Learning methods for the analysis of environmental data for unmanned vehicles

Within this project, the Active Vision Group is researching solid-state LIDAR sensors for measuring environmental data. The advantage of this new sensor technology over classic LIDAR systems is that it does not require any moving parts, thus the robustness of the sensors increases.

In addition, segmentation and classification of environmental data is improved. For this purpose, the temporal relationship of sequential data is used to achieve a higher consistency of segmented and classified images.

Furthermore, a data format is developed for storing environmental measurement data and annotated data of different sensor modalities. In order to accelerate the annotation of new measurement data for the creation of basic truths, an AI-based annotation method is implemented.