图像亮度信息法选取多光谱莲花白叶片特征波段

利用液晶可调谐滤波器(LCTF)和CMOS相机组合的多光谱成像系统,在波长400~720nm内以5nm为间隔对莲花白叶片进行多光谱成像。首先根据图像亮度信息法的原理,计算得到各波段莲花白叶片的可识别度;然后对莲花白叶片的可识别度进行大小排序,综合图像的信息特征和可识别度,得出555nm、715nm、710nm、575nm、535nm、520nm、720nm、605nm和650nm 9个波段有较好的识别度;最后根据欧氏距离法和光谱角度匹配法分别对莲花白叶片的特征波段的分类精度予以统计,得到两种方法的分类精度分别为95.56%和93.13%。实验证明,选取的9个波段对莲花白叶片具有较好的分类精度,可作为莲花白叶片的特征波段。

Selecting feature bands for multi-spectral images of cabbage leaves based on imaging brightness information

The experiment takes cabbage leaves as research object to capture images based on multi-spectral imaging system with combination of liquid crystal tunable filter(LCTF) and CMOS camera by every 5 nm interval from 400 nm to 720 nm.Firstly,according to the principle of image brightness information,the value of distinguish degree for cabbage leaves are calculated for each band.Then,by sorting the values of distinguish degree for cabbage leaves,along with information features of the image and distinguish degree,it can be concluded that bands 555 nm,715 nm,710 nm,575 nm,535 nm,520 nm,720 nm,605 nm and 650 nm have better distinguish degree.Finally,the classification accuracy statistic of feature bands for cabbage leaves are 95.56% and 93.13% by using the principle of Euclidean distance and spectral angle match respectively.The experiment demonstrates that the nine bands are of ideal classification accuracy for cabbage leaves and these bands can be used as feature bands for cabbage leaves.