基于声波的玉米含水量测定

Corn moisture measurement based on acoustic analysis

利用声音传感器采集玉米籽粒流从高处落到相同状态玉米堆时发出的撞击声音,对声音信号进行滤波预处理,提取声波信号强度、功率谱能量、谱峰值等特征值来描述信号。应用多元线性回归、二项式回归和神经网络等方法进行全面分析。试验结果表明:声波信号强度功率谱能量及谱峰值等特征值与玉米籽粒含水量之间存在较强的相关性,其中信号强度相关性最大。利用这些特征能够准确地测定籽粒含水量,其中二项式回归模型最接近实测值。

A method to measure the corn moisture by acoustic analysis was proposed. The sound wave was acquired by a computer system equipped with a sound sensor. After the filtration of noise from the sound signals, the features, such as the acoustic intensity, spectrum energy and peak value were calculated to describe the sound waves. Multiple linear regression, binominal regression and neural network methods were applied in the analyses. Results show that there exist prominent relationships between these features and the moisture content, and among these features acoustic intensity is the most significant one. Using these relationships the corn moisture content can be accurately measured. The predicted data by the binominal regression model is the closest one to the real data.