysical images plays an important role in geophysical interpretation and the detection of subsurface structures. The method was applied on geophysical images in an effort to recognize the linear patterns of subsurface architectural structures that exist in archaeological sites. To our knowledge, the problem of identification of curvilinear structures in geophysical images for archaeological sites is faced for the first time. The method efficiently combines a rotation- and scale-invariant filter and a pixel-labeling method, providing a robust enhancement and detection of mostly line structures in 2-D grayscale images, respectively. Experimental results on real and synthetic images and comparison with existing methods in the literature indicated the reliable performance of the proposed scheme.