Purity identification of maize seed based on discrete wavelet transform and BP neural network

摘要

In order to identify the maize seed purity efficiently,after researching on image processing methods and classification algorithm in terms of the image characteristic of maize seed,a purity identification calculation based on discrete wavelet transform(DWT) and BP neural network was presented.Through this method,the RGB color model character parameters of the maize seed crown part were obtained,then the three color values were processed and analyzed by the two-level DWT.The average of every band was selected as the input samples for BP neural network,and purity identification results of maize seed as the output samples of neural network.Results demonstrated that this method can identify the maize purity effectively with accurate identification rate reaching 94.5％.%为快速有效地识别玉米种子纯度,针对玉米种子图像特征,对其图像处理方法和分类算法进行研究,提出一种基于离散小波变换和BP神经网络玉米种子纯度识别算法.该方法首先提取玉米种子冠部核心区域的RGB颜色模型特征参数,然后对三个色彩分量分别进行二层离散小波变换,提取各频带区域均值作为BP神经网络的输入样本,玉米种子的纯度分类作为神经网络的输出样本.实验结果表明该方法可准确识别玉米纯度并分类,正确识别率达94.5％.