摘要
种植环境下的玫瑰花分布紧密、互相遮挡,为准确检测并提取玫瑰花的特征,基于机器视觉对玫瑰花的颜色和形状进行识别与处理。首先选取双边滤波对玫瑰花图像去噪,然后采用六角锥体模型颜色空间(hexagonal cone colour model,HSV)提取玫瑰花颜色,创建滚动条函数对六角锥体颜色模型各分量图阈值分割从而确定最佳阈值,最后运用形态学运算、面积阈值、孔洞查询填补等方法提取玫瑰花轮廓,并提出玫瑰花内切圆形状拟合算法,将拟合内切圆的圆心和半径作为玫瑰花图像特征。结果表明,玫瑰花颜色阈值能够有效去除玫瑰花枝叶、泥土等图像,形状拟合算法能有效提取玫瑰花的形状特征,并擦除玫瑰花苞。运用该算法单朵玫瑰花识别率为98.17%,3朵及以下重叠玫瑰花的识别率为92.67%,4朵及以上重叠玫瑰花的识别率为74.07%,被枝叶遮挡的玫瑰花识别率为83.03%,该套机器算法在复杂的种植环境中能有效识别并提取玫瑰花的特征值,结果可为玫瑰花采摘机器人研究提供重要技术支撑。
Roses in the planting environment are closely distributed and shielded from each other.In order to accurately detect and extract the characteristics of roses,the colour and shape of roses were recognized and processed based on machine vision.Firstly,bilateral filtering was selected to denoise the rose image,and then the colour of the rose was extracted by using the hexagonal cone colour model(HSV),and a scrollbar function was created to segment the threshold of each component of the hexagonal cone colour model to determine the optimal threshold.Finally,the contour of the rose was extracted by means of morphological operation,area threshold,hole query and filling.The fitting method of the shape of the inner circle of the rose was proposed,and the center and radius of the fitting inner circle were used as the image features of the rose.The results showed that the rose colour threshold could effectively remove the images of the rose branches and leaves,and the shape fitting algorithm could effectively extract the shape features of the rose and erase the rose bud.By using this method,the recognition rate of single rose was 98.17%,that of overlapping roses with 3 or less roses was 92.67%,that of overlapping roses with 4 or more roses was 74.07%,and that of roses blocked by branches and leaves was 83.03%.This set of machine algorithm could effectively recognize and extract the characteristic values of roses in a complex planting environment,which provided important technical support for rose-picking robot.
作者
李锐风
杨云福
杨永发
于永顺
LI Ruifeng;YANG Yunfu;YANG Yongfa;YU Yongshun(College of Mechanical and Transport,Southwest Forestry University,Kunming 650224,China)
出处
《中国农业科技导报》
CAS
CSCD
北大核心
2024年第4期106-113,共8页
Journal of Agricultural Science and Technology
基金
云南省科技厅科技技术项目(202104AR040012)。
关键词
玫瑰花
机器视觉
图像识别
特征提取
rose
machine vision
image recognition
feature extraction
