Towards the Crowdsourcing of Massive Smartphone Assisted-GPS Sensor Ground Observations for the Production of Digital Terrain Models.

Digital Terrain Models (DTMs) used for the representation of the bare earth are produced from elevation data obtained using high-end mapping platforms and technologies. These require the handling of complex post-processing performed by authoritative and commercial mapping agencies. In this research, we aim to exploit user-generated data to produce DTMs by handling massive volumes of position and elevation data collected using ubiquitous smartphone devices equipped with Assisted-GPS sensors. As massive position and elevation data are collected passively and straightforwardly by pedestrians, cyclists, and drivers, it can be transformed into valuable topographic information. Specifically, in dense and concealed built and vegetated areas, where other technologies fail, handheld devices have an advantage. Still, Assisted-GPS measurements are not as accurate as high-end technologies, requiring pre- and post-processing of observations. We propose the development and implementation of a 2D Kalman filter and smoothing on the acquired crowdsourced observations for topographic representation production. When compared to an authoritative DTM, results obtained are very promising in producing good elevation values. Today, open-source mapping infrastructures, such as OpenStreetMap, rely primarily on the global authoritative SRTM (Shuttle Radar Topography Mission), which shows similar accuracy but inferior resolution when compared to the results obtained in this research. Accordingly, our crowdsourced methodology has the capacity for reliable topographic representation production that is based on ubiquitous volunteered user-generated data.