Evonik Digital Research

A joint research project of EVONIK and the University of Duisburg-Essen

Detecting Rescue Targets with Video

We use the image information to detect people lying on the floor or waving with their hands. For privacy reasons, we will use a preprocessing approach to anonymize the character information from the cameras. Then we use these anonymous image data for people gesture and situation perception detection. Recognition of humans and their body position in video streams are achieved in this stage.

For the purpose of protecting privacy, using face recognition technology, the face in the image is recognized, and then the face part is deducted from the recorded image. Thus the camera data become anonymized. The data before deducting the face is not saved.

Using computer vision technology achieve people positioning with the image. With the help of a deep learning algorithm, the gesture of people can be recognized. By analyzing people's movements and posture patterns, the perception of the situation in the scene can be known.

By analyzing the above-anonymized data, we can find out whether the people in the current scene has an abnormality, thereby determining whether the current scene is normal or abnormal.