基于差分图像边界距离的粮粒孔洞自动检测  

Automated Detection of Holes in Wheat Kernel Based on Boundary Distance of the Differential Image

作　　者：张红涛[1] 胡玉霞[2] 剧森[1] 张恒源[1] 

机构地区：[1]华北水利水电大学电力学院,郑州450011 [2]郑州大学电气工程学院,郑州450001

出　　处：《中国粮油学报》2015年第6期117-120,共4页Journal of the Chinese Cereals and Oils Association

基　　金：国家自然科学基金(31101085);河南省基础与前沿技术研究计划(122300410145);河南省高等学校青年骨干教师资助计划(2011GGJS-094);华北水利水电大学教学名师培育项目(2014108)

摘　　要：粮粒孔洞的自动检测是近红外高光谱图像技术检测粮粒内部害虫中的一个关键问题。提出基于差分图像边界距离的粮粒孔洞自动检测方法,该方法通过求取粮粒(内部)轮廓与阈值分割后二值图像的差分,若差分图像中的目标与粮粒边界的最远距离大于某个阈值时,则该目标应判别为边界(内部)孔洞。用米象的幼虫、蛹和成虫3个侵染阶段粮粒的900帧近红外图像进行训练,用450帧近红外图像进行检验,结果表明该方法不仅可以判断粮粒是否存在孔洞,还能检测出孔洞的数量及形态,其中边界孔洞和内部孔洞的识别率分别为97.33%和95.56%,证实了基于差分图像边界距离的粮粒孔洞检测方法是可行的。The automatic detection of holes in wheat kernel is the key to detect insects in wheat kernels based on near-infrared hyperspectral imaging technology. A method of detecting holes in wheat kernel automatically was proposed based on boundary distance of differential image. The differential image was acquired between the kernel （internal） contours image and the binary image after thresholding. If the maximum distance between the object and the kernel boundary was more than a threshold in differential image, the object should be acted as a boundary or internal hole. The near infrared-images were acquired from the infested wheat kernels with the larva, pupa and adult stages of Sitophilus oryzae （L.） The method was trained by 900 images and tested by 450 images. The results showed that the method could determine whether there were holes in wheat kernel, and the number of holes and morphology could be detected. The correct identification ratio of boundary holes and internal holes was over 97.33% and 95.56% respectively. The experiment showed that it was practical and feasible.

关 键 词：仓储害虫 粮粒 边界孔洞 内部孔洞 差分图像 检测 

分 类 号：S24[农业科学—农业电气化与自动化]