本文以李果实作为研究对象，基于高光谱图像技术对不同成熟度的李果实（未熟、半熟、成熟和过熟）样本的图像信息进行采集，对采集样本的图像进行中值滤波去噪处理。运用Matlab软件编程对各种成熟度样本的图像进行颜色特征值提取，分别获得RGB和HSV彩色图像模型不同分量的平均值（μ）和标准差（σ）作为颜色特征值，并建立RGB、HSV颜色特征值以及RGB-HSV特征值相融合的样本成熟度PLS判别模型，并对所建立的判别模型进行预测。结果表明，基于RGB-HSV相融合颜色特征值的判别模型准确率优于RGB与HSV，其对未熟、半熟、成熟、过熟的判别准确率达到了98.4%、90.0%、85.6%及90.9%。该方法建立的判别模型不仅简化，而且增强了模型的判别能力，为实现李果实成熟度的无损检测和判别提供理论依据。

In this paper, plum was regarded as the research object, and image informations of different mature degrees of plum fruit (unripe, ripe, mature and overmature) samples were collected based on hyperspectral image. Then the images were treated with median filtering denoising. RGB and HSV color image models were obtained from different mature degrees by Matlab software and average and standard deviation of different components were regarded as the color feature value. RGB, HSV and RGB-HSV color characteristic value were established to identify the plum maturity PLS, and the established models were predicted. The results showed that the accuracy rate of discriminated model based on RGB-HSV color feature value was better than that of RGB and HSV. The accuracy rate of immaturity, instrumental, mature and overmature plum reached 98.35, 90.00%, 85.85% and 90.85%, respectively. The results showed that this discriminated model not only simplified but also enhances the discriminated ability of models, and provided the theoretical basis for the nondestructive detection and discrimination of plum maturity.

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）；山西农业大学青年科技创新项目（2016005）