Abstract

This paper discusses an approach to human action recognition via local feature tracking and robust estimation of background motion. The main contribution is a robust feature extraction algorithm based on KLT tracker and SIFT as well as a method for estimating dominant planes in the scene. Multiple interest point detectors are used to provide large number of features for every frame. The motion vectors for the features are estimated using optical flow and SIFT based matching. The features are combined with image segmentation to estimate dominant homographies, and then separated into static and moving ones regardless the camera motion. The action recognition approach can handle camera motion, zoom, human appearance variations, background clutter and occlusion. The motion compensation shows very good accuracy on a number of test sequences. The recognition system is extensively compared to state-of-the art action recognition methods and the results are improved.

No abstract

For a Three-Level Three-phase T-type (3LTTC) rectifier and inverter of a high efficiency Uninterruptible Power Supply (UPS) with an output power of 20 kVA, most suitable semiconductor components are selected. For this purpose, this paper details conduction and switching loss models of T-type rectifiers and inverters, compares the total semiconductor losses achieved for RB-IGBTs and for different types of conventional IGBTs, and evaluates the improvements achieved if the Si rectifier diodes are replaced by SiC Schottky Barrier Diodes (SiC SBDs). The switching loss model is parameterized with measured switching losses. According to the results of this comparison, the rectifier preferably employs RB-IGBTs to realize the bi-directional switch and SiC SBDs for the rectifier diodes; switching frequencies up to 32.5 kHz are feasible for total semiconductor losses of the rectifier of 250 W. For the inverter, a realization of the bi-directional switch using an antiseries connection of conventional IGBT / SiC SBD modules is found to be most suitable and facilitates a switching frequency of 19.7 kHz for maximum allowed losses of 250 W.

No abstract

No abstract

In this paper, an optimization regarding volume, efficiency and costs of AC boost inductors in 3-phase PWM converters based on detailed multi-domain models is presented. The optimization is performed for a wide switching frequency range of 5-80kHz and a wide current ripple range of 5-100 %, considering ferrite, amorphous and powder core materials in combination with round, Iitz, foil and flat wire windings. The shown analysis and optimization identifies the best core mate riaVwinding type combinations for both thermally and efficiency constrained inductor designs. Furthermore, the investigations reveal that simplified scaling assumptions, e.g. a proportional relationship between the inductor volume and the inverse of the frequency or the stored energy, are only accurate in special cases.

This paper describes a robust color tracker employing an updateable two-dimensional color histogram with an anisotropic asymmetric Gaussian distribution model. A robust color tracking is achieved by updating the color model employing two-dimensional histogram of the h-s plane of the hsv color space. Performance of the proposed technique is robust to unknown illumination change, because the color model is acquired directly from the object and updated iteratively. Moreover, an anisotropic asymmetric Gaussian distribution model is adopted as a model of changeable color range of the object. The color model is added and deleted one by one in the update process. Not very different color is added and quite different color is discarded by estimating the changeable color range of the object. A color tracking system with a pan-tilt camera was developed and, using the system, performance of the proposed technique was examined in a various illumination conditions. The system achieved robust tracking of specified color in the experiment.