论文部分内容阅读
基于老化不同时间的稻种的生理学和物理学特性,提出一种基于多尺度小波变换和灰色神经网络的稻种发芽率红外热预测模型,实现稻种发芽率的快速、无损检测,解决传统发芽实验法实验周期长、操作复杂等问题。从不同发芽率稻种的胚芽部位提取144组数据,通过多尺度小波变换,分析逼近信号和细节信号,得出第3层细节信号(d3)贡献最大。以第3层细节信号作为模型的输入,随机分为校正集和预测集,校正集96组,预测集48组。分析和比较老化不同时间的稻种的红外热差异,通过偏最小二乘算法(PLS)、BP神经网络、径向基神经网络(RBFNN)和灰色神经网络(GNN),建立稻种发芽率红外热预测模型。结果表明,GNN建立的稻种发芽率模型预测效果最优,其中校正集相关系数(RC)和标准偏差(SEC)分别为0.9619、2.5013,预测集相关系数(RP)和标准偏差(SEP)分别为0.9554、2.4172,相关性达到较高水平且误差较小。研究表明采用小波分解和灰色神经网络建立稻种发芽率红外热预测模型的方法是可行的。
Based on the physiological and physical characteristics of rice seeds aged at different times, an infrared thermal prediction model of rice germination rate based on multi-scale wavelet transform and gray neural network was proposed to achieve fast and nondestructive detection of rice germination rate and to solve the problems of traditional germination Experimental method Experimental cycle is long, complicated operation and other issues. 144 sets of data were extracted from the embryo sites of rice varieties with different germination rates. The multi-scale wavelet transform was used to analyze the approximation signal and the detail signal, and the third layer of detail signal (d3) was the largest contributor. The layer 3 detail signal is used as the input of the model and is randomly divided into a calibration set and a prediction set, with 96 sets of calibration sets and 48 sets of prediction sets. By analyzing and comparing the infrared difference of the rice varieties aged at different times, the germination rate of rice seeds was established by partial least squares (PLS), BP neural network, radial basis neural network (RBFNN) and gray neural network (GNN) Thermal prediction model. The results showed that the germination rate model of rice established by GNN had the best prediction effect. The correlation coefficient (RC) and standard deviation (SEC) of the calibration set were 0.9619 and 2.5013 respectively, and the correlation coefficient (RP) and standard deviation (SEP) 0.9554,2.4172, the correlation reached a higher level and the error is smaller. The research shows that it is feasible to establish the infrared thermal prediction model of rice germination rate by using wavelet decomposition and gray neural network.