综合利用计算机视觉、图像处理技术,增加茶叶的形状参数,改进神经网络算法,实现了茶叶品质识别的自动化。研究中通过数码相机等直接得到茶叶图像,经过对图像格式进行转换和预处理,然后基于HSI模型提取的茶叶颜色特征参数和二值化后图像提取的茶叶形状特征参数,通过遗传神经网络,最后完成对茶叶的自动识别。实验结果表明此方法能取得更好的识别效果,计算机的检测结果与人工检测结果高度吻合。
The figure characteristics of tea and improved neural-network, computer vision and image processing were combined together to realize automatic identification of external quality of tea leaf. Firstly a tea-leaf image was obtained by a digital camera directly. The parameters of tea HSI model and parameters of the figure was extracted to identify tea leaf after image conversion and preprocess. Then completed automatically identify of tea-leaf through the Genetical-Neural network. The experiments reveal that the method improves the consistence between computer inspection and manual inspection.
[1] 陆建良, 梁月荣, 龚淑英, 等. 茶汤色差与茶叶感官品质相关性研究[J]. 茶叶科学, 2002, 22(1): 57~61.
[2] 王文杰, 罗守进, 黄建琴, 等. 电脑测定茶叶色泽的方法研究[J]. 茶叶科学, 2005, 25(1): 37~42.
[3] Borah.S, Bhuyan.M.Quality indexing by machine vision during fermentation in black tea Manufacturing . Proceedings of the SPIE The International Society for Optical Engineering, v5132, 2003, 468~475.
[4] 刘忠伟, 章毓晋. 十种基于颜色特征图像检索算法的比较和分析[J]. 信号处理, 2000, 16(3): 77~82.
[5] 田玉敏, 梁若莹. 计算机彩色输入输出设备常用颜色空间及其转换[J]. 计算机工程, 2002, 28(9): 111~115.
[6] Gonzalez R C Woods R E. 数字图像处理[M](第2版). 阮秋琦, 阮宇智译. 北京电子工业出版社, 2004: 224~270.
[7] 李庆中, 张漫, 汪懋华. 基于遗传神经网络的苹果颜色实时分级方法[J]. 中国图象图形学报, 2000, 5(9): 784~799.
[8] 张代远. 神经网络新理论与方法[M]. 北京: 清华大学出版社, 2006: 155~207.
[9] 李峥嵘, 刘月峨, 向东健, 等. 基于内容的小麦害虫图像检索系统研究与实现[J]. 农业工程学报, 2007, 11(23): 210~215.
浙公网安备 33019902000101号 
