玉米种子净度超声波检测方法及装置研究

发布时间：2018-03-28 21:46

本文选题：空气耦合超声波　切入点：玉米种子净度　出处：《中国农业大学》2016年博士论文

【摘要】：我国在玉米种子净度加工处理过程中,对于虫蛀孔洞颗粒和霉变颗粒的筛选精度还不够,导致加工后的种子发芽率低,仍然无法满足“单粒播种”的需求,且如果霉变玉米进入食物链则会严重危害人、畜健康。针对上述问题,在总结国内外研究成果的基础上,提出基于空气耦合超声波无损检测技术、信号处理方法、计算机技术和电子技术的玉米种子净度检测方法及装置研究。主要研究内容和成果包括:(1)研究了超声波检测玉米种子净度原理和方法。首先采用有无经过超声波照射的玉米种子做了对比性发芽试验,验证了超声波检测玉米种子的安全性。然后分析了玉米虫蛀和霉变的超声波响应特征,即检测原理。随之分析了玉米颗粒的力学特性,根据其物理量计算出了玉米颗粒的纵波声速,并根据声速和检测精度需求设定了激励信号频率。最后针对实验对检测精度和获取高信噪比信号的需求,集成优化了超声波信号采集设备。(2)提出了基于空气耦合超声波技术的玉米虫蛀种子检测方法。对在室温、空气中采集的玉米种子超声波信号进行了去噪和特征提取,采用玉米颗粒单面和双面数据建立了KNN、DT、SIMCA和LDA等四种识别模型。结果验证,种子的摆放方位不影响检测,且所有模型对两类玉米的正确识别率都在85.00%以上。其中,KNN模型性能最佳,识别率高且稳定,对所有数据识别率均高于97.00%。研究结果验证了采用空气耦合超声波技术检测玉米虫蛀颗粒的可行性。(3)提出了基于空气耦合超声波技术的玉米霉变颗粒检测方法。首先对霉变与完好种子信号数据做了Hilbert-Huang变换,提取了频率段为300-498Hz的Hilbert边际谱作为两类种子的分类特征,建立了BPNN、SVM、SIMCA和KNN4种分类识别模型,并对模型性能进行了测试。结果显示,对霉变和完好颗粒的识别效果最好的是BPNN模型,识别率均达到了80.00%以上,特别是隐含层神经元个数为6时,识别率分别达到了85.71%、91.43%;SVM模型次之,识别率分别为80.00%、85.71%。结果表明采用空气耦合超声波技术检测玉米霉变颗粒是可行的。(4)研制了基于空气耦合超声波技术的玉米种子净度检测装置。根据实际的检测需求集成了玉米种子空气耦合超声波信号采集设备,同时设计实现了软件检测系统。设计进行了检测系统的准确性、包容性、稳定性和运行速率测试实验,测试结果表明,系统对各品种和类别玉米颗粒的识别率均在80%以上,且实验结果较稳定,即系统能够达到一定的准确性、包容性和稳定性,因此认为研制的检测装置基本可以满足实际的玉米种子净度检测需求。
[Abstract]:In the process of processing corn seed cleanliness, the screening accuracy of wormhole and moldy grain is not enough, which leads to the low germination rate of processed seeds, which still can not meet the demand of "single seed seeding". And if mildew corn enters the food chain, it will seriously harm human and animal health. In view of the above problems, based on the summary of domestic and foreign research results, based on the air coupled ultrasonic nondestructive testing technology, signal processing method, Research on the testing method and equipment of corn seed cleanliness by computer and electronic technology. The main research contents and results include: 1) the principle and method of ultrasonic detection of corn seed cleanliness are studied. Firstly, the ultrasonic wave is used to detect the purity of maize seed. A comparative germination test of irradiated maize seeds was carried out. The safety of ultrasonic detection of corn seeds was verified. Then the ultrasonic response characteristics of corn borer and mildew were analyzed, that is, the principle of detection. The mechanical properties of corn granules were analyzed. According to its physical quantity, the longitudinal wave velocity of corn grain is calculated, and the frequency of excitation signal is set according to the demand of sound velocity and detection precision. Finally, aiming at the demand of experiment for detecting precision and obtaining high signal-to-noise ratio signal, Integrated and optimized ultrasonic signal acquisition equipment. (2) A method for detecting corn worm-borer seeds based on air-coupled ultrasonic technology is proposed. The ultrasonic signals collected in air at room temperature are de-noised and feature extraction is carried out. Four recognition models were established by using the single-sided and double-sided data of corn grains. The results showed that the orientation of seeds had no effect on the detection, and the correct recognition rate of all the models was over 85.00% for the two kinds of maize, among which the KNN model had the best performance. The recognition rate is high and stable. For all the data, the recognition rate is higher than 97.00. The research results verify the feasibility of using air-coupled ultrasonic technology to detect corn worm-borer particles. A method based on air-coupled ultrasonic technology is proposed to detect corn mildew particles. First of all, the Hilbert-Huang transformation of mildew and intact seed signal data is done. The marginal spectrum of Hilbert with frequency band of 300-498Hz was extracted as the classification feature of two kinds of seeds, and the classification and recognition models of BPNNNSVMIMCA and KNN4 were established, and the performance of the model was tested. The results showed that the best recognition effect for mildew and intact particles was BPNN model. The recognition rate is over 80.00%, especially the number of hidden layer neurons is 6, and the recognition rate is 85.71 / 91.43SVM model. The results show that it is feasible to use air-coupled ultrasonic technology to detect corn mildew particles.) A device for detecting corn seed cleanliness based on air-coupled ultrasonic technology has been developed. Integrated with cornseed air coupling ultrasonic signal acquisition equipment, At the same time, the software testing system is designed and implemented. The accuracy, inclusiveness, stability and running rate of the system are tested. The test results show that the recognition rate of the system for all varieties and types of corn grains is above 80%. The experimental results are stable, that is, the system can achieve a certain accuracy, inclusiveness and stability, so it is considered that the developed detection device can basically meet the actual needs of corn seed cleanliness detection.
【学位授予单位】：中国农业大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：S513;S237

【相似文献】