近红外高光谱成像技术的桃轻微损伤早期检测

摘要

In this paper,the near-infrared hyperspectral imaging technology (900~1700nm) is applied for slight bruises detection on the Beijing-8 peach after 12 hours of the injury occurred .The principal component analysis and independent component analysis algorhtm is used for data dimensionality reduction on the injure region .The median filtering , threshold segmentation and mathematical morphology processing is used for damage region segmentation .60 normal peaches and 60 injury peaches is used in our expeirment for early detection of slight bruises .And the expeirmnt result shows that NIR hy-perspectral imaging technology is effective for early detection of slight bruises on peaches .%以“北京8号”桃为研究对象，应用900～1700nm 范围内近红外高光谱成像技术对损伤发生12h 后的桃进行损伤早期检测研究。通过光谱比较，确定出识别光谱区域为950～1350 nm。基于此光谱区域，利用主成分分析与独立成分分析不同方法进行降维，结合中值滤波、阈值分割等数字图像处理算法进行损伤区域的检测。对60个正常样本和60个损伤样本进行检测，主成分分析方法对损伤果的识别率为85％，独立成分分析方法对损伤果识别率为96．6％，两种方法对正常果的识别率均为100％。结果表明：近红外高光谱成像技术能有效地进行桃损伤的早期检测。