论文部分内容阅读
酸皮病是一种重要的洋葱采后病害,对洋葱的采后生产造成了巨大的经济损失。本研究利用短波红外高光谱成像系统对洋葱采后酸皮病进行检测,分别从健康的和已感染酸皮病的洋葱上采集高光谱反射图像(950~1 650 nm),通过对两种洋葱的光谱进行主成分分析的结果表明,洋葱的颈部区域在1 070、1 400 nm波长下对酸皮病的发生最具指示性。采用两种最佳波长下获得的对数比率图像对两种不同的图像进行分析研究。第一种分析方法是利用全局阈值(0.45)将感染酸皮病的图像区域从对数比率图像中分离出来,通过对分离图像区域的象元素进行Fisher判别分析,可辨别出80%感染酸皮病的洋葱。第二种分析方法是从对数比率图像上获得三个参数(最大值、对比度及匀质性),再将这三个参数作为输入特征向量导入到支持向量机(Gaussian kernel,γ=1.5)中,利用这种方法,可辨别出87.14%感染酸皮病的洋葱。应用本研究结果可进一步开发出一种多光谱成像系统,用于洋葱包装流水线上酸皮病的检测。
Acidic skin disease is an important onion postharvest disease, causing huge economic losses to onion production. In this study, shortwave hyperspectral hyperspectral imaging system was used to detect the post-harvest acidosis onions. Hyperspectral reflectance images (950 ~ 1 650 nm) were collected from healthy and acidophilic onions, The results of principal component analysis showed that the onion neck region is the most indicative of the occurrence of acidosis at the wavelength of 1 070 and 1 400 nm. Two different images were analyzed using logarithmic ratio images obtained at the two best wavelengths. The first method used the global threshold (0.45) to separate the region of the acidified skin from the logarithmic ratio image. By Fisher’s discriminant analysis of the image elements in the segmented image area, 80% of infected acid skin Sick onions. The second method is to obtain three parameters (maximum value, contrast ratio and homogeneity) from the logarithmic ratio image and then use the three parameters as input eigenvectors to a support vector machine (γ = 1.5) In this way, 87.14% of the oncidium infected with onion can be distinguished. The results of this study can be further used to develop a multispectral imaging system for the detection of acidosis on onion packaging lines.