基于多光谱图像参数的茶叶摊青评价模型研究  

The model research of tea green airing evaluation based on imaging technology

作　　者：张宪[1] 贾广松 赵章风[1] 钟江[1] 乔欣[1] ZHANG Xian JIA Guangsong ZHAO Zhangfeng ZHONG Jiang QIAO Xin(College of Mechanical Engineering, Zhejiang University of Technology, Hangzhou 310014, China)

机构地区：[1]浙江工业大学机械工程学院,浙江杭州310014

出　　处：《浙江工业大学学报》2017年第2期125-129,共5页Journal of Zhejiang University of Technology

基　　金：浙江省自然科学基金资助项目(LQ12C13004);国家"十二五"科技支撑计划项目(2014BAD06B06);浙江省重大科技专项重大农业项目(2013C02024-2);国家自然科学青年基金资助项目(31201138)

摘　　要：茶叶加工过程中,第一道摊青工艺的处理是提高茶叶品质的关键环节,其中水分含量多少直接影响茶叶加工品质,含水率传统检测方法损坏样品且检测速度较慢,所以建立一套准确、无损和快速的茶叶叶片水分检测方法对于评价茶叶加工质量有着现实意义.利用高精度数码相机,对摊青过程中10个含水率梯度600个样本进行数据采集,通过研究茶叶叶片形状、纹理及颜色的变化,实时监测摊青过程中含水率的变化,利用与含水率相关性大于0.92的特征参数,通过回归值验证与BP神经网络建立含水率非线性预测模型,准确度在90%以上.证明此模型对于研究茶叶叶片含水率和指导茶叶生产具有一定准确度和可靠性.Traditional detection method of water content includes drying and weighing method and sensor changed into electric quantity detection method.Although it is more accurate,but it will damage the samples,and the detection rate is slow.Establishing a nondestructive,accurate and rapid detection of biochemical parameters of tea leaf method to evaluate the quality of tea processing has important practical significance.By studying the change of the tea leaf shape,texture and color,the change of moisture content,the change of moisture content in the process of spreading green is real time monitored.And according to the parameters which correlation with moisture content is greater than 0.92,the nonlinear prediction model is established and verified using BP neural network.The accuracy is above 90% which could prove that this model piece of moisture content research and guidance for tea production has a certain accuracy and reliability.

关 键 词：图像处理 参数提取 茶叶含水率 BP神经网络 

分 类 号：S571.1[农业科学—茶叶生产加工]