The Research On Ker Technology Of Hyperspectral Imaging System Based On LCTF

The technology of hyperspectral imaging is a combination of traditional imaging and spectroscopy technology,which is widely used in space remote sensing and measurement,food testing,environmental monitoring,biomedical and other fields.As a new and non-contact optical diagnosis technology,hyperspectral imaging provides an effective means of diagnosis through the spectral image information for clinical medicine,which has great potential for development.In this paper,a microscopic hyperspectral imaging system based on liquid crystal tunable filter is designed to obtain the microscopic hyperspectral images of biological tissue samples in the visible range,human diseases can be rapidly detected through spectral differences between normal cells and diseased cells.This paper firstly introduces the principles and theories of hyperspectral biological tissue imaging;the working principle of liquid crystal tunable filter is analyzed and simulated;the spectral image preprocessing,feature extraction and classification and recognition methods are studied;Finally,the relevant experimental research is carried out.The main contents are as follows:1.The basic principles of hyperspectral imaging and microscopic imaging are briefly introduced.The interaction between light and biological tissue and the optical diagnosis mechanism of biological tissue are discussed.The basic structure and working principle of LCTF are analyzed,and the spectral transmittance characteristics of the filter are simulated.2.The micrographic imaging system based on LCTF is designed and built.The selection and performance parameters of the main devices are introduced in detail,and the comprehensive performance of the system is analyzed.The hardware system of the various parts of the software platform and the corresponding function is described detailly.3.The preprocessing method,feature extraction and classification of spectral image are studied.In this paper,the adaptive median filter and joint calibration algorithm are introduced for the salt and pepper noise of the system and the influence caused by the system light source,lens and CCD.Aiming at the problem of large amount of hyperspectral image data,a feature extraction method including spectral analysis of principal component analysis and wavelet transform is studied.The classification and recognition methods of spectral images are introduced.The minimum distance,KNN and BP neural network algorithms are studied.4.Experimental study of lung cancer tissue slices is carried out.The classification accuracy of the algorithm under different feature extraction methods,including minimum distance,KNN and BP neural network is analyzed and compared.According to the related experimental results,the microscopic hyperspectral assistant diagnosis system is designed to realize the one-button recognition of cancer cells and normal cells,and the accuracy rate is over 90%.Through the above research shows that the application of microscopic hyperspectral imaging system to the diagnosis of lung cancer and other diseases is feasible,with significant clinical value.