| Multispectral images can offer valuable information on the reflectance distribution of an object’s surface across multiple wavelengths,which can aid in enhancing the performance of systems involved in various tasks,such as object classification,detection,and identification.In this thesis,we present a novel multispectral imaging method that is based on LED time-sharing strobe.Using this imaging principle,we have designed and built two multispectral imaging systems,one at the macroscopic level and the other at the microscopic level,both of which are compact and highly portable.Compared to existing multispectral imaging systems,the one constructed in this thesis offers several advantages,including low cost,compact size,and ease of portability.These features make it highly suitable for use in environmental monitoring,biological research,geological exploration,and other fields,thus demonstrating its significant potential and value.In particular,this multispectral imaging system utilizes the narrow-band power LEDs as its light source,and is paired with a camera that is sensitive to the wavelengths in the operating spectral range.The system is designed to acquire spectral data at various wavelengths and to analyze and process multispectral images using image processing techniques.The main research contents and innovations are as follows:(1)At the macro level,we develop a lightweight and compact multispectral imaging system based on pulse modulation technology.The system is capable of acquiring 9-band images(365nm,390 nm,460nm,515 nm,585nm,620 nm,650nm,730 nm,840nm),and the imaging conditions such as camera shooting time,distance,and light intensity are optimized using two image quality evaluation indicators namely sharpness and blurriness.(2)At the microscopic level,we develop a multispectral imaging system based on pulse modulation technology.The system is capable of acquiring 7-band microscopic images(390nm,460 nm,515nm,585 nm,620nm,650 nm,730nm),and we investigate the impact of different imaging conditions such as objective magnification and light intensity on the quality of the multispectral images using two image quality evaluation indicators namely sharpness and blurriness.(3)Two applications are carried out at two scales,macro and micro,respectively: a)Application of the macromultispectral imaging system to fingerprint image recognition tasks.Experimental results show that the system performs better than the RGB imaging system in the fingerprint recognition task,and its recognition accuracy is improved by up to 10.5%.b)Application of the microscopic multispectral imaging system for thyroid cancer pathology analysis.The experimental results show that the reflectance of cancer cells and normal cells show a large difference in the microscopic multispectral images The maximum reflectance difference between cancer cells and normal tissue cells reaches9.2% and 4.2% on stained and unstained slides,respectively.The system is therefore expected to be used in the future for the microscopic pathological analysis of cancerous and normal cells. |
