Machine vision standard software
Focus on deep learning: MVTec improves software
MVTec has launched the new version 19.05 of its standard software Halcon. The release is intended to improve processes in industrial image processing with revised functions, and deep learning technologies in particular have been further developed.
One highlight: The new version also allows the deep learning inference to be executed on CPUs with the common Arm processor architecture. This allows customers to use the latest deep learning technologies on standard embedded devices. In addition, deep learning-based object detection has been improved: the process, which clearly localizes and identifies objects in the image using bounding boxes, now also optimally detects the orientation of the objects.
Improvements to matching and code reading
Other new features in Halcon 19.05 optimize the matching processes: Halcon's shape-based matching, for example, now makes it possible to specifically define model areas in which no contours should occur. This leads to more robust matching results in the context of repetitive structures.
The new release also offers some improvements to surface-based 3D matching: for example, it is now possible to better control the quality of 3D edges using additional parameters. This leads to even more robust matches, especially in noisy 3D data.
"With version 19.05, we are taking our proven standard machine vision software Halcon to a new level. Developers and users alike will benefit from numerous new features. These include not only additional, sophisticated deep learning functions, but also noticeably improved usability in the embedded environment," explains Johannes Hiltner, Product Manager Halcon at MVTec. as









