Study On Quality Detection And Analysis Of Small White Apricot Based On Machine Vision And Spectral Technology

The small white apricot is one of the most popular local fruits in Xinjiang.The quality testing and grading technology of small white apricot is not mater enough.The backward production and processing technology will restrict the development of small white apricot industry.The small white apricot is still in the use of manual sorting.Comprehensive use of advanced technology in multiple disciplines,such as machine vision technology,near infrared spectroscopy technology,it was coducted to develop the rapid detection method of small white apricot quality.By acquiring the image information and spectral information of the small white apricot,a fast quality detection model of the small white apricot was established.It has theoretical guiding significance for promoting the rapid detection for small white apricot,and has theoretical support for further research and development of rapid nondestructive testing equipment for quality of small white apricot.The main research contents and conclusions are as follows.(1)Research on fruit weight prediction of small white apricot based on machine vision technology.The RGB image of small white apricot was obtained by constructing machine vision system.The collected sample images will be converted from RGB model to HSV model.The saturation channel of HSV color model was selected,and the maximum inter-class variance method was used for threshold segmentation of S-component graph.The binary image was obtained after the threshold processing.Canny operator edge detection technology was used to extract the edge of the target region.Extract the geometric feature information of the target region in the preprocessed image of small white apricot.Using the area,perimeter,long axis,short axis and long axis*short axis of the target region in the extracted image of the small white apricot,the unary linear regression and multiple linear regression models of the fruit weight prediction of the small white apricot were established.The results shown that the multiple linear regression model,which based on the geometric features of the target area extracted from the image of the small white apricot has the best effect on predicting.To predict fruit weight of the small white apricot,and its coefficient of determination is 0.977.The results show that machine vision technology can be used to predict the fruit weight of small white apricot.(2)Research on small white apricot color and size classification based on machine vision.In order to solve the problem of low efficiency of artificial classification of small white apricot,the fast classification method is studied by using machine vision technology.The machine vision system of small white apricot was set up to obtain the sample image.The first,second and third moments of RGB color space,the first and second moments of HSV color space,and the smallest circumscribed rectangle of target area were extracted by MATLAB software programming.Combined with probabilistic neural network method,the classification prediction model of small white apricot was established.Sixteen features of color and geometry were used to construct vector,and the classification accuracy is 94.64%.Eight features with the highest correlation were used to construct the feature vector,and the classification accuracy of each grade of small white apricot reached more than 90%.The results show that it is feasible to use machine vision technology to extract the color and geometric features of small white apricot.(3)Research on prediction of soluble solids of small white apricot by near infrared spectral.near-infrared diffuse reflectance spectroscopy was used to collect spectral data of small white apricot in the spectral range of 230-1069nm.Four kinds of preprocessing methods,including first derivative,second derivative,multiple scattering correction and standard normal variable transformation,were used to preprocess the collected near-infrared spectra date of small white apricot.The spectral date of 230-649nm,the spectral date of 649-1069nm,and the full spectral date of 230-1069nm were used for prediction modeling combined with partial least squares and least squares support vector machine for the sample soluble solids of small white apricot.Compared with different modeling results,the results shown that the PLS model established in the full spectrum range of 230-1069nm has the best effect in predicting the soluble solids of the test samples with the first derivative preprocessing of the near-infrared diffuse reflectance spectral data of the test samples.The R,RMSEC,R_test and RMSEP were 0.896,1.654,0.973and 2.054,respectively.The results show that the detection of soluble solids of small white apricot can be predicted by near-infrared spectroscopy.