PARK is mainly composed of urban site with elevation change. Various characteristics including forest, pedestrian road, narrow corridors and more can be found.
FOREST is a terrain where trees are sporadically planted. People are continuously appearing in this sequence.
LAWN focuses on poor environments including under-canopy. In this sequence, difficult situations such as the robot stepping on a high grass or hitting it directly are included.
Our dataset provides survey-grade map and ground-truth trajectory of our campus sequences
We utilized RTC360, a automated survey device which provides a precise photo-realistic point-cloud map to generate Ground-truth map in environments where each platforms obtained data.
Each obtained map has thoroughly removed dynamic objects and noisy point-cloud generated from glass, water, and sun glare.
Using scan-to-map localization method on survey-grade prior map, we generated Ground Truth Trajectory in each environment.
RTC360 is a survey device which can output very dense and colored point cloud up to 130m with mm-level ranging accuracy.
Accuracy of RTC360 comes as follows :
Angular accuracy 18”
Range accuracy 1.0 mm + 10 ppm
3D point accuracy
1.9 mm @ 10 m
2.9 mm @ 20 m
5.3 mm @ 40 m
Ground-truth map is processed through Leica mobile-device app to automatically register scans in the field, and post-processed for fine registration and optimization.
Our Ground-truth map offers multiple spatial and spectral information, which comes with intensity and color information in each point-cloud.