Research On Application Of Detection Algorithm For Ketone Gas Based On HSI Color Space

Ketone body level in human body is of great significance in clinical practice,especially for patients with diabetes,which can provide effective diagnostic basis for doctors and can promptly curb the further deterioration of the disease,therefore,long-term and frequent monitoring of ketone body level is necessary.Along with the continued development of POCT technology and home application of medical instruments at present,there is an urgent need for a noninvasive,intuitive and easy method of ketone body detection.As an important indicator of ketone body in human body,acetone from exhalation brings new ideas for the noninvasive detection of ketone body.Based on the detection requirements such as non-invasive,intuitive and easy,etc.,this paper summarizes and compares the existing detection technology for ketone body,and designs a detection system for acetone from exhalation based on HSI color space by combing with color response mechanism of colorimetric sensor array and the wide application of color space in analytical chemistry,and conducts further study mainly for the contents of the system such as feature extraction,feature analysis,etc.,and the main research content includes the following points:(1)Based on the color cross-response mechanism of the colorimetric sensor array,by combining the basic requirements of the analytical chemistry instruments based on color information at present,the overall design of detection system for acetone from exhalation is achieved.(2)By combining the characteristics of the colorimetric sensor array image and the main noise source,an image feature extraction algorithm based on HSI color space is designed to efficiently and accurately realize the extraction of the color information of the colorimetric sensor array.In the feature extraction process of whole image,a FCM image segmentation algorithm based on H&I weighting component is innovatively proposed on the basis of conventional Fuzzy C-means(FCM)algorithm.Compared with the commonly used Otsu algorithm and the conventional FCM image segmentation algorithm,the segmentation algorithm proposed in this paper shows 96.54% of the optimal segmentation accuracy at different types images of array points.(3)Based on the development environment of Qt and OpenCV,according to the design process of algorithm,the image feature extraction algorithm designed in this paper is achieved,and the transformation from algorithm design to practical engineering application is completed.(4)By taking the detection system designed in this paper as the experiment platform,the detection of acetone gas of different concentrations is achieved.The dynamic response analysis,hierarchical clustering analysis and BP neural network identification,etc.are carried out in connection with HSI difference characteristics of acetone gas of various concentrations.Through the above characteristic analysis,it can be concluded that the concentration of acetone gas has a nonlinear relationship with the HSI characteristic response of the colorimetric sensor array,and BP neural network can achieve the HSI difference characteristics recognition,with the overall relative error of 6.67% as the recognition result.Besides,this paper also implements initial exploration in the relationship between the color information obtained from detection system and spectral information.