In this paper, we propose a new algorithm to boost performance of traditional Linear Discriminant Analysis (LDA)-based face recogNtion (FR) methods in complex FR tasks, where highly nonlinear face pattern distributions are often encountered. The algorithm embodies the principle of "&vide and conquer", by whch a complex problem is decomposed into a set of simpler ones, each of which can be conquered by a relatively easy solution. The AdaBoost technique is utilized within this framework to: 1) generalize a set of simple FR sub-problems and their corresponding LDA solutions; 2) combine results from the multiple, relatively weak, LDA solutions to form a very strong solution. Experimentation performed on the FERET database indicates that the proposed methodology is able to greatly enhance performance of the Daditional LDA-based method with an averaged improvement of correct recognition rate (CRR) up to 9% reported. against the SSS problem, performing well even when L << J , which is the case in many FR tasks.Although successful in many cases, linear methods including the LDA-based ones oflen fail to deliver good performance when face patterns are subject to 1 q e variations in viewpoints, illumination or facial expression, which result in a highly nonlinear and complex distribution. The limited success of these methods should be attributed to their linear nature. There are two ways to handle the complex pattern distribution: 1 ) with nonlinear models, or 2) with a mixture of locally linear models (AMLLM). The main problem with most nonlinear methods such as those based on kernel maclunes is that the involved nonlinear parameters which significantly influence the performance of the FR systems, are veryThe authors would like to thank the FERET Technical Agent, Ihe US. National Institute of Standards and Technology (NISI) for providing the FERET datshase. difficult to he optimized. In addition, these methods are computationally expensive compared to their linear counterparts, and tend to overfit quite often. On the other hand, AMLLM-based approaches embody the principle of'divide and conquer", by which a complex FR problem is decomposed into a set of simpler ones, in each of which a locally linear face distribution can he generalized and dealt with by a relatively easy linear solution. As such, the AMLLM-based methods are simpler, more cost effective and easier to implement compared to the nonlinear ones.In this paper, we propose a new AMLLM-like method to boost the performance of the traditional LDA-based approaches in complex FR hks. The main novelty existing in the method is the introduction of the machine-learning technique known as "boosting", which is able to boost an ensemble of weak learners slightly hetter than random guessing to a very accurate leamer [3]. Boosting seems ideal to deal with two issues central to the AMLLM-like approaches: 1) the generalization of a set of simple linear solutions, each one aimed to a particular sub-problem; 2) the formation of a globally strong solution through the combi...
We present an infrared microscopy technique, capable of measuring the temperature of both the bounding surfaces and the oil film in an elastohydrodynamic contact. This technique can, for the first time, spatially resolve the oil film temperature in three dimensions. The contact is produced by loading a steel ball against a sapphire disc, and the film is viewed using an infrared microscope focussing through the disc. Two band pass filters are used to isolate the radiation from the oil film, and Planck's law is applied to data obtained at a known temperature as part of the calibration procedure. The proposed technique requires the emissivity of the oil film to be measured, which is acquired in situ and is shown to vary strongly as a function of thickness and temperature. The technique is validated under pure rolling conditions, when the temperature of the oil film is equal to the controlled lubricant reservoir temperature, and also compared to an equation commonly used to predict average film temperatures, confirming the value of the unknown constant. The technique is then used to gain insights into the thermal/rheological behaviour within a contact. This is important since the temperature of elastohydrodynamic contacts is critical in determining friction and hence the efficiency of machine components and this technique enables much needed validation and provides input data for CFD and numerical simulations.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.