摘要
为实现对李果实成熟度的快速、准确判别,采用高光谱成像技术(450~1 000nm)采集不同成熟阶段(未熟期、半熟期、成熟期、过熟期)的李果实共计640个样本的高光谱信息进行判别研究。对不同成熟阶段的李果实样本测定表征成熟度的理化指标(可溶性固形物和硬度值)并进行单因素方差分析,结果表明:不同成熟度样本的两项指标均存在极显著差异,硬度值差异最大。采用连续投影算法(Successive projections algorithm,SPA)和主成分分析(Principal component analysis,PCA)分别提取得到不同成熟度样本光谱数据的10个特征波长(381、3 8 2、3 8 7、4 0 8、4 4 3、4 9 4、5 9 6、8 1 3、9 6 3、1 0 0 8 nm)和前5个主成分值(累积贡献率达9 7.8 3%)。基于RGB、HSV颜色模型对不同成熟度李果实样本图像进行颜色特征提取,最终得到6项颜色特征指标(R、G、B及H、S、V分量图像的平均值和标准差)。分别建立基于光谱信息、图像信息及融合信息的偏最小二乘(Partial least squares,PLS)判别模型,结果表明:基于特征波长和RGB特征信息融合值建立的PLS模型判别结果最佳,准确率达9 1.2 5%。由此可见,采用高光谱成像技术在光谱和图像信息方面对不同成熟度李果实进行判别是可行的、有效的,该研究为实现李果实成熟度在线检测提供了理论依据。
In order to achieve the rapid and accurate determination of the plums maturity, the hyperspectral imaging technology with the band of 450~1000nm was adopted to collect altogether 640 hyperspectral information samples of the plum for judgment.The samples were classified into four maturity, unripe, mid-ripe, ripe and over-ripe.Firstly, the soluble solid content （SSC） and firmness were chosen as physical and chemical indexes on the plum samples during various mature stages, and one-way analysis of variance was conducted, the results of which indicated that there were remarkable differences in the two indexes among samples , and the greatest differences was found in firmness.Afterwards, successive projections algorithm （SPA） was adopted, and 10 wavelengths at 381, 382, 387, 408, 443, 494, 596, 813, 963 and 1008nm were selected as the optimal sensitive wavelengths;principal component analysis （PCA） was used to compress spectral data of plum samples, the analysis suggested that the cumulative contribution of the top 5 principal components with rate of 97.83%.Finally, 6 color feature indexes including the mean and the standard deviation of the R, G, B, H, S and V component images were obtained through extraction based on the RGB and HSV color feature information of the plum images.Spectral information, image information and fusion information were based on the partial least squares （PLS） models to judge the plums maturity.The results demonstrated that the PLS model based on the sensitive wavelength and RGB characteristic information was the best, and accuracy rate of 91.25%.Thus it could be seen that the adoption of hyperspectral imaging technology to judge the plums with various degrees of maturity on the aspects of spectrum and image information was feasible and effective, and this research had provided theoretical foundation for achieving the online detection of the plums maturity.
出处
《农机化研究》
北大核心
2017年第12期141-145,共5页
Journal of Agricultural Mechanization Research
基金
国家自然科学基金项目(31271973)
山西省自然科学基金项目(2012011030-3)
