3D Virtual Environments (3DVE) are more and more used in different applications such as CAD, games, or teleoperation. Due to the improvement of smartphones hardware performance, 3D applications were also introduced to mobile devices. In addition, smartphones provide new computing capabilities far beyond the traditional voice communication. They are permitted by the variety of built-in sensors and the internet connectivity. In consequence, interesting 3D applications can be designed by enabling the device capabilities to interact in a 3DVE. Due to the fact that smartphones have small and flat screens and that a 3DVE is wide and dense, mobile devices present some constraints: the environment density, the depth of targets and the occlusion. The pointing task faces these three problems to select a target. We propose a new classification of the existing interaction techniques, according to three axis of classification: a) the three discussed problems (density, depth and occlusion); b) the first two subtasks of the pointing task (navigation, selection); and c) the number of targets selected by the pointing technique (1 or N). In this paper we will begin by presenting a state of the art of the different pointing techniques in existing 3DVE, structured around three selection techniques: a) Ray casting, b) Curve and c) Point cursor. Then we will present our classification, and we will illustrate the classification of the main pointing techniques for 3DVE. From this classification, we will discuss the type of interaction that seems the most appropriate to perform this subtask optimally.