This project addresses the challenges of detecting objects in industrial processes that are difficult to detect due to overlap, poor lighting conditions, and reflective surfaces. Vision systems available on the market are either specialized for a specific application, have a high technical complexity or are not very flexible. In addition, such vision systems are quite expensive. In industrial applications, lighting units are used for cameras, which can compensate for external influences, but exacerbate the problem of reflections. The limitations mentioned above result in a limited possibility of broad and easy integration into existing automation processes.
The project presents an image processing software that is characterized by its extensibility due to its unique way of providing training data using an automated dataset generator. The integration of the vision system into existing industrial processes is possible, given its relatively low hardware requirements, which make it a viable solution even for small-scale industries. The software is based on modern deep learning algorithms and traditional image processing techniques. The aim is to increase the efficiency of industrial processes and to exploit previously unused optimization potential.

Based on our bachelor thesis, the following milestones are targeted from the start of the project in April 2025 until the end of the First Venture “Proof” project in October 2025:

Refactoring of the bachelor thesis and evaluation of the bin-picking system
Revision and optimization the existing research and development work
Systematic performance analysis of the bin-picking system
Implementation of the vision system as a proof of concept in the Swiss Digital Learning Factory “SmartPro 4.0”
Integration of the system into a practical environment to validate its functionality
Gaining insights through iterative trial-and-error analyses

Specification of the business model and financing options
Development of a detailed business model for the market launch of the system (SMC Schweiz AG and early adopters)
Examination and preparation of proof-of-concept license agreements
Application for the First Venture “Validate” program for follow-up funding

In the following project phase (First Venture Validate), the proof of concept will be further developed with the aim of integrating a prototype into the industrial process at SMC and early adopters.