摘要
采用数字图像处理技术实现了对玉米种子表面裂纹的识别和检测。选择冷阴极荧光灯(CCFL)设计了图像采集的光照环境,建立了玉米种子图像的采集系统,然后针对玉米种子图像提出了一种基于籽粒形态学特征的表面裂纹检测方法。该方法采用水平和垂直边缘检测算子处理得到裂纹、种子边界和噪声等边缘信息;然后通过玉米籽粒的形态特征寻找其尖端位置,并使用图像代数运算的方法去除大部分非裂纹信息;最后根据裂纹的长度和位置特征提取得到裂纹,并计算裂纹的绝对长度和相对长度。对农大4967和农大3138两个品种的玉米分别选取裂纹粒和无裂纹粒各50粒进行图像识别,试验结果表明:识别准确率分别为94%和90%,基本满足玉米种子表面裂纹检测的精度要求。
The surface crack identification and detection of a corn kernel are studied based on digital image processing. Cold Cathode Fluorescent Lamps(CCFL) are chosen to construct the image capturing illumination environment, and a set of image acquisition system of the corn kernel is established. Then, a surface crack detection method is developed based on morphology of the corn kernel. A binary image including the cracks, boundary and noises is picked up with horizontal and vertical Sobel operators. Then, a majority of non-crack information is eliminated by image subtraction after finding the tip of the corn kernel using its morphology. Finally, the cracks are extracted and the absolute and relative lengths of cracks are calculated according to the crack length and position. A detecting experiment is carried out by 50 kernels with cracks and 50 kernels without cracks selected from NongDa 4967 and NongDa-3138 (two novel varieties of corn seed developed by China Agricultural University) respectively. The results indicate that the detecting accuracy are 94% and 90%. It can meet the requirement of the accuracy of surface crack detection of corn kernel.
出处
《光学精密工程》
EI
CAS
CSCD
北大核心
2007年第6期951-956,共6页
Optics and Precision Engineering
基金
国家自然科学基金资助项目(No.30471011)
高校博士点基金资助项目(No.20050019005)