Computer Vision

Our competences

For humans, recognising objects and spatial vision are a matter of course. However, computer programmes have only been able to achieve human-like vision for just over a decade. Specific applications nevertheless present new challenges. We support you in the search for solutions and develop them for your target platform.

Detection of objects and anomalies

Objects are represented in images in a variety of ways. Accordingly, a larger number of categorised images per object class is required to train classic or AI-based methods.

Automatic and human-supervised labelling can be combined to generate categorised image data with minimal effort. High reliability for industrial applications can often be achieved by using additional sensors, e.g. depth cameras or high dynamic range (HDR).

Localisation and tracking in the room

Autonomous navigation of robots is based on cameras and other sensors. Technologies such as GPRS, radio links, lasers, magnetic fields, ultrasound and passive and active markers can also be part of a suitable solution. For their seamless interaction, information from these sensors is fused and context information is taken into account.

Efficient implementation

Computer vision algorithms are computationally complex. Their efficient implementation for a fast runtime is crucial for many practical applications. From modern CNN-based AI to classic image processing algorithms, we realise algorithms on suitable target platforms on microcontrollers or dedicated edge AI platforms.

Algorithms must be skilfully woven into good software. For good maintainability and extensibility, we work together with the Embedded Software Engineering team at ICOM. Our partner institute IMES implements algorithms on FPGAs and in ASICs.

We grow with our tasks

Image processing, computer vision and sensor fusion can be used to realise a wide range of useful applications. We are interested in your task. Please contact us without obligation.