Real-Time 3D Modeling and Visualization Based on RGB-D Camera using RTAB-Map through Loop Closure

This paper, we present a method for real-time 3D modeling and visualization using RTAB-Map with the utilization of 3D loop closure technique. 3D modeling and visualization are performed to create an accurate representation in three dimensions. By modeling in 3D, we can visualize more detailed and realistic representations, making it easier to understand the structure and characteristics. In this method, we collect data through a Kinect camera sensor, and then perform odometry processing to estimate the camera’s position in the observed environment. The data is then used to create a 3D map that is continuously updated in real-time with 3D loop closure techniques, where it will resulting in a more accurate and consistent map. We explained the details of the 3D loop closure process, involving key frame selection, feature extraction and matching, and pose graph optimization. From the results obtained, it requires more than 300 matching data IDs out of 500 image captures to create a 3D model and visualize objects in 3D. When the number of matching data is less than 300, the 3D model and visualization of the object will not be perfect. In conclusion, we found that the proposed method can perform real-time 3D modeling and visualization quite well, based on the testing of the system. This method can be used in various applications such as environment mapping in robotics and others.

I. Introduction

3D mapping is a technique for creating a three-dimensional (3D) model of an environment or object using data from various sensors. Typically, this technique involves gathering data from sensors such as cameras[1], LIDAR[2], which are then processed using certain algorithms to build a 3D model of the environment or object. 3D mapping is often used in the field of robotics, augmented reality, and virtual reality to create 3D models of complex environments[3]. Some common techniques used in 3D mapping include Simultaneous Localization and Mapping (SLAM), Structure from Motion (SfM)[4], Multi-View Stereo (MVS)[5], and so on. Each technique has its own advantages and disadvantages depending on the type of sensor data used, the complexity of the environment, and the purpose of the 3D mapping.