摘要
针对甘蔗茎节与茎间颜色相近和由于表皮上白色果粉的干扰导致茎节难以识别的问题,提出了一种基于高光谱成像技术的茎节识别与定位方法.采集236个茎节和茎间样本的高光谱图像(874~1734 nm),采用连续投影算法(SPA)提取5个特征波长(1022 nm,1062 nm,1456 nm,1609 nm和1649 nm),建立偏最小二乘法(PLS)分类模型,利用该模型对20组甘蔗高光谱图像进行识别,生成甘蔗茎节的二值化图像,采用图像处理的方法进行茎节定位.实验结果表明,高光谱成像技术测量结果的标准差为0.7 mm,绝对误差的最大值为2.6 mm,能够有效识别与定位甘蔗茎节,为蔗种的防伤芽自动切割提供技术支持.
Due to the fact that the color of sugarcane nodes and internodes are similar to each other,and the interference of white fruit powder on the skin,node recognition and location were affected seriously.A method for node identification and localization was proposed based on hyperspectral imaging.236 sugarcane samples（874~1734 nm） were collected by the hyperspectral imaging acquisition system.Using successive projections algorithm（SPA） to extract characteristics band（1022 nm,1062 nm,1456 nm,1609 nm and 1649 nm）,the PLS discriminant model were established by these 5 characteristic bands.The 20 groups of sugarcane hyperspectral images were identified by the established model,the binary image of sugarcane was gotten,image processing was used to locate the position of node.The experimental result showed the standard deviation was 0.7 mm,the maximum absolute error was 2.6 mm,and it could identify and locate the sugarcane node effectively,provide technical support for automatic cutting of sugarcane which could prevent injury buds.
出处
《轻工学报》
CAS
2017年第5期95-102,共8页
Journal of Light Industry
基金
国家自然科学基金项目(61403349
61503173)
郑州轻工业学院博士科研基金资助项目(JSJ20170017)
关键词
高光谱成像
甘蔗茎节识别与定位
连续投影算法
图像处理
hyperspectral imaging
identification and location of sugarcane node
successive projections algorithm（SPA）
image processing
