摘要
为了快速提取农田作物行中心线,设计了"2G-R-B灰度变换—Otsu自动阈值图像二值化—形态学图像处理—左右边缘中间线检测算法"4步图像预处理方法,得到代表作物行中心的特征点;扫描特征点图像,将特征点横坐标存放于二维数组中,通过聚类方法,得到代表作物行的已知点;根据基于已知点的随机直线检测算法,检测出作物行.实验结果表明,利用该算法不仅可以克服光照影响而且可以提取不同作物的作物行.处理一幅640×480的彩色图像平均耗时196ms,正确识别率达95%,能够满足农业机器人田间作业的实际需求.与霍夫变换(HT)、随机霍夫变换(RHT)相比,该算法显示了计算时间和存储空间上的优越性.
In order to rapidly detect crop rows, designing "2G-R-B gray scale transform, Otsu image binarization,morphological image processing and middle line detection between left edge and fight edge of crop rows" four-step method of image pre-processing for extracting the feature points represents the crop rows centers. By scanning feature points image, feature points abscissa values were stored in a two-dimensional array ,known-points through crop rows were obtained by clustering method. Crop rows were detected according to the random line detection algorithm based on known- points. Test results show that the method not only functions well for a wide range of illumination condition but also is suitable for a range of different crops. It takes about 196ms for a 640 ~ 480 pixels color image and the recognition rate reaches 95%. It can meet the actual requirement of agricultural robot in field operation. Compared with Hough Transform( HT)and Randomized Hough Transform(RI-IT) ,this method demonstrates the computational and spatial advantages.
出处
《应用基础与工程科学学报》
EI
CSCD
北大核心
2013年第5期983-990,共8页
Journal of Basic Science and Engineering
基金
国家自然科学基金(61201395
61272394
61005033)
河南省高校创新科技人才计划(13HASTIT039)
河南省高校骨干教师资助计划(2012GGJS-057)
河南省高等学校矿山信息化重点学科开放实验室开放基金(KY2012-09)