ANNOUNCEMENTS
5G, the upcoming mobile network is the fastest network ever developed having potential to provide speed of 10 gigabits per second and ultra-low latency. It consists of high band mmWave frequencies which cannot penetrate buildings and canopies hence requires spatial decision support tools to maximize its coverage and select best possible site for radio frequency (rf) planning. This site selection process is carried out twin’ which requires high resolution GIS inputs called ‘geodata’ using simulation in a ‘digital twin’ with submeter accuracy. Here, a methodology has been proposed to create high resolution geodata which are required for radio frequency planning which can help in site selection and effective planning of 5G networking such as line of sight analysis. Specifically, these geodata are, Clutter i.e. land use surface information, 3D building models, Digital terrain model i.e. bare earth elevation, Tree crown cover or canopy model. The data used for creation of clutter is high resolution Worldview 2 RGB imagery, using which, Support vector machine algorithm was carried out to produce land use with an accuracy of 87% and further merged with building footprint data to produce clutter based on Maryland classification scheme. The 3D building model was created using footprint data consisting of height attribute with vertical accuracy of 50 cm. Terrain model was obtained from lidar data & this lidar data was further segmented using PointCNN, a deep learning model. The precision and recall achieved from PointCNN for canopy is 98.3% and 96% respectively. Analytic hierarchy process (AHP) and weighted overlay analysis were carried out with geodata to locate suitable sites. 90 locations were identified to be highly suitable for rf planning. 5G is more than just data flow, its about IoTs, location intelligence, smart cities and vehicles etc.
Keywords: 5G, spatial rf planning, Geodata, PointCNN, AHP
