Fraunhofer Institute of Optronics, System Technologies and Image Exploitation IOSB
Fraunhofer Institute of Optronics, System Technologies and Image Exploitation IOSB

Image Processing / Image Evaluation

The automated evaluation of image and video data is a key area of expertise at Fraunhofer IOSB and is an important element of many sensor-based applications. This often involves the use of AI and machine learning methods that enable, for example

  • the recognition of objects in images and videos
  • the tracking of these objects
  • the semantic interpretation of entire scenes or activities.

Our core competencies in this area include the adaptation of complex and data-intensive processes to new applications. We also draw on and advance the latest research, for example on the use of foundation models or large language-image models (see interview “Vision language models take interaction to a new level”).

Broad expertise

Our expertise is not limited to conventional image and video data. It also includes the evaluation of data from other imaging sensors, such as infrared and multispectral data, as well as the creation and evaluation of 3D point clouds.

Since the evaluation methods are implemented and integrated right through to the application, our competencies also include aspects such as support for various standards and interfaces to sensor carriers and runtime optimization of complex AI models for sensor platforms with limited energy or space budgets (e.g., vehicle interiors, drones).

Fields of application

The range of applications for image and video analysis methods is very broad. We have experience in the following areas, among others:

  • Production and quality control,
  • Medicine,
  • Autonomous driving and traffic analysis,
  • Civil security, and
  • Defense.

Example projects in the field of image evaluation

 

© Fraunhofer IOSB/M. Zentsch, Adobe Stock/onlyyouqj [M.]

MODEAS

Modular drone detection and assistance system

Find out more

Creation of heavy rain hazard maps

Classification of UAV data

As part of a collaborative project, Fraunhofer IOSB-INA is developing methods for automated structural analysis based on UAV image data. To this end, an annotated image database is being set up, annotation types and classes are being defined, and the images are being annotated. An AI algorithm is being trained to automatically recognize relevant structures for hazard mapping. After verifying the results of the neural network, the network is integrated into the overall system, including the development of a software application for inferring the trained neural network on UAV image data and the integration of the segmented images into a 3D model.

Project page UAV-SRGK
 

© This content is subject to copyright.

Intelligent video analysis

Greater security and data protection in public spaces – through algorithm-based video surveillance for detecting behavior relevant to the police. 

Find out more

Datenfabrik.NRW

Thermal imaging-based anomaly detection on combine harvesters

As part of the “Datenfabrik.NRW” project, the CLAAS Group has introduced heat image-based anomaly detection in the functional testing of its combine harvesters. This has automated and objectified the testing process. The AI-based application for automatic detection of faulty components not only improves product quality, but also reduces complaints and supports employees in the test cabin.

Find out more
 

© Fraunhofer IOSB

ABUL

Automated Image Exploitation for Unmanned Aircraft

Find out more

Image-based quality inspection process for laminated safety glass

Given the critical importance of flawless laminated safety glass for the safety of people and property, the project aimed to replace the current time-consuming and error-prone manual visual inspection with an automated system that efficiently detects defects such as scratches and streaks.

Find out more