Ionut Florescu

An Adaptive Background Model for Camshift Tracking with a Moving Camera

Continuously Adaptive Mean shift (CAMSHIFT) is a popular algorithm for visual tracking, providing speed and robustness with minimal training and computational cost. While it performs well with a fixed camera and static background scene, it can fail rapidly when the camera moves or the background changes since it relies on static models of both the background and the tracked object. Furthermore it is unable to track objects passing in front of backgrounds with which they share significant colours. We describe a new algorithm, the Adaptive Background CAMSHIFT (ABCshift), which addresses both of these problems by using a background model which can be continuously relearned for every frame with minimal additional computational expense. Further, we show how adaptive background relearning can occasionally lead to a particular mode of instability which we resolve by comparing background and tracked object distributions using a metric based on the Bhattacharyya coefficient.

Efficient visual servoing with the ABCshift tracking algorithm

Visual tracking algorithms have important robotic applications such as mobile robot guidance and servoed wide area surveillance systems. These applications ideally require vision algorithms which are robust to camera motion and scene change but are cheap and fast enough to run on small, low power embedded systems. Unfortunately most robust visual tracking algorithms are either computationally expensive or are restricted to a stationary camera. This paper describes a new color based tracking algorithm, the Adaptive Background CAMSHIFT (ABCshift) tracker and an associated technique, mean shift servoing, for efficient pan-tilt servoing of a motorized camera platform. ABCshift achieves robustness against camera motion and other scene changes by continuously relearning its background model at every frame. This also enables robustness in difficult scenes where the tracked object moves past backgrounds with which it shares significant colors. Despite this continuous machine learning, ABCshift needs minimal training and is remarkably computationally cheap. We first demonstrate how ABCshift tracks robustly in situations where related algorithms fail, and then show how it can be used for real time tracking with pan-tilt servo control using only a small embedded microcontroller.

Bayesian fusion of thermal and visible spectra camera data for region based tracking with rapid background adaptation

This paper presents a method for optimally combining pixel information from an infra-red thermal imaging camera, and a conventional visible spectrum colour camera, for tracking a moving target. The tracking algorithm rapidly re-learns its background models for each camera modality from scratch at every frame. This enables, firstly, automatic adjustment of the relative importance of thermal and visible information in decision making, and, secondly, a degree of “camouflage target” tracking by continuously re-weighting the importance of those parts of the target model that are most distinct from the present background at each frame. Furthermore, this very rapid background adaptation ensures robustness to large, sudden and arbitrary camera motion, and thus makes this method a useful tool for robotics, for example visual servoing of a pan-tilt turret mounted on a moving robot vehicle. The method can be used to track any kind of arbitrarily shaped or deforming object, however the combination of thermal and visible information proves particularly useful for enabling robots to track people. The method is also important in that it can be readily extended for data fusion of an arbitrary number of statistically independent features from one or arbitrarily many imaging modalities.

An Appraisal of the Classic Forest Succession Paradigm with the Shade Tolerance Index

In this paper we revisit the classic theory of forest succession that relates shade tolerance and species replacement and assess its validity to understand patch-mosaic patterns of forested ecosystems of the USA. We introduce a macroscopic parameter called the “shade tolerance index” and compare it to the classic continuum index in southern Wisconsin forests. We exemplify shade tolerance driven succession in White Pine-Eastern Hemlock forests using computer simulations and analyzing approximated chronosequence data from the USDA FIA forest inventory. We describe this parameter across the last 50 years in the ecoregions of mainland USA, and demonstrate that it does not correlate with the usual macroscopic characteristics of stand age, biomass, basal area, and biodiversity measures. We characterize the dynamics of shade tolerance index using transition matrices and delimit geographical areas based on the relevance of shade tolerance to explain forest succession. We conclude that shade tolerance driven succession is linked to climatic variables and can be considered as a primary driving factor of forest dynamics mostly in central-north and northeastern areas in the USA. Overall, the shade tolerance index constitutes a new quantitative approach that can be used to understand and predict succession of forested ecosystems and biogeographic patterns.

Probability of Detection and Optimal Sensor Placement for Threshold Based Detection Systems

This paper provides a probabilistic analysis of simple detection systems which are based on thresholding feature values extracted from a sensor signal. For such systems, this paper explains how to calculate the probability of detection as a function of range from the sensor to the object of interest. This function is important in that it enables optimal positioning of a group of sensors, either maximizing detection rates for a given number of sensors or informing the minimum number of sensors necessary to achieve a desired probability of detection throughout an area. An example case study is presented, based on a novel approach to passive acoustic diver detection in noisy environments.

Modelling of forest stand dynamics using Markov chains

Understanding forest complexity and self-organization across multiple scales is essential for both ecology and natural resource management. In this paper, we develop a Markov chain approach for the modelling of forest stand dynamics. The aim of this work is to generalize the recently developed Perfect Plasticity Approximation (PPA) model for scaling of vegetation dynamics from individual level to the landscape level through the ecosystem hierarchical structure. Our basic assumption is that the forested ecosystem and disturbance regimes can be modelled on 3 hierarchical scales (levels): individual trees, forest stand (or patch, defined as a spatial unit about 0.5-1 ha of the same forest at one successional stage.) and landscape (collection of forest patches of different forest/soil types at different successional stages) levels. In our modelling approach the PPA model is an intermediate step for scaling from the individual level to the forest stand level (or patch level). In this paper we develop a Markov chain model for stage-structured dynamics of forest stands (patches). In order to determine the structure of the Markov chain model and estimate parameters, we analyze the patch-mosaic patterns of forest stands of the Lake States (MI, WI, and MN) recorded in the USDA FIA database as well as data for other US states and Canada. The distribution of macroscopic characteristics of a large collection of forest patches is considered as an estimate of the stationary distribution of the underlying Markov chain. The data demonstrates that this distribution is unimodal and skewed to the right. We identify the simplest Markov chain that produces such a distribution and estimate the upper bound of the probability of disaster for this Markov chain.

NUMERICAL SCHEMES FOR OPTION PRICING IN REGIME-SWITCHING JUMP DIFFUSION MODELS

In this paper, we present algorithms to solve a complex system of partial integro-differential equations (PIDEs) of parabolic type. The system is motivated by applications in finance where the solution of the system gives the price of European options in a regime-switching jump diffusion model. The new algorithms are based on theoretical analysis in Florescu et al. (2012) where the proof of convergence of the algorithms is carried out. The problems are also solved using a more traditional approach, where the integral terms (but not the derivative terms) are treated explicitly. Another contribution of this work details a novel type of jump distribution. Empirical evidence suggests that this type of distribution may be more appropriate to model jumps as it makes them more clearly distinguishable from the signal variability.

Bayesian fusion of thermal and visible spectra camera data for mean shift tracking with rapid background adaptation

This paper presents a method for optimally combining pixel information from thermal imaging and visible spectrum colour cameras, for tracking an arbitrarily shaped deformable moving target. The tracking algorithm rapidly re-learns its background models for each camera modality from scratch at every frame. This enables, firstly, automatic adjustment of the relative importance of thermal and visible information in decision making, and, secondly, a degree of “camouflage target” tracking by continuously re-weighting the importance of those parts of the target model that are most distinct from the present background at each frame. Furthermore, this very rapid background adaptation ensures robustness to rapid camera motion. The combination of thermal and visible information is applicable to any target, but particularly useful for people tracking. The method is also important in that it can be readily extended for fusion of data from arbitrarily many imaging modalities.

Tools for Change: An Examination of Transformative Learning and Its Precursor Steps in Undergraduate Students

In this quantitative study of college students spanning three waves, the 10 theoretical precursor steps of transformational learning did predict its occurrence. The most consistent predictor was the step of reflection. Maturity and ethnicity also showed a predictive value, but college major was not a significant differentiator for transformative learning.

Regulating noisy short-selling of troubled firms?

Purpose - The purpose of this paper is to examine the efficacy of recent policy initiatives taken by the US Securities and Exchange Commission banning naked “short-selling” of specific financial stocks. The paper also considers the merits of reinstating “uptick rule” 10a-1, which prohibits short-selling securities on a downtick. Design/methodology/approach - The paper studies theoretical implications of short-selling in a simple state-claim model, reflecting varying amounts of short interest in a representative firm and noise trading in the market. Price discovery depends on the proportion of noise trading compared to rational short-selling. The empirical analysis focuses on price volatility under short-selling constraints employing simple regressions, EGARCH analysis and simulated price behavior under a hypothetical uptick rule. Findings - The EGARCH results suggest short-selling constraints had non-uniform impacts on the persistence and leverage effects associated with price volatility. The corresponding price simulations indicate a hypothetical uptick rule might have helped stabilize price behavior in some cases, depending on the nature of the stochastic process and whether or not quantity constraints on short-selling are binding. Originality/value - The theoretical arguments and empirical findings suggest a “focused approach” to market regulation would be a more efficient means of discouraging trend chasing without compromising “informed trading” – that is to say, safeguarding price discovery and market liquidity without impeding arbitrage or confounding probability beliefs regarding firm survival. These conclusions are largely in accord with recent policy analysis and proposals outlined in Avgouleas.