苹果的粉质化是指苹果果肉发软、汁液减少等一系列物理和生理变化现象, 采用高光谱散射图像技术结合信号稀疏表示分类算法(SRSA)研究了苹果的粉质化分类问题。首先利用平均反射算法(MEAN)提取了600~1 000 nm的高光谱散射图像特征; 引入遗传算法(GA)解决分类样本的不均衡问题, 在此基础上, 把苹果的粉质化分类问题, 转化为一个求解待识别样本对于整体训练样本的稀疏表示问题。仿真结果表明, 基于信号稀疏表示分类算法的苹果粉质化分类精度为79.8%, 高于偏最小二乘判别分析(PLSDA)的74.8%, 为苹果的粉质化分类提供了一种新的有效的方法。

Mealiness, characterized by the sensation of deteriorative texture and lack of juiciness, is a symptom of physical and physiological change of apple fruit. Hyperspectral scattering image technique coupled with sparse representation of signals algorithm (SRSA) were used for assessing apple mealiness. Firstly, mean reflectance (MEAN) algorithm was utilized to extract spectral features of hyperspectral scattering images of apple sample between 600 and 1 000 nm. Genetic algorithm (GA) was applied to deal with the problem of unbalanced sample. And apple mealiness classification was changed into a math problem, that the solution of sparse representation between trained samples and test samples. Compared with partial least squares discriminant analysis (PLSDA) (74.8%), sparse representation of signals algorithm yielded a higher accuracy (79.8%). Therefore, sparse representation of signals algorithm can be used as a new method for mealiness classification of hyperspectral scattering images.