Study On The Detection System And Application Method Of Fruit Firmness Based On Acoustic Characteristics

Acoustic characteristic detection and analysis technology is widely applied in industry,medicine,voice and other fields with the advantages of simplicity,rapidity and non-destructive.In foreign countries,the acoustic characteristics detection technology was started earlier in the field of agricultural products quality detection,and its development speed was fast enough to reach the online detection stage.However,the domestic started relatively late in this respect,the development is slow,and and the detection of several kinds of fruits in the main agricultural products is carried out,and the product level acoustic characteristic detection and analysis device is almost zero.This study has developed a complete set of acoustic characteristics detection and analysis system,including all the hardware system design,the selection of key components,and the processing and assembly.Functional design of software system,program development of control system of upper and lower machines,acoustic characteristics detection,time-domain signal and frequency-domain signal transformation,spectrum preprocessing and qualitative classification of chemometrics methods of implementation.The establishment and prediction accuracy analysis of two kinds of four fruit hardness grading models were carried out,and it was proved that the system could be used for fruit hardness grading.The main study contents are as follows:1.The working principle and structure of the acoustic characteristic detection and analysis system are introduced and analyzed.The overall framework of the detection system was proposed,and the software and hardware design of the acoustic characteristic detection and analysis system was completed.The important key components were selected and analyzed,and the stepper motor with 17h40h-0404a2-21 is used as a percussion device for the acoustic characteristics detection and analysis system.Based on the comparison of several kinds of acoustic sensors,4015 acoustic sensor is used as the acquisition sensors for acoustic characteristic signals in the acoustic characteristic detection and analysis system.2.To make the acoustic characteristic detection system more stable and reliable,the key parameters of the detection system were optimized.On six kinds of 4015microphones with different sensitivity were studied and analyzed,the results showed that the stability of the acoustic characteristic detection and analysis system was better with sensitivity-42dB microphones,and the relative standard deviation of multiple detection frequency was less than 2.8%;With fifteen green apples as the research object,3 kinds of percussion head materials such as rubber,wood and plastic were designed,F₁,F₂,and F₃ three different percussion forces were designed,and the optimal test scheme of the fruit's sunny side,nightside,and flank,three different lateral striking positions is optimized.The experiment shows that the impact of plastic hitting on the spectrum signal is minimal,the standard deviation is only 2.2 Hz,and the maximum hitting force F₃ is used.The signal strength and stability are good.The conclusion of multi-point acquisition of acoustic characteristic signals is also adopted.3.Based on the hardware control and signal processing requirements of acoustic characteristic detection and analysis system,the overall framework of acoustic characteristic detection and analysis system was designed.In the Microsoft,NET platform of Microsoft,PC software was developed using the C#programming language,hardware control?parameter setting for controlling the knock process?,time domain acoustic signal acquisition,transformation of time-domain signal and frequency-domain signal,and data processing and other functional modules of the program development were implemented.In order to facilitate the experiment and acoustic signal analysis research,the functions of map display,preservation,and help were added.4.The process of establishing fruit hardness grade model by acoustic characteristic analysis technology was introduced.The effects of different spectrum pretreatment methods and four qualitative classification chemometrics methods on modeling precision were studied.Four groups of fruit samples such as Yantai Red Fuji Apple,Jinshuai Apple,Fengshui pear and Gongli were used as hardness classification objects.Through the actual modeling and analysis of 60 samples in 4 groups,60 samples were divided into a correction set and a prediction set by 7:3 using the K-S method.The time-domain signal was transformed into a spectrum signal by FFT.The spectral data preprocessed by SGF+normalization method is used for PCA feature extraction,in the calibration models established by Fisher,KNN,QDA,and LIBSVM,the experiments showed that the model established by QDA method has the best effect,the accuracy rate of the hardness grade of fruit of each variety is more than 95%,and the accuracy rate of the prediction set is more than 80%.However,the accuracy rate of the hardness grade model established by other methods is more than 82%.In addition,the accuracy of LIBSVM model is 70%,and the accuracy of other predictions is above 80%.The results showed that it is feasible to use the acoustic properties of fruit to detect the hardness of fruit and satisfy the fast test of fruit in practice.