| The annual output of watermelon in China accounts for 60% of the global production,making China the largest watermelon producer.However,the export volume of watermelon in China only accounts for 1.2% of the total,which means it is not a large exporter of watermelon.The backward watermelon maturity testing technology and low standardization degree of watermelon industry are the main reasons for the weak competitiveness of watermelon export in China.Choosing the right time to harvest watermelons depends too much on the experience of farmers,which leads to the transportation and sale of watermelons of different maturity,thus increasing the transportation loss and reducing the quality of watermelon on the market.To address the problems of time cost,low accuracy and subjectivity in manual watermelon maturity detection methods,this thesis designed a portable watermelon maturity detection system with STM32 based on the acoustic characteristics of watermelon.Firstly,based on the recent study of watermelon maturity detection technology and non-destructive watermelon testing system,this thesis analyzed the advantages of using acoustic characteristics in watermelon maturity detection,and introduced watermelon maturity detection methods based on acoustic characteristics through pretreatment,feature extraction and classification.Aiming at the disadvantages of poor applicability and low accuracy of mainstream acoustic feature extraction methods,a short-time Fourier Transform(STFT)acoustic feature extraction method based on adaptive frequency band division was proposed.Based on the difference of signals in the frequency domain,this method adaptively divides multiple frequency bands with different bandwidths,and characterizes the frequency domain information of STFT by amplitude of frequency band.The STFT with adaptive frequency band division can display the time-frequency information of signals with less data amount.The maturity of watermelon is not only related to acoustic characteristics,but also closely related to the weight of watermelon.However,the acoustic characteristic matrix and weight characteristics are different in dimension and order of magnitude,thus making them not suitable to be processed with the classical classification algorithm directly.In order to improve the accuracy of watermelon maturity detection,the Information Fusion Knearest Neighbor(IKNN)algorithm was proposed.The fusion distance was obtained by combining acoustic features of samples and Euclidic distance of weight features with quality value.Besides,the fusion distance of K nearest neighbor samples was weighted by Gaussian function to solve the problem that KNN was susceptible to the unbalanced sample number interference.Then,according to the main functions of watermelon maturity detection system,this thesis presents a portable system hardware solution with STM32F407 as the control core.The system is divided into five parts: data acquisition unit,information processing and management unit,knocking unit,photoelectric unit and power unit.The circuit design of each part is introduced in detail.The software system is designed in the MDK5 software development environment.The software design scheme of the system is described through the main program flow,system initialization flow,data processing module and system management module.Besides,a low-power software design of the system is carried out to further improve the system endurance.Finally,experiments are conducted based on 73 collected Kirin watermelon samples,and the results show that the combined model of STFT and IKNN classifier with adaptive band division proposed in this thesis can detect watermelon maturity with 97% accuracy,which is higher than many other feature extraction methods and classification methods.What’s more,the model has better capability to avoid the impact of interference after weighting on extracted features and better real-time performance using the fast IKNN classification algorithm.The watermelon maturity detection system proposed in this thesis is user-friendly,and has high detection accuracy,portability,and certain practicality. |
PDF Full Text Request