基于在线连续极限学习机的图像分类改进算法

结合LBP算子提取图像的局部纹理特征,在分类阶段根据优化解进行矩阵逆的区别计算并加入正则因子,最后结合在线学习方法,提出准确在线连续极限学习机的图像分类改进算法.实验结果表明,改进算法在图像分类方面比传统的极限学习机有更快的学习速度,更好的泛化性能.

A modified algorithm for image classification based on online sequential extreme learning machine

Traditional Extreme Learning Machine (ELM) has some problems such as size of image dataset will influence effect of classification and low efficiency because of ignoring singularrity of hidden output matrix. In order to improve the defect and get better effect of Classification, Accurate Online Sequential Extreme Learning Machine (AOS-ELM) for the image classification is proposed. Firstly, AOS-ELM combined with LBP which could extract textrual features from subregion of image and using regularizetion factor and different manipulation matrix secondly, online sequential learning phase is put into in the end. Experimental results demonstrate that the AOS-ELM can learn faster and achieve better performance than traditional ELM.