The retroreflective nature of the markers ensures that they appear very bright to the upward-looking camera, allowing them to be identified in the image using relatively simple image processing methods, and ensuring that they are not swamped by the studio lights. A small upward-looking monochrome camera surrounded by a ring of LEDs is mounted on each studio camera to be tracked ( Fig. After the rings have been merged to a single-layered surfel representation, reflectance samples for each surfel are collected into lumitexels and used for texture reconstruction according to the Phong reflectance model, which is then rendered and edited using Pointshop3D as described in Section 5.2.Īn example of such a system is the BBC-developed free-d system which uses circular retroreflective bar-coded markers on the studio ceiling ( Fig.
MONOCHROME CAMERA REGISTRATION
Some challenges of registration, such as error accumulation, play a minor role in this case because the rings are already consistent at high accuracy, leading to a simpler registration problem. The current system uses ICP for registration. Some methods for mutual registration of reconstructions are geometry based, such as iterated closest points (ICP) described in Section 3.1.4, and others are image based, such as that by Bernardini et al. Therefore, the resulting rings have to be mutually registered, treating each ring as a separate rigid model. The optimal solution would be to perform these transformations again around a calibrated axis.īut since this tends to be mechanically demanding, the object is tilted by hand and each time a ring of reconstructions is acquired by rotation of the turntable. This is accomplished by either tilting the object, or by moving projector and camera to a new position between the scans. Often several rings of reconstructions are needed in order to get a complete reconstruction. ( b) Local range = 1 and minimal distance = 2 mm (18,452 surfels removed). For similar reasons, it is also avoided to rectify the images and instead the lens distortion is modeled.įigure 3.14. This complicates calibration, smoothes the reconstruction, and can produce artifacts comparable to those in Figure 3.14 shown later. The reason is that cameras using a Bayer pattern for color acquisition (each pixel has a red, green, or blue filter attached) exhibit increased blur and artifacts due to undersampling and interpolation. It is preferred to use monochrome cameras and to illuminate the object (or calibration pattern) with red, green, and blue projector light in order to acquire colors.
Currently, an IEEE-1394 video camera with a resolution of uncompressed 1, 024 × 768 pixels is used to acquire the images. This becomes important if small objects have to be acquired. Another aspect is the projector's minimal image diagonal, or in other words, the minimal focal distance. This way the calibration is not lost when the projector is turned off between scans. It is beneficial if the projector allows for disabling features such as automatic synchronization and image size adaption. In our setup, the system contains an analog-input 1, 024 × 768 DLP video projector. Mid-infra-red cameras were also employed in order to recognize the stalk-ends and calyxes ( Cheng et al., 2003), but their high price prevents their use in commercial grading machines for the moment. The system had the potential for industrial application. (2003, 2004) selected four wavelength bands in the visible and NIR spectra and developed a four-band multispectral vision system dedicated to defect detection on apples.
Taking into account practical considerations, Kleynen et al. Since this imaging technique provides a large amount of data, which it takes a great deal of time to acquire and to process, it cannot be transferred to an industrial machine. (2004) have used a hyperspectral imaging system to detect apple surface defects.
As a guide, in the 1980s and earlier, monochrome cameras were used in the 1990s, color cameras were considered. The spectral sensitivity of the image acquisition devices and the number of “channels” acquired depend on the development of technology. Vincent Leemans, Olivier Kleynen, in Computer Vision Technology for Food Quality Evaluation, 2008 2.3 Image acquisition devices
0 kommentar(er)
