An architecture for automatic multi-modal video data anonymization to ensure data protection.

摘要

To perform a data protection concept for our mobile sensor platform (MODISSA), we designed and implementedan anonymization pipeline. This pipeline contains plugins for reading, modifying, and writing di erent imageformats, as well as methods to detect the regions that should be anonymized. This includes a method todetermine head positions and an object detector for the license plates, both based on state of the art deeplearning methods. These methods are applied for all image sensors on the platform, no matter if they arepanoramic RGB, thermal IR, or grayscale cameras. In this paper we focus on the whole face anonymizationprocess. We determine the face region to anonymize on the basis of body pose estimates from OpenPose whatproved to lead to robust results. Our anonymization pipeline achieves nearly human performance, with almostno human resources spent. However, to gain perfect anonymization a quick additional human interactive postprocessingstep can be performed. We evaluated our pipeline quantitatively and qualitatively on urban exampledata recorded with MODISSA.