Improved Tracking of Multiple Humans with Trajectory Prediction and Occlusion Modeling

A combined 2D, 3D approach is presented that allows for robust tracking of moving bodies in a given environment as observed via a single, uncalibrated video camera. Tracking is robust even in the presence of occlusions. Low-level features are often insufficient for detection, segmentation, and tracking of non-rigid moving objects. Therefore, an improved mechanism is proposed that combines low-level (image processing) and mid-level (recursive trajectory estimation) information obtained during the tracking process. The resulting system can segment and maintain the tracking of moving objects before, during, and after occlusion. At each frame, the system also extracts a stabilized coordinate frame of the moving objects. This stabilized frame is used to resize and resample the moving blob so that it can be used as input to motion recognition modules. The approach enables robust tracking without constraining the system to know the shape of the objects being tracked beforehand; although, some assumptions are made about the characteristics of the shape of the objects, and how they evolve with time. Experiments in tracking moving people are described.

Estimation and Prediction of Evolving Color Distributions for Skin Segmentation Under Varying Illumination

A novel approach for real-time skin segmentation in video sequences is described. The approach enables reliable skin segmentation despite wide variation in illumination during tracking. An explicit second order Markov model is used to predict evolution of the skin color (HSV) histogram over time. Histograms are dynamically updated based on feedback from the current segmentation and based on predictions of the Markov model. The evolution of the skin color distribution at each frame is parameterized by translation, scaling and rotation in color space. Consequent changes in geometric parameterization of the distribution are propagated by warping and re-sampling the histogram. The parameters of the discrete-time dynamic Markov model are estimated using Maximum Likelihood Estimation, and also evolve over time. Quantitative evaluation of the method was conducted on labeled ground-truth video sequences taken from popular movies.

A Neural Network Architecture for Autonomous Learning, Recognition, and Prediction in a Nonstationary World

In a constantly changing world, humans are adapted to alternate routinely between attending to familiar objects and testing hypotheses about novel ones. We can rapidly learn to recognize and narne novel objects without unselectively disrupting our memories of familiar ones. We can notice fine details that differentiate nearly identical objects and generalize across broad classes of dissimilar objects. This chapter describes a class of self-organizing neural network architectures--called ARTMAP-- that are capable of fast, yet stable, on-line recognition learning, hypothesis testing, and naming in response to an arbitrary stream of input patterns (Carpenter, Grossberg, Markuzon, Reynolds, and Rosen, 1992; Carpenter, Grossberg, and Reynolds, 1991). The intrinsic stability of ARTMAP allows the system to learn incrementally for an unlimited period of time. System stability properties can be traced to the structure of its learned memories, which encode clusters of attended features into its recognition categories, rather than slow averages of category inputs. The level of detail in the learned attentional focus is determined moment-by-moment, depending on predictive success: an error due to over-generalization automatically focuses attention on additional input details enough of which are learned in a new recognition category so that the predictive error will not be repeated. An ARTMAP system creates an evolving map between a variable number of learned categories that compress one feature space (e.g., visual features) to learned categories of another feature space (e.g., auditory features). Input vectors can be either binary or analog. Computational properties of the networks enable them to perform significantly better in benchmark studies than alternative machine learning, genetic algorithm, or neural network models. Some of the critical problems that challenge and constrain any such autonomous learning system will next be illustrated. Design principles that work together to solve these problems are then outlined. These principles are realized in the ARTMAP architecture, which is specified as an algorithm. Finally, ARTMAP dynamics are illustrated by means of a series of benchmark simulations.

ARTMAP: Supervised Real-Time Learning and Classification of Nonstationary Data by a Self-Organizing Neural Network

This article introduces a new neural network architecture, called ARTMAP, that autonomously learns to classify arbitrarily many, arbitrarily ordered vectors into recognition categories based on predictive success. This supervised learning system is built up from a pair of Adaptive Resonance Theory modules (ARTa and ARTb) that are capable of self-organizing stable recognition categories in response to arbitrary sequences of input patterns. During training trials, the ARTa module receives a stream {a^(p)} of input patterns, and ARTb receives a stream {b^(p)} of input patterns, where b^(p) is the correct prediction given a^(p). These ART modules are linked by an associative learning network and an internal controller that ensures autonomous system operation in real time. During test trials, the remaining patterns a^(p) are presented without b^(p), and their predictions at ARTb are compared with b^(p). Tested on a benchmark machine learning database in both on-line and off-line simulations, the ARTMAP system learns orders of magnitude more quickly, efficiently, and accurately than alternative algorithms, and achieves 100% accuracy after training on less than half the input patterns in the database. It achieves these properties by using an internal controller that conjointly maximizes predictive generalization and minimizes predictive error by linking predictive success to category size on a trial-by-trial basis, using only local operations. This computation increases the vigilance parameter ρa of ARTa by the minimal amount needed to correct a predictive error at ARTb· Parameter ρa calibrates the minimum confidence that ARTa must have in a category, or hypothesis, activated by an input a^(p) in order for ARTa to accept that category, rather than search for a better one through an automatically controlled process of hypothesis testing. Parameter ρa is compared with the degree of match between a^(p) and the top-down learned expectation, or prototype, that is read-out subsequent to activation of an ARTa category. Search occurs if the degree of match is less than ρa. ARTMAP is hereby a type of self-organizing expert system that calibrates the selectivity of its hypotheses based upon predictive success. As a result, rare but important events can be quickly and sharply distinguished even if they are similar to frequent events with different consequences. Between input trials ρa relaxes to a baseline vigilance pa When ρa is large, the system runs in a conservative mode, wherein predictions are made only if the system is confident of the outcome. Very few false-alarm errors then occur at any stage of learning, yet the system reaches asymptote with no loss of speed. Because ARTMAP learning is self stabilizing, it can continue learning one or more databases, without degrading its corpus of memories, until its full memory capacity is utilized.

Prediction and prevention of venous thrombosis in pregnancy

Venous thromboembolism (VTE) remains the leading cause of maternal mortality. Reports identified further research is required in obese and women post caesarean section (CS). Risk factors for VTE during pregnancy are periodically absent indicating the need for a simple and effective screening tool for pregnancy. Perturbation of the uteroplacental haemostasis has been implicated in placenta mediated pregnancy complications. This thesis had 4 main aims: 1) To investigate anticoagulant effects following a fixed thromboprophylaxis dose in healthy women post elective CS. 2) To evaluate the calibrated automated thrombogram (CAT) assay as a potential predictive tool for thrombosis in pregnancy. 3) To compare the anticoagulant effects of fixed versus weight adjusted thromboprophylaxis dose in morbidly obese pregnant women. 4) To investigate the LMWH effects on human haemostatic gene and antigen expression in placentae and plasma from the uteroplacental , maternal and fetal circulation. Tissue factor pathway inhibitor (TFPI), thrombin antithrombin (TAT), CAT and anti-Xa levels were analysed. Real-time PCR and ELISA were used to quantify mRNA and protein expression of TFPI and TF in placental tissue. In women post CS, anti-Xa levels do not reflect the full anticoagulant effects of LMWH. LMWH thromboprophylaxis in this healthy cohort of patients appears to have a sustained effect in reducing excess thrombin production post elective CS. The results of this study suggest that predicting VTE in pregnant women using CAT assay is not possible at present time. The prothrombotic state in pregnant morbidly obese women was substantially attenuated by weight adjusted but not at fixed LMWH doses. LMWH may be effective in reducing in- vivo thrombin production in the uteroplacental circulation of thrombophilic women. All these results collectively suggest that at appropriate dosage, LMWH is effective in attenuating excess thrombin generation, in low risk pregnant women post caesarean section or moderate to high risk pregnant women who are morbidly obese or tested positive for thrombophilia. The results of the studies provided data to inform evidence-based practice to improve the outcome for pregnant women at risk of thrombosis.

The Exodus evacuation model applied to building evacuation scenarios

The purpose of this paper is to demonstrate the potential of the EXODUS evacuation model in building environments. The latest PC/workstation version of EXODUS is described and is also applied to a large hypothetical supermarket/restaurant complex measuring 50 m x 40 m. A range of scenarios is presented where population characteristics (such as size, individual travel speeds, and individual response times), and enclosure configuration characteristics (such as number of exits, size of exits, and opening times of exits) are varied. The results demonstrate a wide range of occupant behavior including overtaking, queuing, redirection, and conflict avoidance. Evacuation performance is measured by a number of model predicted parameters including individual exit flow rates, overall evacuation flow rates, total evacuation time, average evacuation time per occupant, average travel distance, and average wait time. The simulations highlight the profound impact that variations in individual travel speeds and occupant response times have in determining the overall evacuation performance. 1. Jin, T., and Yamada T., "Experimental Study of Human Behavior in Smoke Filled Corridors," Proceedings of The Second International Symposium on Fire Safety Science, 1988, pp. 511-519. 2. Galea, E.R., and Galparsoro, J.M.P., "EXODUS: An Evacuation Model for Mass Transport Vehicles," UK CAA Paper 93006 ISBN 086039 543X, CAA London, 1993. 3. Galea, E.R., and Galparsoro, J.M.P., "A Computer Based Simulation Model for the Prediction of Evacuation from Mass Transport Vehicles," Fire Safety Journal, Vol. 22, 1994, pp. 341-366. 4. Galea, E.R., Owen, M., and Lawrence, P., "Computer Modeling of Human Be havior in Aircraft Fire Accidents," to appear in the Proceedings of Combus tion Toxicology Symposium, CAMI, Oklahoma City, OK, 1995. 5. Kisko, T.M. and Francis, R.L., "EVACNET+: A Computer Program to Determine Optimal Building Evacuation Plans," Fire Safety Journal, Vol. 9, 1985, pp. 211-220. 6. Levin, B., "EXITT, A Simulation Model of Occupant Decisions and Actions in Residential Fires," Proceedings of The Second International Symposium on Fire Safety Science, 1988, pp. 561-570. 7. Fahy, R.F., "EXIT89: An Evacuation Model for High-Rise Buildings," Pro ceedings of The Third International Sym posium on Fire Safety Science, 1991, pp. 815-823. 8. Thompson, P.A., and Marchant, E.W., "A Computer Model for the Evacuation of Large Building Populations," Fire Safety Journal, Vol. 24, 1995, pp. 131-148. 9. Still, K., "New Computer System Can Predict Human Behavior Response to Building Fires," FIRE 84, 1993, pp. 40-41. 10. Ketchell, N., Cole, S.S., Webber, D.M., et.al., "The Egress Code for Human Move ment and Behavior in Emergency Evacu ations," Engineering for Crowd Safety (Smith, R.A., and Dickie, J.F., Eds.), Elsevier, 1993, pp. 361-370. 11. Takahashi, K., Tanaka, T. and Kose, S., "An Evacuation Model for Use in Fire Safety Design of Buildings," Proceedings of The Second International Symposium on Fire Safety Science, 1988, pp. 551- 560. 12. G2 Reference Manual, Version 3.0, Gensym Corporation, Cambridge, MA. 13. XVT Reference Manual, Version 3.0 XVT Software Inc., Boulder, CO. 14. Galea, E.R., "On the Field Modeling Approach to the Simulation of Enclosure Fires, Journal of Fire Protection Engineering, Vol. 1, No. 1, 1989, pp. 11-22. 15. Purser, D.A., "Toxicity Assessment of Combustion Products," SFPE Handbook of Fire Protection Engineering, National Fire Protection Association, Quincy, MA, pp. 1-200 - 1-245, 1988. 16. Hankin, B.D., and Wright, R.A., "Pas senger Flows in Subways," Operational Research Quarterly, Vol. 9, 1958, pp. 81-88. 17. HMSO, The Building Regulations 1991 - Approved Document B, section B 1 (1992 edition), HMSO publications, London, pp. 9-40. 18. Polus A., Schofer, J.L., and Ushpiz, A., "Pedestrian Flow and Level of Service," Journal of Transportation Engineering, Vol. 109, 1983, pp. 46-47. 19. Muir, H., Marrison, C., and Evans, A., "Aircraft Evacuations: the Effect of Passenger Motivation and Cabin Con figuration Adjacent to the Exit," CAA Paper 89019, ISBN 0 86039 406 9, 1989. 20. Muir, H., Private communication to appear as a CAA report, 1996.

The role of evacuation modelling in the development of safer air travel

Computer based mathematical models describing the aircraft evacuation process have a vital role to play in the design and development of safer aircraft, in the implementation of safer and more rigorous certification criteria and in post mortuuum accident investigation. As the risk of personal injury and costs involved in performing large-scale evacuation experiments for the next generation 'Ultra High Capacity Aircraft' (UHCA) are expected to be high, the development and use of these evacuation modelling tools may become essential if these aircraft are to prove a viable reality. In this paper the capabilities and limitation of the air-EXODUS evacuation model are described. Its successful application to the prediction of a recent certificaiton trial, prior to the actual trial taking place, is described. Also described is a newly defined parameter known as OPS which can be used as a measure of evacuation trial optimality. Finally, the data requirements of aircraft evacuation models is discussed along with several projects currently underway at the University of Greenwich designed to obtain this data. Included in this discussion is a description of the AASK - Aircraft Accident Statistics and Knowledge - data base which contains detailed information from aircraft accident survivors.

Investigating the impact of occupant response time on computer simulations of the WTC North Tower evacuation

This work explores the impact of response time distributions on high-rise building evacuation. The analysis utilises response times extracted from printed accounts and interviews of evacuees from the WTC North Tower evacuation of 11 September 2001. Evacuation simulations produced using these “real” response time distributions are compared with simulations produced using instant and engineering response time distributions. Results suggest that while typical engineering approximations to the response time distribution may produce reasonable evacuation times for up to 90% of the building population, using this approach may underestimate total evacuation times by as much as 61%. These observations are applicable to situations involving large high-rise buildings in which travel times are generally expected to be greater than response times

Prediction of reinforcement corrosion in concrete and its effects on concrete cracking and strength reduction

Based on extensive research on reinforcing steel corrosion in concrete in the past decades, it is now possible to estimate the effect of the progression of reinforcement corrosion in concrete infrastructure on its structural performance. There are still areas of considerable uncertainty in the models and in the data available, however This paper uses a recently developed model for reinforcement corrosion in concrete to improve the estimation process and to indicate the practical implications. In particular stochastic models are used to estimate the time likely to elapse for each phase of the whole corrosion process: initiation, corrosion-induced concrete cracking, and structural strength reduction. It was found that, for practical flexural structures subject to chloride attacks, corrosion initiation may start quite early in their service life. It was also found that, once the structure is considered to be unserviceable due to corrosion-induced cracking, there is considerable remaining service life before the structure can be considered to have become unsafe. The procedure proposed in the paper has the potential to serve as a rational tool for practitioners, operators, and asset managers to make decisions about the optimal timing of repairs, strengthening, and/or rehabilitation of corrosion-affected concrete infrastructure. Timely intervention has the potential to prolong the service life of infrastructure.

Modelling and Prediction of NOx Emission in a Coal-Fired Power Generation Plant

In this paper NOx emissions modelling for real-time operation and control of a 200 MWe coal-fired power generation plant is studied. Three model types are compared. For the first model the fundamentals governing the NOx formation mechanisms and a system identification technique are used to develop a grey-box model. Then a linear AutoRegressive model with eXogenous inputs (ARX) model and a non-linear ARX model (NARX) are built. Operation plant data is used for modelling and validation. Model cross-validation tests show that the developed grey-box model is able to consistently produce better overall long-term prediction performance than the other two models.

Prediction of adipose tissue composition using Raman spectroscopy: average properties and individual fatty acids.

Abstract: Raman spectroscopy has been used for the first time to predict the FA composition of unextracted adipose tissue of pork, beef, lamb, and chicken. It was found that the bulk unsaturation parameters could be predicted successfully [R-2 = 0.97, root mean square error of prediction (RMSEP) = 4.6% of 4 sigma], with cis unsaturation, which accounted for the majority of the unsaturation, giving similar correlations. The combined abundance of all measured PUFA (>= 2 double bonds per chain) was also well predicted with R-2 = 0.97 and RMSEP = 4.0% of 4 sigma. Trans unsaturation was not as well modeled (R-2 = 0.52, RMSEP = 18% of 4 sigma); this reduced prediction ability can be attributed to the low levels of trans FA found in adipose tissue (0.035 times the cis unsaturation level). For the individual FA, the average partial least squares (PLS) regression coefficient of the 18 most abundant FA (relative abundances ranging from 0.1 to 38.6% of the total FA content) was R-2 = 0.73; the average RMSEP = 11.9% of 4 sigma. Regression coefficients and prediction errors for the five most abundant FA were all better than the average value (in some cases as low as RMSEP = 4.7% of 4 sigma). Cross-correlation between the abundances of the minor FA and more abundant acids could be determined by principal component analysis methods, and the resulting groups of correlated compounds were also well-predicted using PLS. The accuracy of the prediction of individual FA was at least as good as other spectroscopic methods, and the extremely straightforward sampling method meant that very rapid analysis of samples at ambient temperature was easily achieved. This work shows that Raman profiling of hundreds of samples per day is easily achievable with an automated sampling system.

Time domain wave separation using multiple microphones

Methods of measuring the acoustic behavior of tubular systems can be broadly characterized as steady state measurements, where the measured signals are analyzed in terms of infinite duration sinusoids, and reflectometry measurements which exploit causality to separate the forward and backward going waves in a duct. This paper sets out a multiple microphone reflectometry technique which performs wave separation by using time domain convolution to track the forward and backward going waves in a cylindrical source tube. The current work uses two calibration runs (one for forward going waves and one for backward going waves) to measure the time domain transfer functions for each pair of microphones. These time domain transfer functions encode the time delay, frequency dependent losses and microphone gain ratios for travel between microphones. This approach is applied to the measurement of wave separation, bore profile and input impedance. The work differs from existing frequency domain methods in that it combines the information of multiple microphones within a time domain algorithm, and differs from existing time domain methods in its inclusion of the effect of losses and gain ratios in intermicrophone transfer functions.

In Silico Prediction of the Mechanobiological Response of Arterial Tissue: Application to Angioplasty and Stenting

One way to restore physiological blood flow to occluded arteries involves the deformation of plaque using an intravascular balloon and preventing elastic recoil using a stent. Angioplasty and stent implantation cause unphysiological loading of the arterial tissue, which may lead to tissue in-growth and reblockage; termed “restenosis.” In this paper, a computational methodology for predicting the time-course of restenosis is presented. Stress-induced damage, computed using a remaining life approach, stimulates inflammation (production of matrix degrading factors and growth stimuli). This, in turn, induces a change in smooth muscle cell phenotype from contractile (as exists in the quiescent tissue) to synthetic (as exists in the growing tissue). In this paper, smooth muscle cell activity (migration, proliferation, and differentiation) is simulated in a lattice using a stochastic approach to model individual cell activity. The inflammation equations are examined under simplified loading cases. The mechanobiological parameters of the model were estimated by calibrating the model response to the results of a balloon angioplasty study in humans. The simulation method was then used to simulate restenosis in a two dimensional model of a stented artery. Cell activity predictions were similar to those observed during neointimal hyperplasia, culminating in the growth of restenosis. Similar to experiment, the amount of neointima produced increased with the degree of expansion of the stent, and this relationship was found to be highly dependant on the prescribed inflammatory response. It was found that the duration of inflammation affected the amount of restenosis produced, and that this effect was most pronounced with large stent expansions. In conclusion, the paper shows that the arterial tissue response to mechanical stimulation can be predicted using a stochastic cell modeling approach, and that the simulation captures features of restenosis development observed with real stents. The modeling approach is proposed for application in three dimensional models of cardiovascular stenting procedures.