A new generation of machine vision intelligence

HALCON 26.05 introduces automatic contour optimization for shape-based matching (SBM), enabling users to automatically remove unstable or misleading contours from SBM models — eliminating labour-intensive manual tuning.

Reflections, shadows, and random texture often introduce unreliable contours that reduce matching robustness. With sample images of real object instances, HALCON now identifies which contours appear consistently across the dataset and refines the SBM model automatically.

1. Speed and Performance Optimization The core focus of this release is accelerating machine vision applications for demanding industrial automation scenarios. Both deep learning methods and rule-based algorithms have been heavily optimized to ensure operations are not only precise and robust but also execute at exceptional speeds, meeting the strict cycle-time requirements of modern production lines.
2. Automatic Contour Optimization for Shape Matching Shape-based matching is critical for locating objects, but reflections, shadows, and random textures often create unstable contours that require time-consuming manual cleanup. HALCON 26.05 introduces a data-driven automatic contour optimization feature. By analyzing sample images of real object instances, the system automatically identifies and retains only stable, consistent contours while discarding unreliable ones. This makes matching significantly faster, more accurate, and highly robust, especially for reflective mechanical parts or electronic components in feeder-based pick-and-place systems.
3. Data Matrix Rectification Reading codes on non-flat materials is a common industrial challenge. HALCON 26.05 expands its code-reading capabilities with a new Data Matrix rectification function. This feature compensates for geometric distortions on curved or deformed surfaces (such as cylindrical components, curved packaging, or flexible materials) before decoding. While it adds a slight processing overhead, it dramatically improves decoding reliability in demanding physical environments.
4. Enhanced Data Augmentation for Deep Learning The software introduces a modernized, operator-based approach to data augmentation, replacing older procedure-based methods. Developers can now define flexible augmentation pipelines programmatically directly within HALCON. This includes applying geometric transforms, color variations, and blurring, with the ability to preview the resulting image variations. This streamlined workflow improves model robustness and generalization, reducing the reliance on massive training datasets, and is initially available for object detection and instance segmentation tasks.
5. Next-Generation AI Object Detection HALCON 26.05 delivers a new generation of deep-learning-based object detection that achieves up to five times faster inference speeds while maintaining high detection accuracy. Utilizing an anchor-free detection approach, it performs reliably even with small objects or strongly varying object sizes. Combined with integrated data augmentation, the models are highly robust against real-world challenges like illumination changes, rotation, distortion, and partial occlusion. Pretrained models from MVTec are available to accelerate application development.
6. HDevelopEVO and AI Assistant Enhancements The integrated development environment, HDevelopEVO, receives a significant preview update. Developers can now integrate scripts created in HDevelopEVO directly into their own custom applications using the .NET interface. Furthermore, multimodal Large Language Model (LLM) support has been expanded to include visual prompting, allowing developers to incorporate actual image data directly into prompts for the HALCON AI Assistant, streamlining the development and troubleshooting process.