CPU Inference for Deep Learning
With HALCON 18.05, you are able to perform the deep learning inference, i.e., applying your trained Convolutional Neural Network (CNN) to new data, on a CPU.
This CPU inference has been highly optimized for Intel® compatible x86 processors. As a result, a standard Intel CPU can reach inference execution times comparable to a mid-range graphics processor (GPU).
Removing the need for a dedicated GPU greatly increases the operational flexibility. E.g., industrial PCs that are usually not designed for housing large and powerful GPUs can now easily be used for deep-learning-powered classification.
Variable Input Image Size for Deep Learning Classifier
You can now adjust the input image size of your deep learning classifier. This can help to improve classification results by preserving relatively small relevant information, e.g., a very small defect in a large image. If smaller images are sufficient for the given purpose, reducing the input image size can speed up the classification.
Enhanced Deflectometry
The deflectometry functionality introduced in HALCON 17.12 now includes a new pattern type that improves the precision and robustness of error detection especially on partially specular surfaces like varnished metal sheets.
Improved Bar Code Reader
HALCON 18.05 features optimized edge detection, which improves the ability to reliably read bar codes with very small line widths as well as strongly blurred codes. Moreover, the quality of the bar codes is also verified in accordance with the most recent version of the ISO/IEC 15416 standard.
HDevEngine Improvements
The HDevelop library export feature has been expanded: Developers can now access HDevelop procedures not just in C++, but also in .NET via an exported wrapper – as easily and intuitively as a native function. This significantly facilitates the development process.
Automatic Handle Clearing
HALCON 18.05 makes it much more comfortable to work with handles by clearing these automatically once they are no longer required. This significantly reduces the risk of creating memory leaks because you no longer have to manually release unused memory. This way, writing “safe code” is now much simpler.
3D Improvements
HALCON 18.05 offers optimized functions for surface-based 3D matching: These can be used to determine the position of objects in 3D space more reliably, making development of 3D applications easier. In addition, HALCON now also includes a new helper procedure that allows developers to quickly inspect and debug parameters and results of a surface-based matching application.
Support for Hypercentric Lenses
A new camera model within HALCON now allows the corrections of distortions in images that were recorded with hypercentric (also known as pericentric) camera lenses. These lenses can depict several sides of an object simultaneously, thus enabling a convergent view of the test object. With this technology, users only need a single camera system for inspection and identification tasks, e.g., the inspection of cylindrical objects.