Login

Annoucements

The WRM Concept from High Energy Physics to Artificial Vision

06 Apr 2016

Lebanese University, Faculty of Science, Rafic Hariri University Campus – Hadath - Lebanon

Monday, April 11th, 2016, h. 14.30

Conference room (Physics Department)

Exploring the frontiers of High Energy Physics is an extreme challenge always pushing the invention of new tools. The WRM is an analog computing concept, invented by R. Cardarelli in 1992, aimed to find in a few ns vertexes in LHC particle detectors. While it was not implemented as it was too avant garde for its times, its ability of detecting patterns within an high level of noise can push artificial vision to a new level of performance.

Lectures

  • J. Abdallah At the frontier of High Energy Physics: What does it mean to search at the edge of knowledge? How do we test new ideas from cosmology to particle physics? We will seek answers to these questions with concrete examples from the latest research into the secret structure of the universe. We will discuss the experimental challenges in High Energy Physics and see their intricate relation with innovation.
  • G. Aielli “The WRM concept, from the invention to the innovation”: The weighting resistor matrix (WRM) is conceptually a pattern recognition analog processor, based on a special resistive network. The working principle is based on transforming a given analog or digital data pattern from the real space to an arbitrary parametric space, scoring it with a likelihood estimator, which measures how good each given set of parameters describes the data. The proposed technique recalls some aspects of the Hough Transform, so that we may consider that both techniques fall under the same generalized category. However, there are crucial differences between the WRM and Hough Transform or other pattern recognition techniques:
    • the WRM embeds the concept of matching uncertainty, and it is able to extract information even in presence of approximate matching, missing portions and severe background noise
    • while the Hough Transform is a very highly expensive algorithm in terms of processing power, the WRM performs its operation on the chip in a single clock cycle, thus is extremely fast;
    • in the WRM there is no need of memory growing dramatically with data size and pattern complexity.
    By this means the WRM calculates arbitrary pattern fit at the nanosecond scale with least square fit effectiveness. The original idea was conceived in 1993 as a result of an R&D program for designing a very fast topological trigger scheme for the ATLAS experiment. A reduced size prototype IC was subsequently designed and built according to this principles, in a very simplified implementation. We will illustrate the basic principles of the WRM technique and its potential for high energy physics experiments and how this concept can be exploited for a very innovative artificial vision system.
  • A. Abdallah “Feature Detection and Classification”: Classifying pattern is a fundamental problem in general in any science field. The weighting resistor matrix (WRM) is developed as a fast trigger processor in high energy where detecting signal (pattern) in data at high speed is crucial. In this talk I explain the use of the WRM technology to implement a method for detecting discrete straight lines segments in binary images. This solution has been developed and adopted within the EDUSAFE project [1].
    We give a brief description of the WRM chip and its mathematical properties; we also provide directions of further researches based on the fast Fourier transform to achieve a real time fast general feature detector.
    [1] http://edusafe.web.cern.ch/edusafe/site.php

Announcements