Development And Application Research Of Portable Spatial Frequency Domain Imaging Intelligent All-in-One Device

Spatial Frequency Domain Imaging(SFDI)is a non-contact,wide-field optical characterization technique that can quantify the distribution of tissue optical properties in the form of images.By analyzing and studying the optical properties,the chemical composition and physical structure of the tissue can be understood for non-destructive testing.SFDI technology is therefore widely used in medical testing and surgical guidance,but is relatively less used in the field of agricultural products and food safety.Traditional SFDI systems are large and difficult to move,which is not conducive to field inspection in various complex scenarios.At the same time,the SFDI technique customarily uses a least-squares based method to fit the tissue optical properties parameters pixel by pixel,a process that is not only time inefficient but also needs to be improved in terms of accuracy.In this study,a portable SFDI intelligent all-in-one system was constructed,which can achieve portable,automatic,rapid and accurate measurement of tissue optical properties.The accuracy of the SFDI system was verified based on liquid mimic experiments,and the verified SFDI system was used to detect the change pattern of optical properties of pear bruised tissues.In order to measure the tissue optical properties quickly and accurately,a model that can map the tissue optical property parameters quickly and accurately was developed based on the Monte Carlo(MC)forward model and machine learning techniques.The main findings and conclusions of this study are as follows:In this study,a portable SFDI intelligent all-in-one system was designed and developed with micro-components and an embedded control system to enable portable and automated measurement of tissue optical property parameters.The study determines the system’s positional attitude by projector-camera system calibration;corrects the world coordinate system structured light modulation frequency by camera calibration;ensures the sinusoidally modulated structured light is close to the theoretical sinusoid by projector linearity correction;and corrects the medium modulation transfer function by a set of standard diffuse reflectance white plates.System calibration and alignment improve the linearity of the system and the prediction accuracy of the optical characteristic parameters.The study proceeded to carry out accuracy verification experiments and application experiments of the portable SFDI intelligent all-in-one system.The results of the liquid mimic-based validation experiments showed that the errors of the absorption coefficient and the approximate scattering coefficient were 3.5% and 5.6%,respectively,at 525 nm wavelength.By comparing the results with those reported in the existing literature,it can be seen that the developed SFDI system has a good measurement accuracy,which proves the feasibility and accuracy of the system scheme.The changes of absorption coefficient and approximate scattering coefficient with time during the formation of bruised tissues in pears were investigated,and the intrinsic connection between the optical properties of pears and the formation of bruised tissues was analyzed.The results showed that the parameters of optical properties of bruised tissues and the surrounding normal tissues have obvious differences,which can be used for the discrimination of fruit bruises,and the feasibility and effectiveness of the developed system for the detection of actual samples were verified.To address the problem of time-consuming and insufficient accuracy of the inversion process of SFDI technology,this study also conducted experiments on the inversion of tissue optical properties based on machine learning algorithms to improve the speed and accuracy of the inversion process by taking advantage of machine learning technology.Based on the MC simulation forward model and various machine learning algorithms,a mapping model from the spatial frequency domain diffuse reflectance to the optical properties parameters is established.The results of the test set prediction and affine experiments show that the Long Short Term Memory Regressor(LSTMR)inverse mapping model has the most accurate prediction results;the test set absorption coefficient The test set absorption coefficient and the approximate scattering coefficient errors are 0.32% and 0.21%,respectively,and the results of the mimic experiments show that the LSTMR inversion method is 2-3 orders of magnitude faster than the traditional least squares fitting method.This study shows that the constructed portable SFDI intelligent all-in-one system can realize automatic,rapid and accurate detection of tissue optical characteristic parameters,and the portability and integration of the system are of great value and significance to advance the application of relevant analysis and detection based on tissue optical characteristic parameters.