Surveying private pools in suburban areas with neural network based on drone photos

Abstract

In many cases, there is a need for local authorities to carry out a territorial survey of municipalities. The purpose of surveys may be to refine the data or to verify the information that forms the basis of local taxes. Such data include the possession of private outdoor swimming pools belonging to family houses. The knowledge of the number of pools is important in many ways. Drainage load will increase significantly when pools are drained for which the channel operator needs to prepare. In the same way, filling also has a significant water demand for which the water supplier needs to be prepared. For all these reasons, it is understandable if the local government requests an extra contribution from the pool owners in the form of a tax. Up-to-date and accurate data are important for municipalities. Aerial photography can be used to recognize outdoor pools with ease. With drones, the process is fast and cost-effective. At the same time, the numerical quantification of pools from a large amount of data produced during the photo shoot (high-resolution orthophoto covering the area) is a time-consuming task. The method we have developed can automatically detect basins and to quantify them by using a neural network on orthophotos.