The international market for machine vision systems is growing rapidly and standard software solutions play an especially important role. Mario Bohnacker, Technical Product Manager HALCON, MVTec Software GmbH, tells us why MVTec provides outstanding new and optimised features in its latest release: HALCON 20.11.
In HALCON 20.11, which was released on November 20,2020, the MVTec experts have optimised a number of core technologies. A new 2D code ty pe known as DotCode has been added that is based on a matrix of dots and can be printed very quickly, making it especially suitable for high-speed applications – for example, in the tobacco industry.With another new feature called Deep OCR, MVTec introduces a holistic deep-learning-based approach for optical character recognition(OCR). Deep OCR can localise numbers and letters much more robustly, regardless of their orientation, font type, or polarity.The ability to group characters automatically allows whole words to be identified.This significantly improves recognition performance and avoids the misinterpretation of characters with similar appearances. Deep OCR contains algorithms that bring optical character recognition a crucial step closer to human reading capabilities.
Deep OCR also demonstrates that MVTec does not view deep learning exclusively as a unique technology for solving applications that were previously difficult or even impossible to implement. It vividly shows how “traditional” machine vision technologies can also rise to an entirely new level with the aid of deep learning. An important goal, and part of MVTec’s product strategy, is to establish synergies between deep learning and existing core technologies in order to further improve HALCON’s quality and user-friendliness for customers wherever possible.
Improved Usability and faster 3D matching
The core technology shape-based matching has also been optimised in HALCON 20.11. More parameters are now estimated automatically, which improves both user-friendliness and the matching rate in low contrast and high noise situations. To give an example, the workpiece holder on a machine has special markings, known as reference marks. These marks must be located quickly right from the start in order to align the remain in algorithms with them. However, these workpiece holders are typically used on a machine for very long periods.
Over time, the marks become dirty and start to rust.Their contrast in the image diminishes more and more, and they become difficult to detect. MVTec optimised shape-based matching for this very application. A new operator is now available that enables users to find the right parameters very easily. In these cases, the reference marks can also be detected robustly, which improves usability significantly. The new release demonstrates substantial improvements in the 3D environment as well. Edge-supported, surface-based 3D matching is now much faster for 3D scenes with many objects and edges.Usability has also been improved by eliminating the need to set a viewpoint.
HALCON 20.11 makes things much easier, not only for users but also for developers. A new language interface enables programmers who work with Python to seamlessly access HALCON’s powerful operator set. In this way, MVTec takes into account the increasing importance of Python as a programming language, especially in scientific and university settings. The integrated development environment HDevelop has also been given a facelift. The entire design has been made more attractive thanks to a simpler and more consistent icon language. In addition, it now offers more options for individual configuration, such as a modern window docking concept. Moreover, themes are available to improve visual ergonomics and adapt HDevelop to personal preferences.
Precise edge detection with deep learning
HALCON 20.11 includes a new and unique method for robustly extracting edges with the aid of deep learning.Especially for scenarios where a large number of edges are visible in an image, it takes very few images to train the deep-learning-based edge extraction function to reliably extract only the desired edges. This greatly reduces the programming effort for processes of this type. Out of the box, the pre trained network is able to robustly detect edges in low contrast and high noise situations, which also makes it possible to extract edges that cannot be identified using conventional edge detection filters. In addition, “Pruning for Deep Learning” now enables users to subsequently optimise a fully trained deep learning network. They can now control the priority of the parameters speed, storage, and accuracy and, in this way, precisely modify the network according to application-specific requirements.
Coming soon: A new version of the deep learning tool
The new release is available in both a Steady and aProgress edition. This means that the full range of newProgress features is now also available to HALCONSteady customers. For the first time, they can also use functions like deep-learning-based Anomaly Detection.This feature enables users to achieve outstanding results during automated error inspections within just a short period of time, with a small amount of training data, and without any labelling. Moreover, a new version of the MVTec Deep Learning Tool will also be available in December soon after the release of HALCON20.11. Users will then be able to evaluate their trained network directly in the tool. With this addition, the DeepLearning Tool now covers the entire deep learning workflow for the first time.
