The technique of using aerial laser scanning and neural networks to obtain reliable spatial data of the real estate cadastre

The state cadastral registration as a regulator of the real estate market has a direct impact on the country`s economy. As practical experience shows, at the moment there are many objects that are not on the state cadastral register, and this fact, of course, negatively affects taxation and, consequently, the state and regional economies. To solve the problems of three-dimensional identification of objects, it is proposed to use a technique for recognizing private objects, which includes conducting aerial laser scanning with subsequent processing of the data obtained using a neural network. For this method, the DGCNN software is used, which allows solving the tasks of recognizing objects in a cloud of laser reflection points. The most time-consuming and important stage at working with neural networks is training. Therefore, in order to obtain the highest accuracy of segmentation of laser reflection points, a number of experiments were carried out