In many industrial environments physical process values are still measured using analog gauges. An automatic reading of these instruments using wireless or wired data connections is often not feasible, due to cost- or technical issues. In such cases manual reading using human personnel is currently state-of-the-art. Especially in large plants (e.g. oil gas refinery) these tasks are very error-prone, time-consuming and often dangerous due to exposed positions of the gauges. Recent research in the domain of autonomous robots and machine vision broadened the way of implementing either assistive or fully autonomous systems for reading analog gauges in hazardous environments. Therefore this work introduces a novel method for the detection of the gauge itself, the detection of the pointer needle and the extraction of information from the markings on the gauge. This allows the extraction of a physical measurement value without any prior knowledge about the gauge.