Focus set based reconstruction of micro-objects

摘要

With a light optical microscope and a motorized Z-table, the focus depth of a micro-object can be determined. This paper presents an algorithm for determining the depth map of the surface of a non-translucent micro-object based on focus set data. If a robot-manipulator wants to grip a known object, such a depth map can be used to determine its position. Surface data is also useful for automatically determining gripping points. Problems such as this are addressed by the current MiCRoN 1 project. Many algorithms addressing the problem of extracting depth maps can be found. However, most of them suffer from system- atic errors. The algorithm presented in this paper efficiently suppresses these systematic errors by using a multi-scale ap- proach.