An open system for 3D data acquisition from multiple sensor Francesco Isgr` o, Francesca Odone, Alessandro Verri INFM - DISI Universit` a di Genova Genova, Italy {fisgro,odone,verri}@disi.unige.it Abstract— This paper describes a work in progress on a multi- sensor system for 3D data acquisition. The system core structure is a 3D-range scan based on the well known active triangulation procedure and made of a camera, a laser light emitter and a software driven motor. The core system allows us to acquire dense point clouds of objects of about 50 cm. The system today hosts a second camera and thus is able to perform 3D reconstruction from two slightly different viewpoints and produce more dense point clouds. Also, since the motor can be driven back to the original position multiple scans can take place, to obtain smooth surfaces, and multiple information, such as texture and reliability measures. An alternative way of obtaining texture information is by means of a linear camera, also included in the system. We present results obtained with the current system, and describe extensions of the system in estimating noise and producing a more complex geometry description. I. I NTRODUCTION The target of machine vision is to make computers see. In less philosophical words the main objective of this discipline is to extract some kind of information from images, that can be used for preforming some tasks. Among the various tasks that may need computer vision modules we mention remote sensing [14], [13], inspection [17], robot guidance [19], medical tasks [7], etc. For most applications some kind of 3D description of the scene is required. A variety of 3D reconstruction algorithms do exist, and they have been grouped in several classes of algorithms [23], [6]. We can think of dividing all the approaches into two main classes named passive and active methods. In the first class fall all those methods that not use any kind of energy to help the sensors, such as stereopsis [21] or shape from shading [25]. They do count only on the imaging hardware, so that they need, in general, very simple set-ups, but have a certain number of challenges to overcome. Active methods project energy (e.g. a pattern of light, sonar pulses) on the scene and detect its position to perform the measure; or exploit the effect of controlled changes of some sensor parameters (e.g. focus). Active range sensors exploit a variety of physical principles; examples are radars and sonars [24], Moir` e interferometry [10], focusing [16], and triangulation [3]. The system we describe in this paper is based on the active triangulation paradigm. The basic geometry for an active Fig. 1. Structured light systems use triangulation methods to obtain the 3D measures. This is a description of the system in the camera reference frame: Z is the optical axis, O the origin, b the baseline, i.e., the distance between the optical centre and the laser. triangulation system is shown in Figure 1. A light projector (typically a laser) is placed at a certain distance from the centre of projection of a pin-hole camera. The projector emits a plane of light intersecting the scene surfaces in a planar curve called the stripe, which is observed by the camera. If the position of the camera with respect the laser plane is known it is possible to recover the 3D by simple triangulation (see Figure 1). The system we are developing (shown in Figure 2) is a multi-sensor system, the core of which is a multi-camera 3D range scan based on the active triangulation principle. The objects are scanned while moving on a conveyor-belt controlled by a software driven motor. The high accuracy of the motor control permits to have easily multiple scans of the same objects. The advantage of a multi-camera system is that some problems generated by self-occlusion on the object surface can be overcome since there are different view-points. The measurements from the different views should not need to be registered (e.g, using an ICP algorithm [2]), as the system is calibrated and the measures are with respect the same reference frame. In the paper we discuss the state of the art of the system, and, being a work in progress, we also describe the future developments we are planning in terms of algorithms and hardware components. The paper is structured as follows. The next section de- scribes the state of the system and how it currently works. In Section III we discuss the advantages of the addition of more Proceedings of the Seventh International Workshop on Computer Architecture for Machine Perception (CAMP’05) 0-7695-2255-6/05 $20.00 © 2005 IEEE