Verifying the accuracy of machine vision algorithms and systems

Abstract

The authors attempt to summarize important parameters and procedures for verifying the measurement and recognition (classification) accuracy of machine vision algorithms/systems; to alert the machine vision research community to the current, very inadequate practice in this important area; and to propose some measures to improve this situation. Two example applications from the authors' practice are given in order to illustrate experimental verification procedures presented here. The motivation for the present work is based on the fact that machine vision systems are very hard to model or simulate accurately. It is concluded that realistic large-scale experiments are the most reliable means of verifying the measurement and recognition accuracy. However, it is claimed that in the machine vision (research) community, this verification is seldom done adequately.