Bibliographic Metadata

Adaptive machine vision involving nonlinear oscillatory theory for advanced driver assistance systems / Muhammad Ahsan Latif
AuthorLatif, Muhammad Ahsan
CensorKyamakya, Kyandoghere ; Bouchachia, Abdelhamid
DescriptionXV, 184 S. : Ill., graph. Darst.
Institutional NoteKlagenfurt, Alpen-Adria-Univ., Diss., 2011
Bibl. ReferenceOeBB
Document typeDissertation (PhD)
Keywords (DE)Adaptive machine vision / Computer vision / Nonlinear oscillators / Stability analysis / Bifurcation analysis / Routh-Hurwitz stability criterion / Contrast enhancement
Keywords (GND)Fahrerassistenzsystem / Maschinelles Sehen / Bildkontrast / Bildverbesserung / Anharmonischer Oszillator / Nichtlineares System / Stabilität
URNurn:nbn:at:at-ubk:1-24510 Persistent Identifier (URN)
 The work is publicly available
Adaptive machine vision involving nonlinear oscillatory theory for advanced driver assistance systems [11.54 mb]
Abstract (German)

This thesis addresses the two core problems of vision based advanced driver assistance systems, i.e., robustness of the system to the dynamic changes in the lighting conditions and real time processing.

The first problem is related with the continuous variations in the lighting conditions of a dynamic environment. It is well-known that to capture a good image, we need sufficient lighting in the scene which enables the sensor to reproduce a good image out of that real scene.

But, in a real dynamic environment, there is no guarantee that ideal lighting conditions will always sustain. As a result, we may have a sequence of good frames and a sequence of poor frames grabbed by the ADAS. This dependency of ADAS on the dynamic variations in the lighting of the environment leads to an unacceptable performance of the system.

In this thesis, a novel solution is proposed to address this problem which is based on nonlinear oscillatory theory. Settings of the most appropriate parameters of a coupled oscillatory system are derived through stability/bifurcation analysis under which system is highly robust to the dynamic changes in its initial conditions. This property of the system has been successfully tested to handle the dynamic variations in the lighting conditions. The issue of real-time processing has been approached by proposing a novel framework that integrates the proposed oscillatory paradigm with cellular neural network. This integration leads to use any existing analog CNN-UM chip to realize real time processing.

Abstract (English)

Keine Zusammenfassung vorhanden

The PDF-Document has been downloaded 16 times.