Research On The Non-destructive Detection Technology Of Red Grape Quality

The detection of Red Grape internal and external quality is directly related to the processing utilization rate and growth rate. In recent years, the machine vision technique and the near infrared spectroscopy technique have been widely applied to outer quality detection and gradation of farm products. Nevertheless, little literature o research on the size detection of farm products in bunches can be found. The internal and external quality of Red Grape in bunches needs to be further studied. In this study,to detect the external quality of Red Grape by machine vision and image processing technology, the Red Grape varieties of Red Grapes was used as research object.Detecting the internal quality of Red Grape based on the machine vision technology and near infrared spectroscopy technology. The main research contents and results were as follows:(1) The image acquisition system was set to capture the images of Red Grapes.The near-infrared light and fluorescent light were selected, and the 2-CCD industrial camera AD-080 GE with an industrial lens were selected to capture the image of Red Grapes. The camera can collect the RGB image and NIR images at the same time.(2) The image preprocessing method was determined. Compare median filter with mean filter methods, where the effect of the median filter is better than average filter effect, and median filtering was selected as the image pre-processing method to remove noise which was caused by the acquisition process. Contrast Roberts operator,Canny operator, Sobel operator, Log operator and other edge detection operator, Log operator determined by the Red Grapes to extract the grain edge information.(3) The detection methods and models of Red Grape size were determined.Firstly, Red Grapes' central location was extract by mathematical morphology method.Secondly, Red Grapes' edges was detected by edge detecting methods. Thirdly, a single edge was intercepted through central location and edges. Fourthly, edges was split by iterative least-squares fit methods, while, a single central location was divided by according to a single central location, Red Grape edge, and then use the value in iterative least-squares fit to split edges, and two principles, namely maximum 15%downgrade principles and the principle of proportionality, were established RedGrapes size classification model. In this study, 35 of 38 boxes were detected correctly,classification accuracy was 92.1 %.(4) The Red Grape color vision testing methods and models were established.Firstly, the RGB color image is converted to HSV color space, and the stems were removed by an appropriate threshold. Then, calculatting the red pixel, green pixel number and the percentage of red pixel rate. Lastly, Red Grapes size classification model was established by the percentage of red pixel rate. In this study, 36 of 38 boxes were detected correctly, classification accuracy was 94.7%.(5) The Red Grape sugar content visual detection methods and models were established. Ten texture information parameters of NIR images and 9 color information parameters of RGB images was selected, 10 higher degree of correlation of which was establish multiple linear regression model prediction of sugar content.In this model, the correlation coefficients among the training was 0.87, root mean square error of 0.97 and a correlation coefficient of 0.84 for the prediction set, the root mean square error of 0.82. In this study, 34 of 38 boxes were detected correctly, and the results were good.(6) Red Grape sugar content Near Infrared Spectroscopy method and model was determined. Firstly, correction to eliminate spectral pretreatment baselines were elimated correctly by multiple scattering. Secondly, 4 different spectral preprocessing methods, namely, Savitzky-Golay, Norris derivative filt, first derivative, and second derivative were used for removing the interference noise information. Thirdly, three kinds of quantitative analysis moedls was used to detect the Red Grapes' Brix base on full-band near-infrared spectroscopy. The model established in all band was analyzed by using different spectral pretreatments combined with three quantitative analysis models which were multiple linear regression(MLR), partial least squares regression(PLSR) and principal component regression(PCR). The results illustrated that the reliability of PLSR was the best and PCR followed by. In order to get a better prediction model, the number of wavebands was reduced from 1557 to 650 by using quantum genetic algorithm and partial least squares regression(QGA-PLSR). At the same time, the correlation coefficient(RC) of prediction model and its root-mean-square error of prediction(RMSEP) were improved obviously. RC wasincreased from 0.975 to 0.995, and RMSEP was decreased from 0.8 to 0.495.