虫蛀玉米种子的空气耦合超声波检测

发布时间：2018-03-24 15:47

  本文选题：种子胚　切入点：超声波检测　出处：《声学学报》2017年05期

【摘要】：提出了一种基于空气耦合超声波技术的玉米种子虫蛀孔洞颗粒和完好颗粒分类识别方法·首先根据玉米颗粒的弹性模量、泊松比和密度等物理量计算出了玉米颗粒的声速,并根据检测精度需求设定了激励信号频率。然后采用MATLAB对采集的两类种子超声波信号数据进行分析处理,并分析了种子厚度和摆放方位对超声波响应特征的影响。最后建立了K近邻(KNN)、簇类独立软模式法(SIMCA)、Fisher线性判别(LDA)和决策树(DT)识别模型,并对模型性能进行了测试.结果表明;种子孔洞深度、胚部厚度和正反面方位不同,即超声波在种子表面的反射程度不同、在种子中传播声程不同,则起声波信号衰减程度不同,导致接收到信号的幅值不同,且样本点在主成分分析(PCA)特征空间的分布也不同。4种识别模型均可以实现对两类玉米的分类识别,其中KNN模型性能最佳,其对虫蛀孔洞颗粒和完好颗粒的正确识别率分别为98%100%,误差带为2%,0。此结果说明采用空气耦合超声波技术可以实现对玉米种子虫蛀孔洞颗粒的检测。
[Abstract]:In this paper, a classification and identification method of corn seed wormholes and intact particles based on air-coupled ultrasonic technique is proposed. Firstly, the sound velocity of corn grain is calculated according to the physical quantities such as elastic modulus, Poisson's ratio and density. The frequency of excitation signal is set according to the demand of detection precision, and then the two kinds of ultrasonic signal data of seed are analyzed and processed by MATLAB. The effects of seed thickness and placement azimuth on ultrasonic response were analyzed. Finally, the recognition models of K nearest neighbor KNNN, cluster independent soft pattern method SIMCAA Fisher linear discriminant and decision tree DTT were established, and the performance of the model was tested. The depth of the hole, the thickness of the embryo and the positive and negative directions are different, that is, the ultrasonic wave reflects on the surface of the seed with different degrees, and the sound path in the seed is different, the attenuation degree of the sound wave signal is different, and the amplitude of the received signal is different. The distribution of sample points in the feature space of PCA is also different. 4 kinds of recognition models can realize the classification recognition of two kinds of maize, and the KNN model has the best performance. The correct recognition rates of wormhole and intact particles are 98% and 2% respectively. The results show that the air coupled ultrasonic technique can be used to detect corn seed wormhole particles.
【作者单位】： 中国农业大学信息与电气工程学院;河南农业大学机电工程学院;中国科学院声学研究所;
【基金】：国家级星火计划重点项目基金(2015GA600002) 中央高校基本科研业务费专项基金(2016XD002)资助
【分类号】：S435.132;TB559
，

本文编号：1658965