东亚飞蝗光谱特征波长筛选与龄期识别方法研究

Selection of Spectum Feature Wavelength and Recognition of Different Ages of Manilensis

利用高光谱图像采集系统在400~1 000 nm波段范围内采集东亚飞蝗成虫、5龄、4龄和3龄的前胸背甲光谱信息;每个龄期提取15像素×15像素目标区域平均反射率信息作为样本信息;提出了一种基于K均值聚类和主成分分析（K-PCA）相结合特征波段提取方法,对比分析K-PCA和SPA（投影连续变换）2种特征波长提取方法,采用Fisher判别分析方法分别对K-PCA和SPA筛选的特征波长建立东亚飞蝗龄期识别判别模型,实验结果表明K-PCA筛选出的特征波长数少且正确识别率为98.25%。K-PCA筛选的特征波长为468 nm、555 nm、635 nm、710 nm、729 nm、750 nm、786 nm和899 nm。本文提取的东亚飞蝗特征波长为东亚飞蝗的龄期识别奠定基础,进而对蝗灾的监测与预防提供了技术支持。

Manilensis is one of the major pests in China. A method for recognizing different ages of manilensis was presented based on K-means clustering and principal component analysis（ PCA） with selected feature wavelength. The hyperspectral images in the range of 400 ~ 1 000 nm of manilensis back at differnet ages among adult,5-age,4-age and 3-age were collected and the average spectral information of target region on manilensis back with the size of 15 pixel × 15 pixel was extracted. A wavelength secleting method with combined PCA algorithm and K-means clustering（ K-PCA） was proposed. The model for identifying manilensis ages was built by using Fisher algorithm and then compared with K-PCA algorithm and successive projections algorithm（ SPA）. The experiment results showed that the K-PCA algorithm needed fewer wavelengths but with the higher accuracy of 98. 25%. The final feature wavelengths of K-PCA algorithm were 468 nm,555 nm,635 nm,710 nm,729 nm,750 nm,786 nm and899 nm. The proposed method provides a certain technology support for manilensis monitoring and precention.