Structure-Based Object Classification and Recognition for 3D Scenes in Point Clouds

Scene understanding is a critical issue in the advances of intelligent space. Given a 3D point cloud captured for a static scene by laser scanning, we propose a structure-based scene representation and object recognition method. Our method consists of three steps: (1) the scene is segmented and all of the segmented objects in scanned scene are described respectively by features, including size, location, shape, relationship and constraints between multiple-objects which make it easier to construct the logic representation of geometric and semantic features in scene. (2) Those features are represented by first-order logic and corresponding inference rules are designed to different objects, which could provide prior knowledge for scene understanding and recognition. (3) Then concrete knowledge reasoning rules is used to realize objects recognition including scene structure and details of single objects. Experimental results have shown that our proposed method is simple and effective, and can be applied to different outdoor scenes.

I. Introduction

Scene understanding is an important component in the advances of intelligent space. The rapid development of 3D scanning technique has been greatly boosting the growth of point-based scene representation. However, the shape diversity and topology complexity have undoubtedly increased the difficulties in scene understanding and recognition.