Industrial machine vision system for fast and precise 3D object localization

Abstract

Machine Vision for industrial applications requires simple-to-use and robust solutions. Usually industrial machine vision solutions are reduced to a simple 2D application, for which algorithm robustness has already been fully demonstrated, and where a key part of all those visual inspection systems is the efficient commercially-available illumination control systems that ensure the repeatability of captured image intensity. Nevertheless these projects sometimes require the 3D information for applications such as pick and place robotics, and this field of applications is not yet covered by most of the commercial industrial vision brands which actually tend to simplify the programmable visual device and its associated functions. We present in this paper a new fast and robust 3D localization method based on object edge. This process offers an infinite number of possible cases of industrial applications when 3D localization is required, but could also be employed for 3D classification, for example in case of the food industry when fruits are selected depending on their size and must be localized because of the random distribution on a conveyor belt. The presented industrial machine vision system is easy and fast to implement. It results in a robust, precise, and inexpensive industrial solution.