基于改进的主成分分析法的矿工表情识别

针对传统的矿工面部表情识别方法中对矿工面部表情进行特征提取的时间较慢且识别准确率不高的问题,以主成分分析法为基础,运用Fisher线性判别法对传统的主成分分析法进行改进。首先在主成分分析法的基础上增加一个类间离散矩阵,使其投影后不同类别之间特征点的距离更大,同一类别之间特征点的距离更加紧凑,对矿工面部表情图像特征提取的结果更具有代表性和针对性;然后运用径向基神经网络将低维非线性可分的矿工面部表情图像对应的特征矩阵映射到高维空间并使其线性可分,从而实现对矿工面部表情的识别和分类。实验结果表明,所提出的方法对矿工面部表情识别的识别率为89.0%,优于传统矿工面部表情分类识别算法,在矿井安全监控、疲劳驾驶等领域有较好的应用前景。

Miner expression recognition based on improved principal component analysis

Aiming at the problem that the feature extraction of miners'' facial expressions is slower and the recognition accuracy is not high for the traditional miner facial expression recognition method, based on the principal component analysis method, Fisher''s linear discriminant method is used to improve the traditional principal component analysis method. Firstly, based on the principal component analysis method, an inter-class discrete matrix is added to make the distance between the feature points of different categories become larger after projection, and the distance between the feature points of the same category is more compact, so that the result of feature extraction to the miners'' facial expression images is more representative and targeted. Then, the radial basis network is used to map the low-dimensional and nonlinear separable miner''s facial expression feature matrix to the high-dimensional spatially separable class to realize the identification and classification of miners'' facial expressions. The experimental results show that the recognition rate of the miner''s facial expression reaches 89.0%, which is superior to the traditional miner''s facial expression recognition algorithms. The method has a good application prospect in the fields of mine safety monitoring and fatigue driving.