高光谱图像技术检测苹果外部损伤  

Detection of External Damage of Apple by Hyperspectral Image Technique

作　　者：刘晶晶[1] 刘付龙 史铁 孙超 张晋宝 门洪[1] Liu Jingjing;Liu Fulong;Shi Tie;Sun Chao;Zhang Jinbao;Men Hong(Department of Automation Engineering, Northeast Electric Power University, Jinlin 132012)

机构地区：[1]东北电力大学自动化工程学院,吉林132012

出　　处：《中国食品学报》2018年第1期278-284,共7页Journal of Chinese Institute Of Food Science and Technology

基　　金：国家自然科学基金-青年基金项目(31401569);国家自然科学基金面上项目(31772059);吉林省重点科技攻关项目(20170204004SF);吉林省科技发展计划项目(20150520135JH)

摘　　要：利用高光谱分选仪对苹果进行图像获取,基于光谱图像信息实现对苹果的外部损伤检测。由于光谱图像间的数据巨大,具有冗余性和相关性,通过主成分降维法提取640 nm波段下的图像作为特征信息进行外部损伤检测。光谱特征图像上光斑噪声的存在是造成损伤检测效果不佳的原因,分析光谱图像光斑的图像特征,利用二阶巴特沃斯高通滤波器抑除图像频域内的光斑信号,同时增强损伤区域的边缘图像信息。对比滤波前、后的特征图像损伤检测效果,滤波后的图像数据更能真实表现出样本自身的数据特点,结果表明:经过二阶巴特沃斯高通滤波器滤波后的苹果损伤检测效果明显,正确率达到95%。In this study, the image of apples was acquired with the hyperspectral sorter and the detection of external damage on apples had been achieved based on the spectral image information. The principal component analysis（PCA） was applied to extract the feature information in 640 nm in order to reduce the redundancy and correlation. Spot noise in spectral feature images caused the apple external damage detection was ineffective. So the feature of the spot in the spectral image was analyzed and the spot signal in the frequency domain of the image was suppressed and the edge image information of the damaged area was enhanced by second-order Butterworth high-pass filter. Comparing the damage detection effect, it was obviously that the characteristics of the data showed up more factually in the image after filtering.The result showed that the damage detectioneffect of the apple was more significant after Butterworth with the accuracy of95%.

关 键 词：苹果 光斑噪声 高光谱图像 主成分分析 二阶巴特沃斯高通滤波器 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]