摘要
为自动检测自然场景中成熟椪柑果实,提出了基于最优色差分量的树上椪柑机器识别方法。在统计待检测图像椪柑果实平均色调的基础上,以色环单位圆中椪柑果实平均色调为基准线,作红色(R)、绿色(G)、蓝色(B)向基准线投影并求代数和,建立最优色差分量模型,以该模型对自然场景中碰柑图像进行果实识别。以393幅有椪柑果实的裁切图像为研究对象,比较最优色差分量法与G-R、G-B、R-B及2G-R-B分量法的识别效果,试验结果显示最优色差分量法误检率、漏检率最低,果实目标检出完整率最高,整体检出效果最佳。最优色差分量法还可推广到与背景存在色差的其他目标自动检测中。
The study proposes a method of machine recognition of Ponkan in trees based on optimum chromatic aberration component for the automatic detection of mature fruits in natural scenes. Based on the statistics of the average hue of Ponkan in the detected images and taking the average hue in the graph of color-ring as the base line, the study figures out the algebra sum of the reference lines projected by the red(R), green(G), blue(B) and establishes the model of optimum chromatic aberration component, by which Ponkan fruits in natural scenes can be detected. Taking 393 images as the study object, the research compares the recognition effects of optimum chromatic aberration component method with the component methods of G-R, G-B, R-B and 2G-R-B. The results show that the one proposed in this study demonstrates the lowest error rates of false-positive and false-negative judgments while it is of the highest rate of the integrity detection and the best overall detection effects. Moreover, the optimum chromatic aberration component method can be applied to the automatic detection of other targets which are of chromatic aberration with the background.
出处
《湖南农业科学》
2013年第12期120-122,126,共4页
Hunan Agricultural Sciences
基金
湖南农业大学大学生创新性实验计划项目(XCX 12069)
关键词
最优色差分量
机器识别
树上椪柑
平均色调
optimum chromatic aberration component
machine recognition
Ponkan in trees
average hue
