摘要
为准确快速识别出仓储玉米是否包含霉变以及感染黄曲霉毒素的玉米颗粒,提出基于图像处理技术的定性、定量检测方法。基于感染霉菌的玉米颗粒表层会发生颜色褐变、发黑等特点,通过MATLAB对灰度图像二值化、区域填充、去干扰、连续膨胀等操作,统计出霉变颗粒数目;同时,基于感染黄曲霉毒素的玉米颗粒在365 nm紫外光照射下会发出独特的黄绿色荧光(bright greenish-yellow fluorescence,BGYF)特性,通过对彩色图像进行增强、形态滤波、彩图二值化、去除干扰、图像连续膨胀等,统计出感染黄曲霉毒素玉米颗粒的数量。用4组共112粒玉米对所提方法进行检测验证的结果表明,霉变玉米颗粒的正确检出率达93.75%以上,感染黄曲霉毒素玉米颗粒的正确检出率亦可达87.5%以上,能够达到区分检测的要求。
The image processing method based on Matlab has been adopted to screen mouldy or aflatoxin infec- ted corn kernels. The color of corn kernels will become dark or black because of infected fungi, as a result, mildew kernels can be identified by image processing technology such as binaryzation of gray - scale image, region filling, noise removal and continuous image expansion. Meanwhile the corns infected with aflatoxin can emit unique bright greenish -yellow fluorescence （BGYF） when they are exposed under the 365 nm ultraviolet irradiation;thus aflatoxin infected kernels can be detected by image processing methods of color image enhancement, median filtering, binaryza- tion of color image, noise removal and continuous image expansion. Finally, an accuracy of over 93.75 % has been a- chieved for mildew corn kernels. An accuracy of over 87.5 % has been also achieved for aflatoxin infected kernels, which indicates that this method is available to be used to detect mildew infection and aflatoxin level of single corn kernels.
出处
《中国粮油学报》
EI
CAS
CSCD
北大核心
2014年第2期82-88,共7页
Journal of the Chinese Cereals and Oils Association
基金
"十二五"国家科技支撑计划(2012BAK08B04)
关键词
玉米
霉变
黄曲霉毒素
图像处理
检测
corn, mildew, aflatoxin, image processing, detection
