Monday 31 August 2015

Robust pattern recognition and pose estimation.

Stemmer Imaging is extending its machine vision software library Common Vision Blox (CVB) with the new CVB Polimago tool, which enables robust recognition rates at high speed. During the teaching phase, CVB Polimago automatically generates thousands of training images thus significantly reducing the effort for the user.

Pattern recognition is one of the major tasks in machine vision and is used in a variety of applications. The proper and robust functioning of pattern recognition largely depends on the variability of the teaching model. If geometrical transformations such as rotation, size changes and tilting as well as changes in illumination have not been taught during the teaching phase the recognition rate decreases in a later stage of the process. So far, the effort for the user has been significantly high in teaching variable training images as all kinds of geometrical transformations are supposed to be integrated in order to obtain a complete training set.

Stemmer’s CVB Polimago tool for pattern recognition saves its users a lot of time during the teaching phase: the algorithm generates artificial views of the model to simulate various positions of a component in real life. In this way the algorithm is able to learn the variability of the corresponding pattern which enables a reliable identification of the training image at a higher recognition rate even in different views. That is what makes CVB Polimago especially suitable for pattern recognition, pose estimation and object tracking.

The new CVB tool is offered as a module of the existing Common Vision Blox library and in addition to the improved teaching phase, allows a very high speed of execution enabling its use in real time applications. The seamless integration into the existing Common Vision Blox software portfolio makes it simple for those users who are already familiar with using CVB tools for pattern recognition. New users will also be able to quickly use CVB Polimago thanks to numerous application and programme examples.

No comments:

Post a Comment