JENCOLOR Innovation Forum 2012 in conjunction with the 14th SpectroNet Collaboration Forum. Prospects and Applications

Transcription

1 Forschungsinhalte wurden im Rahmen verschiedener Projekte erarbeitet, deren Finanzierung durch das BMBF (Projekte OpMiSen: Förderkennzeichen 16SV5575K und Qualimess: Förderkennzeichen 03IP709) sowie durch das BMWi (Projekt Autopetrographie: KF KM0) erfolgte. Inten nsity Monochrome Image JENCOLOR Innovation Forum 2012 in conjunction with the 14th SpectroNet Collaboration Forum RGB Colour Image Multispectral Imaging Prospects and Applications Department of Quality Assurance and Industrial Image Processing, Ilmenau University of Technology Multispectral Image /Correns 2011/ Hyperspectral Image wavelength 05. September Page 1

2 Agenda 1. Multispectral Image Processing Chances 2. Types of Spectral Images 3. Spectral Imaging Applications 4. Modular Filter Wheel Camera in the Department of Quality Assurance and Industrial Image Processing 5. Applications in Research of the Department Quality Assurance and Industrial Image Processing 6. References Page 2

3 1. Multispectral Image Processing Chances Why does the use of multi channel imaging make sense for metrology tasks? unknown colour edges are a problem for solving measurement tasks a higher spectral resolution of an image increase the chance of access a good contrast between different objects significantly Colour image Intensity image source: importance of multispectral imaging g grows with the increasing in the field of image processing applications, multispectral images provide substantially more spectral information than colour images (for example RGB images) Page 3

4 2. Types of Spectral Images Different types of spectral images in image processing: Monochrome Image RGB Colour Image tensity Int Multispectral Image Hyperspectral Image wavelength Source: /Correns 2011/ Page 4

5 2. Types of Spectral Images Definition iti of Multispectral t l / Hyperspectral Imaging: Methods which generate digital images with spatial and spectral resolution, The spectral range goes from UV to IR and depends on the given problem/task, The resulting data forms a data cube with two spatial axes x and y and a third axis which represented the spectral wavelength (see Cube display), For every pixel we get a individual spectrum (see Spectrum of one pixel). The hyperspectral cube contains the intensity of the pixel at the position (x, y) and the wavelength λ. Cube Display Imaging-Definitions accord. to /Ehlers 02/: Spectrum of one pixel Source: Short, Nicholas M.: AVIRIS and other Imaging Spectrometers, NASA, Ultraspectral: > 250 bands Hyperspectral: bands Multispectral: 2 20 bands Panchromatic 1 band Page 5

6 3. Spectral Imaging Applications Technology Colour imaging Multispectral imaging and Hyperspectral imaging Application object recognition, quality analysis of food and other industrial products, Specific measurement tasks remote sensing, quality analysis of food and other industrial products, biomedical analysis (detection of blood flow, skin cancer..) detection of minerals, monitoring of emissions in the environment Page 6

7 4. Modular Filter Wheel Camera in the Department of Quality Assurance and Industrial Image Processing A multispectral t l camera is designed d and developed d as modular Filter Wheel Camera with 12 colour channels (wavelength bands) 12 band pass interference filters in steps of 50 nm from 400 nm to 950 nm was used for capturing 12 channels CMOS image sensor with a resolution of 640x480 and a frame rate of 60 frames per second maximum filter wheel speed of 250 rpm C-Mount Standard for connection with camera lens /Preißler eße 2011/, /Preißler eße 2012/ Page 7

8 5. Applications in Research of the Department Quality Assurance and Industrial Image Processing Automatic ti Petrographic Inspection of Mineral laggregates Using Multispectral Image Processing sub-classes: Problem: 16 different mineral aggregate classes with very high intra-class variabilities Goal: & low inter-class variability! reach more information of the class specificity for higher discrimination power by using more spectral bands in VIS and NIR mistaken identity = problematic group 1 group Page 8

9 5. Applications in Research of the Department Quality Assurance and Industrial Image Processing 550 nm Automatic ti Petrographic Inspection of Mineral laggregates Using Multi Spectral Image Processing Used technology: 694 nm Filter Wheel Camera with 12 bands ( nm) 766 nm 852 nm 940 nm Rl Relevance-graded dwavelength Process: Capturing of different objects of phenotypical very wavelength [nm] Mean value of F-Score similar classes graywacke, other paleozoic sediments, sandstone and friable sandstone in different wavelength bands for determination of significant bands with highest discrimination power Investigation of colour and texture features in the different spectral bands by using Fisher-Score: intra class variance F = inter class variance Result: Relevance measure of different spectral bands /Garten 2012/ Page 9

10 5. Applications in Research of the Department Quality Assurance and Industrial Image Processing Analysis of Plastic Pipes Problem: Goal: measurement of various fiber-reinforced plastic pipes, low contrast between the different layers complicated measurement of the fiber layer Inspection and measurement of the integrated fiber layer by using contrast enhancement in further wavelength bands Used technology: Filter Wheel Camera with 12 bands ( nm) Process: Image capturing in 12 bands, searching bands with highest contrast between fiber layer and plastic layer, building new colour image with the 3 bands with highest contrast Page 10

11 5. Applications in Research of the Department Quality Assurance and Industrial Image Processing Analysis of Plastic Pipes Result: Fiber 3 channel image of the bands with highest contrast between fiber and plastic Original RGB image with low contrast Page 11

12 Thank you for your interest! 6. References /Correns 2011/ Correns, M.: Subpixelgenaue Kantenortsbestimmung in digitalen Mehrkanalbildern, dargestellt am Beispiel von Sensoren mit Bayer Pattern Color Filter Array, Dissertation, TU Ilmenau, 2011 /Ehlers 2002/ Ehlers, M., Janowsky, R.,Gähler, M.:New remote sensing concepts for environmental monitoring. In: Ehlers, M. (Ed.), Remote Sensing for Environmental Monitoring, GIS Applications, and Geology. Proc. SPIE, vol. 4545, pp Bellingham, WA, 2002 /Short/ Short, Nicholas M.: AVIRIS and other Imaging Spectrometers, t NASA, /Preißler 2011/ /Preißler 2012/ /Garten 2012/ Preißler, M., Rosenberger, M., Correns, M., Schellhorn, M., Linß, G.: Investigations on a Modular High Speed Multispectral Camera,, 20 th Imeko TC2 Symposium on Photonics in Measurement, Linz, Austria, May 2011 Preißler, M., Rosenberger, M., Linß, G.: Smart Multispectral Imager, IEEE Photonics Conference 2012/ Garten, D., Anding, K.: Ergebnisse aus dem Forschungsthema Autopetrographie Brehna Bauzert, Vortrag Arbeitskreistreffen Baustoffprüfstellen, Bauprodukte Überwachungs- und Zertifizierungsverband Ost e.v., Brehna, Germany, June Page 12