摘要
提出一种自动获取玉米籽粒大小特征和总粒数的方法:通过振动给料机实现玉米籽粒快速喂料,利用输送带和线阵相机的无偏差匹配,准确获取玉米籽粒图像,对籽粒图像进行二值化、除杂、筛选、粘连籽粒分割等处理,测量玉米籽粒的粒长、粒宽和籽粒总粒数。研究了图像处理过程中不同阈值、不同粘连籽粒处理方法对测量结果的影响。对100个样品进行的试验结果表明:阈值数设定为24、采用平均面积法时,测量的误差最小,测量结果与人工测量值相比较,玉米籽粒总粒数、粒长、粒宽的相关系数分别为0.999 3、0.900 0、0.930 0,平均相对误差分别为0.26%、3.85%、3.63%。
An automatic method to obtain the size and total number of maize corn was presented. Corn grains are fast fed by vibrating feeder. Maize kernel image was accurately captured by, using error-free match between conveyor belts and linear array camera. Post processing of binary kernel image, removing impurities, filtering and spliting adhesions grain was done to get length, width, and total number of corn grain. The effects of different thresholds and different adhesion grain treatments on the measurement results were studied during image processing. By testing 100 samples of different varieties, the results show that the error of the measurement is the smallest when the threshold number is set to 24 and the average area method is used. Compared with the manual measurement, the correlation coefficient of maize grain total grain number, grain length, grain width is 0.999 3, 0.900 0, 0.930 0, respectively, and the average relative error is 0.26%, 3.85%, 3.63%, respectively.
出处
《湖南农业大学学报（自然科学版）》
CAS
CSCD
北大核心
2017年第3期329-335,共7页
Journal of Hunan Agricultural University(Natural Sciences)
基金
湖北省自然科学基金项目(2015CFB480)
华中农业大学学科交叉研究专项(2662015JC005)
关键词
玉米籽粒
大小特征
二值化
图像处理
corn kernel
size characteristics
binarization
image processing
