Design, test and implementation of a Multispectral-image Reconstruction System Based on a 2D Optical Scanner, a multiwavelength LED-based illuminator, and a compact spectrometer. Andrés R. Vega-Pérez1, Hugo A. Banda-Gamboa2 and César Costa-Vera3,4 1
Department of Automation and Industrial Control, Escuela Politécnica Nacional. 2 Department of Computer Science, Escuela Politécnica Nacional. 3 Department of Physics, Escuela Politécnica Nacional, Ladrón de Guevara E11-253, Quito – Ecuador. 4 Grupo Ecuatoriano para el Estudio Experimental y Teorico de Nanosistemas –GETNano–, Diego de Robles y Via Interoceanica, USFQ, N104E,Quito – Ecuador. E-mail:
[email protected]
Abstract: A 2D+1 scanning imaging spectroscopy system is presented, its performance is tested with patrimonial artwork and biological samples. The device can obtain A5 size sample images with up to 529 pixels per cm2 resolution. OCIS codes: (110.0110) Imaging systems; (100.0100) Image processing.
1. Mechanical and optical instrumentation The 2-D scanner consists of two stepper motors that move horizontally a fiber optic and an illumination head over the sample. The scanner is able to obtain 428µm pixel-side measurements and reaches an effective scanning area of 13cm x 12.5cm. The illumination system is based on ten LEDs that emit light in different wavelengths (375nm - 910nm). The emitted light passes through an opal diffuser (5mm diameter) and then through a plano-convex lens (2cm diameter and 3cm focal length) [1-3]. The supporting structure for the LEDs and optical parts were designed in Solidworks©, and then 3-D printed. The fiber optic is coupled to a structure with a collimating lens (Ocean Optics 74-UV). After, the diffusely reflected light and fluorescence is sent to an USB-2000 Ocean Optics spectrometer (Range: 200 −1100nm). Typically, the illumination is made at 45° on the sample and the fiber optic is pointed perpendicular to the sample [4]. In addition, the plano-convex lens to focus the illumination is attached to a focusing system from an old digital camera. With this system the distance of the lens to the sample is adjusted to concentrate the illumination in a small spot. 2. Electronics and software The control and data acquisition is made in an application based in Matlab 2012b© and with dedicated homemade electronic circuitry. Matlab© permits controlling peripherals, in this case, the Atmel XMEGA128D3 microcontroller and the Ocean Optics spectrometer using the USB port. It can also process signals and the environment permitted implementing a suitable Graphical User Interface (GUI). The whole equipment runs with an 110V@60Hz power input. In order to compensate for variations in temperature and to obtain reproducibility of the multiespectral images, all the LEDs are driven by constant current supplies. These supplies and the corresponding circuitry were designed and tailor-made for this project. With this, the intensity of the LEDs is controlled by a PWM signal from a microprocessor with a 0.25% resolution. The spectra acquired by the spectrometer is brought into the Matlab environment to be digitally processed using both a 2nd order Butterworth filter to reduce the noise and a Savitzky-Golay filter [5], to smooth the final spectra. For each pixel, every LED is turned on for a short period of time and reflectance spectra is recorded. A ratio of the total light recorded in each case with respect to a white standard reference (PTFE) is calculated. By this process, a 3-D matrix is formed with values ranging from 0 to 1 corresponding to these ratios. The matrix dimensions are {i, j, 11}, where i, j represent the number of pixels of the sample, and for each of these pixels there are eleven values corresponding to the calculated ratio. This ratio is used to build the multiespectral images assigning one 2-D matrix to each image component in Matlab (RGB). In order to obtain an image in the visible range the reflection spectra for different LEDs can be combined. For
instance a RGB image can be built with: Red for 625nm, Green for 525nm, and Blue for 475nm. Although, any arbitrary combination may result in an image whose apparent color does not match the original sample as seen by a regular camera or by the naked eye, there are some features that this “false color” configuration may help enhancing. This is useful for the understanding of the sample’s composition and properties. The system permits reconstructing a 3-D image using in the Z-axis one of the 2-D matrix, so it is possible to analyze the sample in up to four LED responses; three in color, and one spatial. The process of scanning and acquiring the data is completely automatic; the user just needs to set the scanning limits, the pixel-side size, and the spectrometer’s integration time. The reconstructed image is shown on the computer screen as the system scans the sample.
Fig. 1. Photographs of the entire system analyzing a segment of an artwork.
3. Results The multispectral decomposition of a patrimonial painting is shown in Fig. 2, also a reconstructive mix combining three LED responses is presented. Others are also possible to make. The RGB combination mentioned above permits the user see a realistic visible image in the GUI, in addition a fourth LED response can be shown in the Z-axis, resulting in a 2D+1 image. This image is a segment of a painting from the early 19th century called “The Resurrection of Christ” (anonymous artist), property of the Instituto Nacional de Patrimonio Cultural of Ecuador Red: 625nm + Green: 525nm + Blue: 475nm
Photograph of a segment of the painting
Red: 625nm
Green: 525nm
Blue: 475nm
Red: 625nm + Blue: 475nm
Red: 625nm + Green: 525nm
Green:525nm + Blue: 475nm
Z-axis: 760nm
Fig. 2. “The Resurrection of Christ” (scanned area limited by the orange rectangle: 5cm 2, resolution: 47x58 pixels).
Natural plant leaves have unique properties outside the visible spectrum [6]. As could be expected, there is no clear distinction between the rim and the center in the image from the NIR LED (Fig. 3), whereas in the visible, the whiter contour is clearly separated from the green area.
Fig. 3. Multiespectral response of a bicolor leaf.
Natural green leaves are relatively transparent in a particular NIR range which includes our 810nm LED. To test this, a small drawing on a paper piece was placed under a leaf, as seen in Fig. 4. The drawing was hidden in the visible spectrum response but was clearly detectable in the NIR images.
Photograph
625nm
700nm
760nm
810nm
Fig. 4. Transparency of a leaf in the Near Infrared (scanned area: 4cm2, resolution: 47x47 pixels). .
4. Conclusions A 2D+1 multispectral device has been constructed using low cost parts and construction processes. This system performance has been tested in different types of samples, including patrimonial artworks and biological samples, with good results. The LED illumination device demonstrates high efficiency for the applications indicated. This source can be modified in the future to include laser diodes, for instance, to enhance fluorescence. The resolution of the system if limited has been sufficient to obtain interesting results. However, future improvements on the device can enhance this feature, reduce the scanning time, and increase the objective area. 5. Acknowledgments We would like to thank Dr. Martha Romero at the Instituto Nacional de Patrimonio Cultural of Ecuador for letting us use one of their paintings to run some tests. 6. References [1] M. Brydegaard et al., “Versatile multispectral microscope based on light emitting diodes”, Rev. Sci. Instrum. 82 (2011). [2] M. Brydegaard et al., “Broad-band multispectral microscope for imaging transmission”, Am. J. Phys., Vol. 77, No. 2, p. 104 – 110 (2009). [3] A. Merdasa, Multispectral Microscopy with application to Malaria Detection (Division of Atomic Physics, Lund University, Sweden, 2010). [4] C. Bonifazzi et al., “A scanning device for VIS–NIR multispectral imaging of paintings”, J. Opt. A: Pure Appl. Opt. 10 064011 (2008). [5] W. Press et al., “Numerical Recipes in C: The Art of Scientific Computing”, Second ed., Cambridge University Press, p. 650 – 655 (1992). [6] S. Lenk et al., “Multispectral fluorescence and reflectance imaging at the leaf level and its possible applications”, Experimental Botany, Vol. 58, No. 4, p. 807–814 (2007).