基于小型化便携式光谱成像技术的分类识别系统研究

Research on classification and recognition system based on miniaturized portable spectral imaging technology

小型化便携式光谱成像越来越多应用于日常生活中,给人们的生活提供了更多的便利。柑橘属水果是人们日常生活中经常食用和储存的水果之一,在智能冰箱中,柑橘由于品种相似等原因不利于识别和分类,光谱成像技术利用其物质不同的特征波长实现其识别。采用新型的单芯片式光谱成像芯片搭建小型化便携式的光谱成像系统,利用光谱成像技术实现对柑橘属水果的分类识别,并通过分批次样本进行交叉验证,同时建立光谱分辨率与准确度关系,通过约束光谱分辨率,有效提高了柑橘属水果分类识别的准确度,光谱分辨率小于20 nm,识别准确度可达到95%以上。

Miniaturized portable spectral imaging was increasingly used in daily life, providing more convenience for people′s lives. Citrus was one of the fruits that people often eat and store in their daily life. In the smart refrigerator, citrus was not conducive to identification and classification due to similar varieties. Spectral imaging technology used its characteristic wavelengths to realize its recognition. A new type of single-chip spectral imaging chip was used to build a compact and portable spectral imaging system. Spectral recognition technology was used to realize the classification and identification of citrus fruits, and cross-validation was carried out by batch samples, simultaneously establish the relationship between spectral resolution and accuracy, by constraining the spectral resolution, which effectively improved the classification and recognition accuracy of citrus fruits. Spectral resolution was less than 20 nm,and recognition accuracy could reach more than 95%.