摘要
为了研究检测苹果表面轻微损伤的有效的方法,以红富士苹果为试验样本,通过高光谱成像采集系统采集苹果样本的高光谱图像,根据正常苹果表面区域和刚损伤、损伤后(3,10,24h)的损伤区域光谱反射率平均曲线得到有效光谱区域;用掩膜法对图像进行背景分割,并基于有效光谱区域做主成分分析,选取第四主成分(PC4)提取损伤区域,运用阈值分割的方法建立提取损伤区域的算法模型;应用该算法模型对正常苹果和损伤苹果进行检测。检测结果表明:正常苹果样本正确检测率达到100%,损伤苹果样本的正确检测率为97.5%,总体检测精确度高达98.75%,说明利用高光谱成像技术可以有效快速检测出苹果表面的轻微损伤。
In order to study the effective method to detect slight surface bruised on apple, Fuji apple was used as experimental samples to complete the experiment. The hyperspectral images of experimental sample were acquired by the hyperspectral imaging system. the effective spectral regions were got according to the average spectral reflectance curves of the surface area of the normal apples, apples of the injury immediately and apples of the injury after（3, 10, 24h）.The mask method was used for image background segmentation. And principle component analysis was transform based on the effective spectral region. The forth component（PC4） scores image was selected as the suitable image for detecting the damaged area in apples and the threshold segmentation method was established a model to extract the damaged area of apples surfaces. Both the intact apples and bruised apples were detected using above the algorithm model. The test results showed that 100% of the normal apples were correctly classified, 97.5% of the damaged apples were correctly recognized and the overall detection accuracy is 98.75%.The study illustrates that slight bruised on apples surface could be detected efficiently using hyperspectral imaging technology.
出处
《沈阳农业大学学报》
CAS
CSCD
北大核心
2015年第4期508-512,共5页
Journal of Shenyang Agricultural University
基金
沈阳市大型仪器设备共享服务专项项目(F15-166-4-00)
辽宁省教育厅科学技术研究项目(L2011114)
辽宁省科学事业公益研究基金项目(2012005015)
关键词
高光谱成像
轻微损伤
检测
苹果
hyperspectral imaging
slight bruise
detection
apple
