苹果质地的近红外光谱无损检测模型研究

Study on nondestructive measurement model of apple fruit texture by near infrared spectroscopy

为了建立快速而无损检测苹果质地的新方法,应用近红外光谱仪研究不同建模方法和光谱预处理方法对苹果质地（脆度、硬度、回复性、凝聚性和咀嚼性）无损检测模型性能的影响。结果表明,波长范围400～2500nm内,采用改进偏最小二乘法、原始光谱结合反相多元离散校正处理所建苹果质地的校正模型最优,脆度、硬度、回复性、凝聚性和咀嚼性预测相关系数均大于0.8,而预测标准误差分别为7.6763N、6.5876N、0.0085、0.0175、1.2466N,残差之和均小于0.2。因此,通过近红外光谱对苹果质地进行快速而无损检测具有一定可行性,但模型精度有待进一步提高。

In order to establish a new method used to measure texture of apple fruit rapidly and nondestructively, the effects of different modeling methods and spectrum pretreatment methods on nondestructive measurement model performance of apple texture,including flesh brittleness,firmness, resilience,cohesiveness and chewiness using near infrared diffuse reflection spectroscopy. The results showed that modified partial least squares （MPLS） model,with respect to original spectrum combined with inverse multiple scatter correction （IMSC）, provided better prediction performance for the flesh brittleness, firmness, resilience, cohesiveness and chewiness of apple fruit,with correlation coefficient of prediction greater than 0.8,with standard error of prediction of 7.6763N,6.5876N,0.0085,0.0175 and 1.2466N respectively,with the sum of residual error less than 0.2. Therefore,apple texture measured by near infrared diffuse reflection spectroscopy rapidly and nondestructively was feasible, but the precision of the models could be improved further.