Study Of AOTF-based Multispectral Imaging System And Its Application In Subcutaneous Veins Imaging

Subcutaneous veins imaging technique plays an important role in many different procedures such as biometric identification, auxiliary intravenous catheterization and blood oxygen saturation detection. In recent years, this technique has attracted increasing interest from research communities and industries. Transillumination imaging, photoacoustic imaging and single wavelength near-infrared imaging technique have been found suitable for subcutaneous veins imaging. Various novel methods such as multiple illuminants with different wavelengths, multispectral imaging and polarized light imaging have subsequently been introduced to enhance the contrast between subcutaneous veins and the adjacent skin tissues. In multispectral imaging of subcutaneous veins, radiation from a broadband source is reflected from the hand-dorsa and collected in dozens of spectral bands within the required spectral ranges, which results in a 3-D image cube that includes two spatial dimensions of high resolution and one spectral dimension. By combining both the spatial and spectral information, multispectral imaging effectively makes up the significant loss of spectral information in the traditional single wavelength near-infrared imaging, resulting in an higher quality subcutaneous veins image and a more flexible subsequent analysis and processing procedure. The multispectral imaging technique for subcutaneous veins imaging is gradually favored by domestic and foreign researchers in this field.With the consecutive support of the PetroChina Innovation Foundation (2009D-5006-01-06,2012D-5006-01-02) and the national science and technology support program (2011BAF02B00), an AOTF-based multispectral imaging analysis system was constructed and the application of multispectral imaging technique to hand-dorsa subcutaneous veins imaging was explored in this thesis. The main contents of this work include:(1) The application fields of research on subcutaneous veins imaging technique and several most commonly used imaging techniques were expounded. The current state of the art of subcutaneous veins near-infrared imaging techniques, as well as various methods introduced to enhance the contrast between subcutaneous veins and the surrounding skin tissues were emphatically summarized.(2) Some related theoretical bases of subcutaneous veins near-infrared imaging technique, including the anatomy structure of skin, the electromagnetic spectrum distribution, the basic interaction forms between light and biological tissues, the near-infrared optical window and the theory of light propagation in biological tissues were discussed.(3) The basic working principle of spectral imaging analysis technique along with various types of existing imaging spectrometers and their spatial and spectral scanning modes were systematically expounded. The design theories and performances of some novel form of spectral scanning devices such as acousto-optic tunable filter (AOTF) and liquid crystal tunable filter (LCTF) were mainly introduced, and their respective merits and drawbacks were also discussed.(4) An AOTF-based multispectral imaging analysis system for subcutaneous veins imaging was constructed. Special considerations were discussed in detail on how to select major components of the system. An AOTF RF driver applied to this system was developed. The direct digital synthesizer (DDS) techniques and the phase locked loop (PLL) techniques were combined for RF signal synthesis, the PLL was driven by the DDS to take advantages of both their merits of high frequency resolution, high frequency scan rate and strong spurious signals resistance capability. A hand-dorsa illumination system consisting of several geometrically arranged wide-bandwidth tungsten halogen lamps and a polycarbonate diffuser was designed, the highest illumination uniformity was achieved through dynamic adjustment. A spectral imaging analysis software with similar functionalities to some commercial ones was developed to collect and process the gathered multispectral images.(5) Through analyzing the system transfer model, a relatively systemic correction, calibration and characterization approach in both spatial and spectral domains for the AOTF-based multispectral imaging system was proposed. A series of tests were conducted to test the spectral accuracy and spectral resolution, measure the spatial resolution and the system DOF, evaluate the stability, SNR, spectral linearity, image distortion and image shift, validate the flat-field correction and spectral sensitivity correction of the multispectral imaging system. And thereafter, the procedure of gathering multispectral image of subcutaneous veins was summarized.(6) Several most commonly used image enhancement methods such as gray level transformation, gray level histogram and neighborhood processing were outlined, along with drawbacks of enhancement with just a single image. A series of multi-band image processing techniques such as color reconstruction and algebraic operations between multi-band images, as well as the feature space transformation of multispectral image data were simply introduced. All of these methods were then applied on the hand-dorsa multispectral images gathered by this AOTF-based system to improve the image for visual perception of humans and making further easy processing steps on the resultant images by machines.(7) The subcutaneous veins were classified automatically from the original hand-dorsa multispectral images through maximum likelihood classification (MLC) method, which achieved a satisfying result. The original high-dimensional multispectral images were then analyzed through principal component analysis (PCA) technique, dimensionality was reduced and high contrast images were created with the first 4 PCs. The MLC method based on these dimensionality-reduced images achieved a similar classification result compared to the result of the original 35-band images, what's more, the classification time was greatly shortened.