Stereovision Mobile Mapping: System Design and Performance Evaluation

Abstract

ABSTRACT:In this paper we introduce a state-of-the-art stereovision mobile mapping system with different stereo imaging sensors and present a series of performance tests carried out with this system. The aim of these empirical tests was to investigate different performance aspects under real-world conditions. The investigations were carried out with different cameras and camera configurations in order to determine their potential and limitations for selected application scenarios. In brief the test set consists of investigations in geometric accuracy, in radiometric image quality and in the extraction of 3d point clouds using dense matching algorithms. The first tests resulted in an absolute overall 3d accuracy of 4-5 cm depending mainly on the quality of the navigation system used for direct georeferencing. The relative point measurement accuracy was approx. 2 cm and 1 cm for the 2 megapixel and the 11 megapixel sensors respectively. In the second series of investigations we present results obtained by applying tone mapping algorithms often used for high dynamic range images (HDR). In our third series of tests refinements of the radiometric calibration of the stereo system and corrections of the vignetting resulted in an improved completeness of the depth map on the roadside environment. We conclude the paper with investigations in obtaining complete and accurate depth maps for a robust and accurate 3d monoplotting support in vision-based mobile mapping web services. For this we present first results obtained by interpolating incomplete areas in the disparity maps and by fusing the disparity maps with additional LiDAR data.

Cloud-Based Geospatial 3D Image Spaces—A Powerful Urban Model for the Smart City

“…It includes the determination of the interior orientations of the 20+ sensor heads; the determination of relative orientation parameters among all sensors heads; and, finally, the calibration of lever arm and misalignment between imaging sensors and the IMU body frame. For a description of the calibration procedure and an evaluation of the results, we refer the reader to [20].…”

“…This temporal coherence is particularly important in urban environments with numerous moving objects such as cars, trams or pedestrians (see Section 4.4). In outdoor urban environments, the requirement of dense and temporally-coherent RGB-D values with a sufficiently high spatial resolution can currently only be met by stereo or trinocular camera setups [6,7,14,20] and a subsequent depth extraction using dense image matching (see Section 6). Once active range imaging sensors will reliably work in bright daylight and provide a sufficiently high spatial resolution, they might become the sensor technology of choice for acquiring urban 3D image spaces.…”

“…Calibration of a multi-sensor mobile mapping platform such as ours is documented in [10,20,[32][33][34]. It includes the determination of the interior orientations of the 20+ sensor heads; the determination of relative orientation parameters among all sensors heads; and, finally, the calibration of lever arm and misalignment between imaging sensors and the IMU body frame.…”

“…These approaches often lead to visually-appealing, photorealistic or abstracted, possibly even semantic, urban models [12,18,19], but they are not well-suited for unstructured environments or for urban models with high to very high accuracy requirements, e.g. for high-quality 3D measurements [7,[20][21][22].…”

“…Data for the subsequent tests was acquired using the multi-stereovision mobile mapping research platform of the Institute of Geomatics Engineering (IVGI) at FHNW [20]. The current acquisition system shown in Figure 3 has the following features [31]:…”