SHOP
English (EN) 
Deutsch (DE) 
Automated tool recognition and measurement based on AI and 3D machine vision
Seamless integration of 2D/3D and AI
Intelligent tool recognition inside the box
Automated tool inspection is becoming increasingly important in modern manufacturing. With the Toolfinder, Ziegelbauer offers an intelligent solution for precise tool recognition using AI, deep learning, and 3D image processing – a future-proof response to rising demands for efficiency and quality.
Developed in close collaboration, the Toolfinder is based on a compact, high-performance system featuring advanced camera technology and intelligent image processing.
At its core are 8 GigE color cameras, each equipped with high-resolution lenses to enable fast and accurate tool identification.
Working principle of the detection unit
- Tools are placed in trays of varying diameters.
- The system detects tool diameters from 6 mm to 40 mm and tool heights up to 180mm.
- For extra-long tools, the camera system can be raised by 90mm, allowing the detection of lengths between 90 mm and 180 mm.
- The measuring range of 200 x 400 x 90 mm is fully covered by 8 cameras, calibrated with a Halcon stereo model.
- Overlapping fields of view from four camera pairs ensure a complete 3D coverage and eliminate occlusions.
- A closed light chamber with diffuse illumination provides consistent lighting conditions, even for highly reflective surfaces.
Software architecture & machine vision
The solution is based on the powerful HALCON image processing library from MVTec. The complete application was developed by phil-vision. From camera integration and calibration to stereo reconstruction and 3D processing to final visualization, the versatility of HALCON proved to be the ideal solution for meeting the high technical demands of this project.
- Deep learning-based object detection is applied to identify tool tips across images from all four cameras. Tips detected in at least three camera views are visually highlighted
- The tool height is determined by averaging the height of the detected tool head.
- A 3D reconstruction of the tool is then performed.
- Deep learning classification is used to categorize tools into one of seven predefined types.
- In parallel, trained object detection models are used to detect the presence of cooling channels.
Boosting efficiency with the Toolfinder
By integrating the Toolfinder, Ziegelbauer has achieved significant improvements in production logistics. Automated tool recognition and precise measurement have led to substantial time and resource savings, while also reducing manual handling and minimizing potential sources of errors. This marks a major step towards a fully automated and streamlined internal logistics process. Powered by AI-driven algorithms, the Toolfinder delivers not only high accuracy, but also future-proof scalability. After almost a year of successful operation during the prototype phase, the Toolfinder is now ready to be introduced to a broader customer base.
Max Ziegelbauer on the collaboration with phil-vision: "Our partnership with phil-vision has been exceptional—fast, straightforward, and built on mutual respect. We always have a knowledgeable and responsive partner at our side, someone who quickly addresses challenges and delivers effective solutions. We're proud to showcase the Toolfinder at automatica in Munich, presenting it for the first time to a wider professional audience as a state-of-the-art example of seamless integrated machine vision in modern tool manufacturing.”
