Dorota Iwaszczuk Technical University of Munich
Zoltan Koppanyi The Ohio State University
Charles Toth The Ohio State University

14E

Due to urbanization and expected growth of the population in the cities, modern solutions for urban planning and data management are required. In response to this need, the Smart City concept has been developed over the past years with a main focus on mobility, including autonomous driving, pedestrian navigation in buildings, and smart building infrastructure. All these objectives require accurate models of urban areas, especially buildings, resulting in increased interest in mapping large urban areas, typically using mobile mapping systems. Data delivered from these systems include detailed geometries of an urban environment, including road networks, vegetation, city furniture, and building façades together with detailed structures, such as windows and doors. Mapping building indoors, however, requires other measurement strategies. Various approaches employing photogrammetric methods, laser scanners or their combination have been proposed to map building interiors. The main challenge to map indoor environments is sensor georeferencing and alignment with outdoor geometries.

In this paper, we propose an approach for mapping building indoors using a laser-based measurement system with additional RGB cameras and connecting them to outdoor geometries by coregistration using indoor-outdoor transitions, such as doors and windows.  In order to localize these transitions, we conduct pixel-level classification on images. For this purpose, we use an encoder-decoder Convolutional Neural Network (CNN). Object labels being output of the CNN-based classification are then transferred to the point clouds measured by the laser scanners, which lead to a 3D localization of indoor-outdoor transitions in indoor datasets. Knowing the position of the transitions in outdoor data, we search for corresponding transitions in the indoor data by creating correspondence hypotheses which are verified using RANSAC approach. After finding the best correspondence hypothesis, we estimating the similarity transformation between them, which leads to the alignment of the two datasets.

In order to test the proposed methodology, we design a prototype mobile mapping system, which can operate in both, indoor and outdoor environment. This mapping system, conceived in a backpack configuration, consists of three LiDAR sensors, six RGB cameras, a depth camera, two GPS receivers, and two inertial measurement unit (IMU); note that Ultrawide Band (UWB) transmitter will be used to obtain reference trajectory data. First, calibration results of this system including calibration of all cameras as well as boresight and leverarm calibration of all sensors with respect to the IMU are presented, and the data collection inside and outside the buildings is described; note that educational building with large number of windows and multiple entrances are used. Then, we present results on pixel-wise labelling. We use a CNN trained on a dataset consisting of 90% public datasets and 10% manually labelled data originating from our data collection. For evaluation of the results on transition identification, we use 100 manually labelled images different from those used for training and calculate the accuracy and intersection over union (IoU). We also investigate the performance of the mapping system in terms of the accuracy of mapping in indoor and outdoor environments.

13:45 Towards Seamless Outdoor and Indoor Mapping, Zoltan Koppanyi