红枣表面损伤的特征光谱提取

Application of Hyperspectral Feature Extraction to Detect Surface Damage in Red Jujubes

以宁夏灵武长枣的损伤为研究对象,利用高光谱成像技术,针对红枣不同损伤形式采集波长650～950nm范围的图像,应用主成分分析(PCA)对图像降维,根据主成分分析中各个波长特征值数据,贡献值比例,提出668nm和715nm为特征波长。使用特征波长光源对样品进行图像识别,解决红枣表面损伤分类的方法,其检测结果对损伤果的识别率达到98.7%。

Spectral images of red jujubes with different types of surface damage were collected by hyperspectral imaging technology over the wavelength range of 650-950 nm. Principal component analysis （PCA） was applied for dimension reduction of the spectra. The characteristic wavelengths of 715 nm and 668 nm, where samples were recognized, were proposed based on eigenvalues obtained from PCA. As a result, classification of surface damage in red jujubes achieved with an accuracv of 98.7%.