Akademik Veri Yönetim Sistemi
Dicle Üniversitesi Mühendislik Fakültesi Mühendislik Dergisi, cilt.13, sa.1, ss.1-9, 2022 (Hakemli Dergi)
A point cloud is a set of points mathematically portrayed in a vectorial space or in the x, y, z coordinate plane of the object. Points are classified in the registered spatial coordinate plane and define semantic information that represents an object or an area. As technology develops, 3D point clouds have recently become eminently prominent in the range of classification, detection, and recognition of objects. Different data sets are attained by transforming the scanned objects into 3D point clouds and transferring them to the virtual environment with lidar scanning systems. Classification methods are developed to successfully classify 3D point data which is analyzed in detail. Dense point clouds of different sizes are created by encompassing distinctive information (depth or RGB) together with the 3D coordinate plane to the point clouds. In addition, for each point in the point cloud dataset; external or internal information has been supplemented and objects are colored with RGB values. In this paper, the success performances, advantages and disadvantages of the methods classifying the 3D point cloud with deep networks have been analyzed. In particular, the applied algorithms, the tried methods, and the created models are compared and discussed. As a result, the existing methods for giving speed and direction to future work are presented in a comprehensive manner.