The new HALCON release includes a wide range of functions for training Convolutional Neural Networks (CNNs). It is the first time that customers can train CNNs with HALCON on the basis of deep learning algorithms, using sample pictures of their specific application. Thus, the resulting networks can be highly optimized to the customers’ needs. With this, image data can now be classified easily and precisely. Users can drastically reduce programming requirements, do not need in-depth machine vision expertise, and can save both time and money. Moreover, the new deep learning feature is fully integrated into the industrially proven MVTec HALCON machine vision library. Consequently, users can benefit from the library’s immense range of advantages, along with regular updates, continuous innovation, and competent support.
Detection of defects on reflective surfaces
In addition to the current deep learning technologies, HALCON 17.12 offers many other new features, product improvements, and revised functions: now, defects on reflective surfaces of objects can be easily detected. Using deflectometry, several “light patterns” are projected onto the surface of the object via a screen. The reflections of the grid can be used to reliably identify errors, which could not be achieved with conventional 3D methods. This method is particularly interesting for applications in car manufacturing and the electronics industry. For example, it can be used to detect tiny paint scratches or flaws on electronic components that are not visible to the naked eye. The new HALCON version 17.12 also simplifies the daily work routines of HALCON developers, since it is now much easier to integrate the HDevEngine into the developer's application. Another highlight is the improved automatic text reader, which has become even better at detecting and reading letters and numbers that touch each other.
In addition, the current HALCON release also offers a new method to fuse the data from different 3D point clouds into one unified model. This 3D-based, cutting-edge technology simplifies the post-processing of such point clouds through a much more accurate reconstruction of objects. This is important for processes such as reverse engineering. For example, if there is no CAD model available, or if objects must be analyzed more precisely regarding their 3D properties.
CNNs for easy application development
“With HALCON 17.12, we are consistently refining our strategy to provide cutting edge technology to solve industrial challenges. With the introduction of deep learning functionality we make the added value of CNNs accessible to a wide user base. It enables many companies to train neural networks with their own resources at a much lower cost, and to profitably apply them in an industrial setting,” explains Johannes Hiltner, Product Manager HALCON at MVTec.
Dr. Olaf Munkelt, Managing Director of MVTec Software GmbH, adds: “The wide variety of functions in HALCON 17.12 also reflects the growing importance of artificial intelligence and self-learning algorithms in industrial machine vision. With this new version, we are targeting the requirements of the market – especially in the context of Industry 4.0, respectively the Industrial Internet of Things”.