Research On Effects Of Forchlorfenuron On The Physical And Chemical Properties Of Kiwifruits And Methods On Discrimination Of Treated Kiwifruits

China is one of the main countries of kiwifruit cultivation and export in the world, however, there is still a big gap between the fruit yield and quality control compared to the developed countries, which has a direct impact on the competitive capacity in the international market. Forchlorfenuron is one of plant growth regulator(commonly known as expansion agent) that is widely used in kiwifruit production. The application of forchlorfenuron increases the yield of kiwifruits, but brings a series of negative issues as well. With the improvement of living standards, people begin to pursue green healthy food, food safety has become the focus of attention. Although the use of expansion agent has been strictly prohibited, the phenomenon of abuse of expansion agent has never stopped due to the interests from it. Pursuit of green and healthy kiwifruit is the goal of consumers. However, it is unclear that effects of forchlorfenuron on the physical and chemical properties of kiwifruits, meanwhile, no reliable methods were reported for discrimination of kiwifruits treated with different concentrations of forchlorfenuron.Three varieties of kiwifruits, i.e., “Hayward”,“Xuxiang” and “Huayou” were treated with six forchlorfenuron concentration levels of 0, 2.5, 5.0, 7.5, 10.0, 12.5 mg/L 15 days after petal fall. In order to explored the effects of forchlorfenuron concentration on physical and chemical properties of kiwifruits, vertical and transverse diameters of selected kiwifruits of each variety during ripening, single fruit mass, fruit shape index, soluble solids content and firmness at postharvest were measured. Then near-infrared spectroscopy of “Xixuan” and “Xuxiang”, and hyperspectral images of “Hayward”, “Xixuan” and “Xuxaing” were acquired. The original spectra were preprocessed by standard normal variate(SNV), two sample division methods, i.e., Kennard-Stone(KS) and Duplex algorithm were compared. Successive projections algorithm(SPA) for classification was used to selected characteristic wavelengths, then based on full spectra(FS) and characteristic wavelengths, discrimination models of partial least squares discriminant analysis(PLS-DA), support vector machine(SVM), extreme learning machine(ELM) and random forests(RF) were established. By comparing the performance of different discrimination models, the optimal model for discrimination of treated kiwifruits was given. The main results are shown as follows:(1) The growth rates of vertical and transverse diameters were fast at early growth stage, and slow at the late growth stage. The maximum average vertical diameters were found at the forchlorfenuron concentration of 7.5 mg/L for all varieties. The maximums of the transverse diameters were found at the concentrations of 5.0, 7.5 and 12.5 mg/L for “Hayward”, “Xuxiang” and “Huayou” kiwifruits, respectively. When the concentration was 5.0 mg/L, “Huayou” and “Xuxiang” had the highest fruit shape index, while other five concentrations decreased fruit shape index of “Hayward”. Forchlorfenuron treatments also improved the single kiwifruit mass and soluble solids content, but decrease firmness. The kiwifruits treated with 7.5 mg/L forchlorfenuron solution provided kiwifruits the largest mass and soluble solids content and the smallest firmness.(2) Effects of sample division methods on the performance of discrimination models were investigated based on the NIR spectroscopy of “Xixuan” kiwifruits. Discrimination models were established based on sample sets divided by KS and Duplex algorithm, respectively. were compared. The results indicated that the accuracy rates in the test set of the former were higher than the latter.(3) In the process of using NIR spectroscopy to discriminate treated kiwifruits, the numbers of characteristic wavelengths selected by SPA for “Xixuan” and “Xuxiang” were 1.40% and 1.35% of the number of full spectral wavelengths. SPA simplified the structure and improved the performance of discrimination models. The best models for treated “Xixuan” and “Xuxiang” discrimination were FS-PLS-DA and SPA-PLS-DA, respectively, both the accuracy rates of test set of the two models were 100%.(4) In the process of using hyperspectral imaging technology to discriminate treated kiwifruits, FS-PLS-DA was the best discrimination model for “Hayward” kiwifruits, the accuracy rates of test set were 97.4% and 100%, respectively. The numbers of characteristic wavelengths selected by SPA for “Xixuan” and “Xuxiang” were 11.76% and 12.22% of the number of full spectral wavelengths, and the optimal models were SPA-PLS-DA and SPA-RF, respectively. The accuracy rates of the teo models of test set were 100%.(5) This study indicated that both NIR spectroscopy and hyperspectral imaging technology can realize nondestruction discrimination of treated kiwifruits. Copmared to NIR spectroscopy, models based hyperspectral imaging technology obtained higher accuracy rates, and is more capable to identify treated kiwifruits.