CAJ | 학술논문

裂纹是衡量鲜枣品质的重要指标之一，果皮裂纹加速鲜枣的腐烂，导致鲜枣货架期的缩短，严重降低鲜枣的经济价值。采用高光谱成像技术在380～1030 nm波段范围内对鲜枣裂纹的位置及大小信息特征进行快速识别。选用偏最小二乘回归（PLSR）、连续投影法（SPA ）和全波段图像主成分分析（PCA ），得到鲜枣裂纹相关的敏感波段。然后利用选取的鲜枣裂纹的敏感波段对建模集的132个样本建立最小二乘支持向量机（LS-SVM ）判别模型，并对预测集的44个样本进行判别。对PLSR-LS-SVM ，SPA-LS-SVM和PCA-LS-SVM 判别模型采用ROC曲线进行评判，得出PLSR-LS-SVM模型对鲜枣裂纹定性判别的结果（area＝1，std＝0）最佳。选取PLSR回归系数挑选出的5条鲜枣裂纹敏感波段（467，544，639，673和682 nm）对应的单波段图像进行主成分分析，其中将主成分PC4的图像结合图像处理技术，最终识别出鲜枣裂纹的位置、大小信息。结果表明，采用高光谱成像技术结合光谱图像处理可以实现鲜枣裂纹定性判别和定量识别的研究，为进一步开发相关仪器的研究提供理论方法和依据。
렬문시형량선조품질적중요지표지일，과피렬문가속선조적부란，도치선조화가기적축단，엄중강저선조적경제개치。채용고광보성상기술재380～1030 nm파단범위내대선조렬문적위치급대소신식특정진행쾌속식별。선용편최소이승회귀（PLSR）、련속투영법（SPA ）화전파단도상주성분분석（PCA ），득도선조렬문상관적민감파단。연후이용선취적선조렬문적민감파단대건모집적132개양본건립최소이승지지향량궤（LS-SVM ）판별모형，병대예측집적44개양본진행판별。대PLSR-LS-SVM ，SPA-LS-SVM화PCA-LS-SVM 판별모형채용ROC곡선진행평판，득출PLSR-LS-SVM모형대선조렬문정성판별적결과（area＝1，std＝0）최가。선취PLSR회귀계수도선출적5조선조렬문민감파단（467，544，639，673화682 nm）대응적단파단도상진행주성분분석，기중장주성분PC4적도상결합도상처리기술，최종식별출선조렬문적위치、대소신식。결과표명，채용고광보성상기술결합광보도상처리가이실현선조렬문정성판별화정량식별적연구，위진일보개발상관의기적연구제공이론방법화의거。
Crack is one of the most important indicators to evaluate the quality of fresh jujube .Crack not only accelerates the de-cay of fresh jujube ,but also diminishes the shelf life and reduces the economic value severely .In this study ,the potential of hy-perspectral imaging covered the range of 380~1 030 nm was evaluated for discrimination crack feature (location and area ) of fresh jujube .Regression coefficients of partial least squares regression (PLSR) ,successive projection analysis (SPA) and princi-pal component analysis (PCA) based full-bands image were adopted to extract sensitive bands of crack of fresh jujube .Then least-squares support vector machine (LS-SVM ) discriminant models using the selected sensitive bands for calibration set (132 samples) were established for identification the prediction set (44 samples) .ROC curve was used to judge the discriminant mod-els of PLSR-LS-SVM ,SPA-LS-SVM and PCA-LS-SVM which are established by sensitive bands of crack of fresh jujube .The results demonstrated that PLSR-LS-SVM model had an optimal effect (area=1 ,std=0) to discriminate crack feature of fresh jujube .Next ,images corresponding to five sensitive bands (467 ,544 ,639 ,673 and 682 nm) selected by PLSR were executed to PCA .Finally ,the image of PC4 was employed to identify the location and area of crack feature through imaging processing .The results revealed that hyperspectral imaging technique combined with image processing could achieve the qualitative discrimination and quantitative identification of crack feature of fresh jujube ,which provided a theoretical reference and basis for develop instru-ment of discrimination of crack of jujube in further work .