DNN-Based Map Deviation Detection in LiDAR Point Clouds

In this work we present a novel deep learning-based approach to detect and specify map deviations in erroneous or outdated high-definition (HD) maps using both sensor data as input neural network (DNN). We first our proposed reference method for deviation detection (MDD) utilizing sensor-only DNN detecting traffic signs, lights, pole-like objects LiDAR data, with obtained by subsequently comparing detected examined map. Second, facilitate the object task additional network. Third, employ specialized MDD directly infer correctness of input. Finally, demonstrate robustness challenging scenes featuring occlusions reduced point density, e.g., due heavy rain. Our code is available at https://github.com/Volkswagen/3dhd_devkit.