The Imaging Source blog

MVTec has just released their latest HALCON Progress Edition: 21.05. Below are the new features and improvements:

• Optimized Subpixel Bar Code Reader
• Deep OCR Improvements
• HDevelop Usability Improvements
• Generic Shape Matching
• HALCON Deep Learning Framework
• Improvements of Basic 2D and 3D Opereators

HALCON's Subpixel bar code readers for low-resolution bar codes: Logistics is experiencing increasing growth in the number of camera-based identification systems. On the warehouse floor, however, there is high-variance in the hardware and other environmental conditions which can lead to less-than-optimal bar code images. HALCON can now read bar codes with even minimal module size. HALCON's improved functionality can increase the decoding rate by up to 50%.

Deep OCR improvements - A holistic deep-learning-based approach: Compared to other algorithms, Deep OCR can localize characters much more robustly - regardless of orientation, font type or polarity. With HALCON 21.05, big images are now handled more robustly, and words of varying lengths - even very long words - can now be detected without parameter tuning. Thanks to the extended set of supported characters in HALCON 21.05, customers also benefit from overall stability improvements and a wider range of possible applications.

HDevelop usability improvements: HDevelop's user interface has also been made more user friendly. The new "Canvas Window" allows users to group and organize graphics windows in a much more convenient manner. Additionally, users now have more options to control and dock the position of floating windows. The new "Auto-hide" feature enables users to quickly shrink unused widgets into the sidebar and easily bring them back when needed.

Generic shape matching: HALCON 21.05 introduces Generic Shape Matching to enhance the usability of MVTec's industry-proven shape matching technologies. By significantly reducing the number of required operators, users can now implement their solution much more quickly and easily. With the new release, it is now possible to strictly delimit the angle and scaling range. Moreover, thanks to the unification of the different shape matching methods into a single set of operators, users can now more efficiently adapt new shape-matching features.

HALCON Deep Learning Framework: With HALCON 21.05, MVTec introduces the first version of the HALCON Deep Learning Framework. This framework allows experienced users to create their own networks within HALCON. With this feature, experts can now realize even the most demanding and highly complex applications in HALCON without having to rely on pretrained models or third-party frameworks.

Improvements of basic 2D and 3D opereators: In HALCON 21.05, 3D-point-cloud sampling supports a new mode called "furthest point" which usually results in a more even sampling of the 3D object. Point cloud smoothing has also been added with the release--a feature which typically significantly improves processing times. Other improvements in HALCON 21.05 include general speedups of basic image operators.

Please Note: The Imaging Source is a certified MVTec distributor for Germany, Austria, Switzerland and number of other countries around the world. Please click here to find the authorized MVTec distributor for your country.

MVTec's newest version of their HALCON Progress Edition, HALCON 21.05, will be released on May 21, 2021. To mark the event, MVTec will host a free webinar (register here) to acquaint users with the latest improvements:

Included in the Upcoming Release:

• Deep OCR improvements
• Generic shape matching
• HDevelop usability improvements
• HALCON Deep Learning Framework
• Subpixel bar code reader improvements
• Improvements to basic operators in 2D and 3D for fast, robust preprocessing

In order to allow maximum participation, the webinar will be held in English and Chinese. In the event you are unable to attend the live webinar, a recording will be available shortly after the live session for those registered.

REGG Inspection designs and manufactures high quality inspection systems using a modular design concept. Based on customer needs, the system uses up to 5 USB 3.0 industrial cameras. (in this case, 5 DMK 33UX252 monochrome cameras) producing 20 images per second. Processing, analysis, reporting and decision-making commands are executed via an industrial computer and PLC using IC Imaging Control SDK and DELPHI-based software. REGG's high-speed systems are used by companies all over the world to ensure production and quality control for a variety of fastener types.

Linux, the highly-stable, open-source operating system, enjoys widespread acceptance among embedded developers because of its low cost, scalability and easy customization. For years, The Imaging Source has continuously maintained and developed its Linux library including the tiscamera SDK for Linux, an open-source project published under the Apache 2.0 license. The latest release for tiscamera SDK includes additions and updates to improve embedded application performance:

• Supports the integration of the NVIDIA Jetson and Raspberry Pi platforms with The Imaging Source's USB and GigE cameras
• Supports The Imaging Source's GStreamer element for MIPI-based cameras on Raspberry Pi 4, enabling standardized camera handling
• General stability and performance improvements

You can find all the changes to the tiscamera SDK in the changelog.

Over 79,000 visitors attended VISION China Shanghai 2021--China's leading trade show for machine vision. The Imaging Source's booth was hosted by reseller Shenzhen SunVision Technology. In addition to a large main island displaying highlights from the entire industrial camera product portfolio, the booth also included a demonstration of The Imaging Source's microscopy cameras (histology specimen) and a live embedded vision demo of a mask-recognition application using The Imaging Source's development kit for NVIDIA Jetson Nano. Many thanks to the SunVision team for their support.