In addition, because it requires high computational complexity to solve the correspondence problem, it makes real-time processing a challenging problem. The authors of [11] strived to make it a real-time process using a single field programmable gate array (FPGA). The 3D laser scanner [14,16], which uses an optical triangulation method, can obtain the most precise 3D face data among all other 3D face data acquisition systems. However, the system is too expensive and the acquisition time is slow. The structured light system consists of one camera and one beam projector [12,13]. To solve the correspondence problem, the structured light system projects structured light patterns onto a face.
However, it is difficult to perform a calibration between the camera and beam projector; moreover, the person whose face is to be acquired may feel uncomfortable from the strong light projected directly onto their face. Table 1 shows the comparison of three systems in terms of cost, scan time, accuracy, and aversion.Table 1.Comparison of three-dimensional (3D) acquisition systems: laser scanner (LS), stereo vision system (SVS), and structured light system (SLS).Recently, real-time 3D data acuiqisiton sensors such as time-of-flight (ToF) and kinect sensors have been introduced, and have become very useful sensors for vision applications [18,19]. Even though real-time 3D depth sensors make it possible to analyze detaied 3D shape information, 3D data acquired by those sensors contain depth noise.2.2.
3D Face Recognition3D face recognition systems generally outperform 2D face recognition systems because the 3D systems exploit depth information to analyze 3D face shapes, which are invariant to external environments. In general, the methodologies of 3D face recognition can be classified Anacetrapib into three types: depth-map-based [20�C22], curvature-based [23,24] and profile-based [25�C27]. Depth-map-based methods use range images, which contain the depth information of a face [20�C22]. The range images provide features that are invariant to changes in internal and external conditions, such as light and pose changes. It shows robust recognition performance against pose and light variation compared to 2D face recognition [20�C22]. However, it has difficulty registering between range images and requires high computational costs. Curvature-based methods use face curvatures and line features of the face [23,24,28].
They account for geometrical information of
Processing capabilities in sensor nodes are typically based on Digital Signal Processors (DSPs) or programmable microcontrollers. However, the use of Field Programmable Gate Arrays (FPGAs) provides specific hardware technology, which can also be reprogrammable thus providing a reconfigurable sensor system. The partial reconfiguration is the process of modifying only sections of the logic that is implemented in an FPGA.