Font Size: a A A

Hyperspectral Microimaging Of Skin Cancer Tissue And Image Processing

Posted on:2019-06-27Degree:MasterType:Thesis
Country:ChinaCandidate:L ShiFull Text:PDF
GTID:2404330575473610Subject:Optical Engineering
Abstract/Summary:PDF Full Text Request
Acousto-optic tunable filter based on TeO2 crystal is a kind of filtering device,and it has been widely used in many fields,such as archaeology,art protection,vegetation water resources control,food quality and safety control,forensic medicine,crime scene detection and biomedicine.Hyperspectral imaging is a frontier field of current medical detection technology.It uses a wide range of very narrow electromagnetic waves to extract valuable information.The research of spectral imaging is relatively late in our country.At present,the domestic research is mainly applied in the direction of remote sensing.The detection,diagnosis and operation of diseases are still in the theoretical and experimental stage.Firstly,the working principle of acoustooptic tunable filter is understood through the idea of sound and light interaction.By combining the acousto-optic tunable filter with the microscope,a hyperspectral microimaging experiment system was built.We used skin cancer tissue as a research object,and obtained experimental hyperspectral microscopic images of skin cancer tissues in 81 different light bands.Because of the high dimension of spectral image data and the characteristics of large data,this gives more information to the description of skin cancer tissues.In the field of image space,the correlation of hyperspectral image of skin cancer tissue is analyzed,and the dimension reduction and RGB image fusion of hyperspectral image are carried out.In the domain of spectral space,the transmission spectrum of each pixel of two-dimensional spatial image can be obtained by constructing the spectral cube of the image.Further,it is found that when the spectrum of the lesion region is known,the potential lesion area can be detected by the spectrum angle matching formula,and the quantitative,fixed point and location detection of the lesion area can be realized.Finally,the classification of hyperspectral images is studied by using support vector machines and BP neural network machine learning methods.
Keywords/Search Tags:hyperspectral microscopic imaging, image processing, spectral matching, Machine Learning
PDF Full Text Request
Related items