基于改进模拟退火算法的木材类型识别方法

Feature selection method based on improved simulated annealing algorithm

木材的优劣直接影响到木材的用途和价格,木材类型的准确识别尤为重要。在改进的模拟退火算法的基础上提出了木材类型特征选择方法FS-ISA,采用分类器的识别率作为评价指标,从14个灰度共生矩阵特征参数中选取最优特征参数组合,通过FS-ISA算法选取熵、对比度、角二阶矩、差异度、方差、逆差距等6个参数,组成最佳参数体系。结果表明,木材类型识别率最高达91.7%。

The strengths and weaknesses of wood directly affect the application and price of wood, and the accurate identification of wood types are especially important. An improved simulated annealing algorithm for feature selection method FS-ISA is proposed. With recognition rate of the classifier as evaluation method, the optimal combination of characteristic parameters including contrast, entropy, energy, difference degree, variance, and inverse difference moments, which resulted in the optimal recognition effects on the types of wood textures, are selected intelligently by the FS-ISA method from the characteristic parameters of 14 gray-level co-occurrence matrixes. The experimental results show that the wood type recognition rate is as high as 91.7%.