基于Roberts算子的牡丹红斑病叶病斑边缘检测方法比较

Comparison of edge detection methods of red spots in infected peony base on improved Roberts algorithm

为了准确快捷地获取牡丹红斑病叶病斑边缘,使用Roberts算子对图像进行边缘检测分析,研究了不同染病阶段病斑的特点及边缘检测阈值。根据不同染病阶段的病斑边缘检测阈值的范围,提出了适用于牡丹红斑病边缘检测的改进Roberts边缘检测方法。结果表明:Roberts算子能够对牡丹红斑病病斑边缘进行检测,传统Roberts算子检测的方式对目标叶片的病斑边缘平均检出率为63%。改进Roberts算子对牡丹红斑病病斑边缘检测的效果有所提高,病斑边缘平均检出率在73%左右,其检测阈值选取较传统算法更加合理。从一定程度上减少了病斑边缘检测的人为干扰因素,客观上提高了检测的自动化水平,增强了模型的鲁棒性,降低了选择阈值时的数据损失。

In order to accurately and rapidly obtain the red spot edge of peony infected by red spot disease, spot images were detected and analyzed by Roberts algorithm, and characters of red spot at different infected stages and threshold values of edge detection were researched. According to the ranges of detection threshold values at different infected stages, the suitable edge detection method is proposed for peony red spot disease based on improved Roberts algorithm. The results indicate that Roberts algorithm can be used to detect red spot edge of infected leaf. Average recognition rate of red spot edge is 63% by using the traditional Roberts algorithm, and it is increased to 73% using the improved Roberts algorithm. The improved Roberts algorithm has more stable thresholds, and it can decrease the human disturbance factors to some extend, objectively increase the automation level of detection, strengthen robustness of the model, and reduce the data loss during threshold selection.