A multi-resolution approach for an automated fusion of different low-cost 3D sensors.

Jan Dupuis,Heiner Kuhlmann,Stefan Paulus,Jan Behmann,Lutz Plümer

doi:10.3390/s140407563

Abstract

The 3D acquisition of object structures has become a common technique in many fields of work, e.g., industrial quality management, cultural heritage or crime scene documentation. The requirements on the measuring devices are versatile, because spacious scenes have to be imaged with a high level of detail for selected objects. Thus, the used measuring systems are expensive and require an experienced operator. With the rise of low-cost 3D imaging systems, their integration into the digital documentation process is possible. However, common low-cost sensors have the limitation of a trade-off between range and accuracy, providing either a low resolution of single objects or a limited imaging field. Therefore, the use of multiple sensors is desirable. We show the combined use of two low-cost sensors, the Microsoft Kinect and the David laserscanning system, to achieve low-resolved scans of the whole scene and a high level of detail for selected objects, respectively. Afterwards, the high-resolved David objects are automatically assigned to their corresponding Kinect object by the use of surface feature histograms and SVM-classification. The corresponding objects are fitted using an ICP-implementation to produce a multi-resolution map. The applicability is shown for a fictional crime scene and the reconstruction of a ballistic trajectory.

Highlights

In recent years,various optical 3D-sensors have become available and are nowadays used in many different fields of work, e.g., reverse engineering or quality management in industrial tasks, cultural heritage, medicine and criminal investigations [1]
We focus on the fusion of unorganized point-clouds coming from different sensor using a characteristic 3D shape description, the so-called “Surface Feature Histograms” [20], in combination with a support vector machine (SVM) classification [21]
If the book was lying on its cover, the bottom side was not imaged by the Kinect sensor in contrast to the David system, resulting in a misallocation of corresponding points in the ICP algorithm and, a higher root mean square error (RMSE)

Summary

Introduction

In recent years,various optical 3D-sensors have become available and are nowadays used in many different fields of work, e.g., reverse engineering or quality management in industrial tasks, cultural heritage, medicine and criminal investigations [1]. Several different sensor technologies can be used for 3D digitizing, like terrestrial laser scanners (TLS), triangulation-based range sensors or photogrammetric approaches, like stereo cameras or bundle adjustment of multiple images [1,2,3,4,5,6,7]. All of these sensors have their own limitations regarding flexibility, measuring volume, spatial resolution and accuracy. These sensors have more strict limitations regarding measuring volume and resolution

Methods

Results

Discussion

Conclusion