利用图像识别技术计算薇甘菊锈病的相对病斑面积

Calculation of spot area of Mikania micrantha rust based on image processing technology

【目的】为了定量评估薇甘菊柄锈菌对薇甘菊的防控效果,研发一种基于图像识别技术的高效、准确的薇甘菊叶片相对病斑面积的计算方法。【方法】利用图像识别、网格法、复印称重法3种相对面积的计算方法,分别计算薇甘菊感染柄锈菌后的相对病斑面积,并结合以手动分割的结果作为标准,计算各方法的绝对准确率和绝对误差并作为评价指标,最终对3种相对病斑面积的计算方法开展系统、科学的评估。【结果】与网格法、复印称重法相比,基于超绿和超红算法分割病斑的图像识别方法能够准确、快速地计算出薇甘菊锈病病斑的相对面积,其绝准确率均值达到98%以上,绝对误差率均值仅有1.81%,处理一张4608×3456像素彩色图像只需要45.77 s。【结论】与传统的病斑面积计算方法相比,图像识别技术能够准确、快速地将病斑区域与健康区域分割,并准确地计算相对病斑面积。

【Aim】To quantitatively evaluate the biological control effect of rust(Puccinia spegazzinii)in Mikania micrantha,an efficiently and accurate image processing technology was developed for calculating the relative lesion area(RLA)of M.micrantha leaf.【Method】In this study,we calculated the RLA of infected leaf in M.micrantha based on three methods,including image processing technology,grid method and photocopy weighing method,respectively.In addition,using the manual segmentation method to calculate the RLA as the standard,the absolute accuracy and absolute error of above three methods were calculated and used as the systematic and scientific evaluation index.【Result】The result shows that compared with grid method and copy method,the image processing technology based on Excess Green+Excess Red(ExG+ExR)algorithm segmentation disease spot can quickly and accurately calculate the RLA of infected leaf in M.micrantha with 98% absolute accuracy and 1.81% absolute error rate.In addition,it only takes 45.77s to process a 4608×3456 pixels color image.【Conclusion】Compared with the traditional method,because the image processing technology could accurately and quickly segment the lesion area and the healthy area,the accurate RLA of infected leaf could be obtained.