摘要
采用高光谱成像技术（450-1 000 nm）对壶瓶枣的5种自然损伤（缩果病、裂纹、虫害、黑斑病、鸟啄伤）进行识别研究。利用高光谱成像系统采集了5种自然损伤及完好枣一共663个壶瓶枣样本的高光谱图像,并提取相应的感兴趣区域（ROI）,得到了样本的光谱数据。应用偏最小二乘回归（PLSR）、连续投影算法（SPA）从全波段中分别提取了9条、10条特征波长,利用Kennard-Stone算法将各类样本按照3∶1的比例随机分成训练集（500个）和测试集（163个）,并对其建立最小二乘支持向量机（LS-SVM）判别模型,结果表明使用SPA-LS-SVM建立的壶瓶枣自然损伤模型的整体判别正确识别率为93.2%。运用主成分分析（PCA）对由SPA提取出的10条特征波长（535、595、657、672、685、749、826、898、964、999 nm）所对应的单波段图像进行数据压缩,分别采用Sobel算子、区域生长算法Regiongrow并结合主成分图像识别出163个壶瓶枣样本的边缘与自然损伤特征区域,得出平均正确识别率达到90.8%。研究结果表明：采用高光谱成像技术可以对壶瓶枣的自然损伤进行光谱判别和图像识别。
Hyperspectral imaging technology covered the range of 450 - 1 000 nm was employed to detect natural defects（ shrink,crack,insect damage,black rot and peck injury） of Huping jujube fruit. 663 sample images were acquired which included five types of natural defects and sound samples. After acquiring hyperspectral images of Huping jujube fruits,the spectral data were extracted from region of interest（ ROI）. Using Kennard-Stone algorithm,all kinds of samples were randomly divided into training set（ 500 samples） and test set（ 163 samples） according to the proportion of 3 ∶ 1. Partial least squares regression（ PLSR） and successive projections algorithm（ SPA） were conducted to select optimal sensitive wavelengths（ SWs）,as a result,9 SWs and 10 SWs were selected,respectively. And then,least squares-support vector machine（ LS-SVM） discriminate model was established by using the selected wavebands. The results showed that the discriminate accuracy of the SPA-LS-SVM method was 93. 2%.Then,images corresponding to ten sensitive bands（ 535,595,657,672,685,749,826,898,964,999 nm） selected by SPA were executed to PCA. Finally,the images of PCA were employed to identify the location and area of natural defects feature through imaging processing. Using Sobel operator,region growing algorithm and the images of PCA,the edge and defect feature of 163 Huping jujube fruits could be recognized,the detect precision was 90. 8%. This investigation demonstrated that hyperspectral imaging technology could detect the natural defects of shrink,crack,insect damage,black rot and peckinjury in Huping jujube fruit in spectral analysis and feature detection,which provided a theoretical reference for the natural defects nondestructive detection of jujube fruit.
出处
《农业机械学报》
EI
CAS
CSCD
北大核心
2015年第7期220-226,共7页
Transactions of the Chinese Society for Agricultural Machinery
基金
国家自然科学基金资助项目(31271973)
山西省自然科学基金资助项目(2012011030-3)
关键词
壶瓶枣
自然损伤
高光谱成像
检测
Huping jujube Natural defects Hyperspectral imaging Detection
