Quality Analysis Of Xinjiang Flat Peach Based On Machine Vision Image And Near Infrared Spectroscopy

As a kind of characteristic fruit in Xinjiang,picking Xinjiang flat peach has a relatively backward technology,which mainly is artificial and low efficiency for sorting.It has a great influence on the competitiveness of flat peach in the international commercialized import and export trade.In order to realize the true value of flat peach by analyzing the their quality of flat peach,this paper studied the prediction of peach quality,size,sugar content and hardness by using machine vision image technology and near infrared spectroscopy technology.The research contents and conclusions are as follows,(1)To built Experimental equipment for acquiring image of peach.It was introduced of the production status and characteristics of flat peach,the equipments or platform for collecting RGB image information and spectral information of flat peach were built on the base of characteristics of flat peach and sensors of imaging and lightness.Finally,some methods and theories for studying the quality of fruit and exterior were introduced.(2)To predict flat peach quality and size based on image information.The image of flat peach is pretreated by clipping and median filtering.And then the threshold was used to segment peach reagion from background to get flat peach binary image.The multivariate linear regression model was established by using area X,perimeter P,long axis L and short axis S.The correlation coefficient of the quality prediction is 0.9877,as the relative error of the forecast set is 0.0291.The linear discriminant analysis method is used to predict the size of the green flat peach,and the accuracy rate of total score is 91.6%.The correlation coefficient of red flat peach quality prediction is 0.9931,as the relative error of forecast set is 0.0104,the analysis of red flat peach size is analyzed by linear discriminant analysis,and the accuracy rate of total score is 90%.(3)To predict flat peach sugar content and hardness based on the near-infrared spectral information.Using the Mahalanobis distance,it was removed for the abnormal samples of near-infrared spectral of flat peach,and the abnormal samples of total sugar of flat peach was eliminated by the method of concentration residuals.The spectral data of flat peach were preprocessed by means of standard normal variable transformation,moving window smoothing,multi-dimensional scattering correction,first order derivative processing,second derivative processing,data center,trend algorithm,data standardization.By using different characteristic wavelength screening methods,the paper finally establishes a predictive model for the sugar content and hardness of flat peach.The results shown that the effect of the pretreatment of flat peach by moving window smoothing is better way to remove noise and peak spectral,and the characteristic wavelength is screened by genetic-partial least squares method,and the prediction model is finally established.Modeling results shown that the prediction of the sugar content of green flat peach is 0.8526,the correlation coefficient of hardness prediction set is 0.8041,the correlation coefficient of red flat peach sugar is 0.8556,and the correlation coefficient of hardness prediction set is 0.9012.