Research On The Acoustic Technique And Instrument Used In Assessment Of Ripeness And Hollow Heart Of Watermelons

China has the largest production and planting area for watermelons in the world,however,the export is not more than 1%annually.It results in low level of commercialization for watermelons due to outdated postharvest detection and process techniques.Watermelons have good taste if harvested at right growth stage.Excessive dependence on the experience of melon farmer leads to the watermelons with various quality in the market.With the ripeness of watermelons,texture trait varies as well.Mechanical and structural determination and ripeness assessment of watermelons will boost income for melon farmers and dealer,meet the requirement of consumers and expand the export overseas.'Qylin' watermelons were used as experimental samples in the thesis.The textural characteristic of watermelons was predicted by regression and ripeness classification and determination of hollow melons was conducted as well after acquiring acoustic feature using self-made acoustic device and parameter analyzing software.The main conclusions of the study include:(1)Study was carried out on the components of acoustic device and other factor which affects the acoustic spectrum of watermelons.It was indicated that the more the elastic modulus of hitting ball and fruit tray,the less effect on acoustic spectrum.Hitting force has no influence on the spectrum.It should be better collect acoustic signal from multiple points along equator of watermelon with only one hitting for each point to get acoustic characteristic precisely.(2)Method of short-term double thresholds,used for endpoint detection of acoustic signal,was combined with linear predictive residual method and digital filtering technique to further remove noise and burr and enhance ratio of signal to noise,which increased about 47%.(3)Index of the first-order moment MI1 and index of the second-order moment MI2 are proposed to solve the problem of resonant frequency splitting into two peaks with approximately the same amplitude and position.Combining with traditional parameters f2m2/3 and f2m,these two parameters were used to conduct regressive analysis with slope of the curve of force-deformation through method of linear regression,polynominal regression,SMR,PCR and GA-ANN.SMR model was proved to have the best performance.(4)Methods of LDA,KNN,ANN and LS-SVM were applied to classifying watermelons into unripe,ripe and overripe group with acoustic characteristics MI1 and MI2.Results showed that best classifier was built up by LS-SVM.(5)Methods of LDA,KNN,ANN,LS-SVM and experienced threshold were employed with parameters MI1,MI2 and energetic ratio Er to determine hollow watermelons from good ones in the unbalanced distributing samples.Results showed that the best classifier was constructed by LS-SVM.The innovations of the thesis are as follows:(1)Linear predicative residual algorithm,combined with method of short-term double thresholds endpoint detection and digit filtering technique,was used to eliminate noise and burr in acoustic signal,keeping key characteristics such as resonant frequency and increasing ratio of signal to noise as much as possible as well.(2)New acoustic parameters of index of the first-order moment MI1 and index of the second-order moment MI2 were proposed to solve the problem of resonant peak splitting into two closed peaks,which makes it hard to extract acoustic features from spectrum.(3)Acoustic parameter energetic ratio Er,combined with experienced threshold method,was used for hollow watermelons determination,which has advantages of simple principle,small computation amount and implementation on computer or programmable hardware easily.