光谱关键变量筛选在农产品及食品品质无损检测中的应用进展

农产品及食品的品质与安全一直以来都是人们关注的焦点，不仅关系着人们的身体健康，而且关系着社会稳定甚至国家安全。由于农产品及食品的品质不合格引发的安全事件备受社会各界的广泛关注。对农产品及食品的品质的监管长久以来都是分析检测领域的重点和难点。我国人口众多，对农产品和食品的消费量非常大。面对如此大量农产品及食品品质的无损快速检测需求，光谱法以其快速、无损、高效、环境友好、可现场检测等诸多特点，为农产品及食品品质的无损快速分析提供了良好的解决方案。然而，传统的光谱法在检测过程中所使用的数据量十分庞大，不仅在建立校正模型过程中会消耗大量时间，而且难以完成大量农产品及食品的品质在线高通量无损快速检测。大量数据的计算成为限制光谱类分析仪器工作效率的主要瓶颈之一，并且大量数据的计算对仪器设备的硬件配置也提出了非常高的要求，从而间接地提高了光谱分析技术的应用成本。近年来，关键变量筛选技术脱颖而出，并成为光谱分析的一个新热点。通过筛选，采用少量关键变量建立校正模型即可得到和全谱数据建模准确度相差无几的分析结果，从而可以有效提高分析仪器的工作效率并间接地降低光谱分析技术的应用成本，进而为农产品及食品品质的高通量检测提供了可靠的技术支持、为满足人民日益增长的美好生活需要提供科技保障。针对光谱关键变量筛选在粮食及粮食作物、蔬菜、水果、经济作物、肉类、食品品质与安全领域的无损检测应用进行综述，对光谱关键变量筛选技术的应用从筛选方法、应用范围、应用效果等方面进行了分类总结归纳，并就光谱关键变量筛选技术在农产品及食品品质无损检测中的应用从变量筛选方法特点及趋势、所选变量的稳定性和可靠性、所选变量的实际意义等方面进行了展望。

Application of Spectral Key Variable Selection in Non-Destructive Detection of the Qualities of Agricultural Products and Food

The quality of agricultural products and food has always been one of the focuses of attention.The quality and safety of agricultural products and food are related to people’s health and related to social stability and even national security.In recent years,the safety incidents caused by the unqualified quality of agricultural products and food have attracted all social circles’attention.The supervision of the quality of agricultural products and food has been the key point even difficulty in analysis and detection for a long time.Given a large population,the consumption of agricultural products and food is enormous in China.In the face of such a large number of the needs of non-destructive and rapid detection of agricultural products and food quality,spectroscopy analysis can provide a good solution for the non-destructive and rapid detection for agricultural products and food with the characteristics of fast,non-destructive,efficient,environmentally friendly,on-site testing.However,due to the large amount of data used in the traditional spectral analysis,it is time-consuming in developing calibration models and difficult to complete the online,high-throughput,non-destructive and rapid detection of the large number of agricultural products and food quality.On the other hand,the calculation of such a large number of data has also become one of the main bottlenecks limiting the efficiency of spectral analysis instruments,and the calculation of a large number of data also puts forward very high requirements for the hardware configuration of the instruments,which will increase the application cost of spectral analysis technology indirectly.In recent years,key variable selection has emerged and become a new hotspot of spectral analysis.According to the selection,calibration models can be developed by a few numbers of the key variables,which are of almost the same accuracy to the models developed by the full spectra.Thus it can improve the analytical instruments’working efficiency effectively and reduce the application cost of the spectral analysis technology.It will also provide reliable technical support for the high-throughput detection of agricultural products and food quality and provide the scientific and technological support for meeting the increasing demand of the people for a better life.In this paper,the applications of spectral key variable selection in the non-destructive detection of grain and grain crops,vegetables,fruits,cash crops,meat,food quality and safety were reviewed.With summarizing others’works in recent years,the applications of spectral key variable selection technology were summarized from the aspects of selection method,application scope,application effect,and so forth.Finally,the application of spectral key variable selection technology in non-destructive detection of agricultural products and food quality prospected from the aspects of the characteristics and trends of the variable selection methods,the stability and reliability and the practical significance of the selected variables.