The principal component analysis network (PCANet) is a kind of deep subspace network based on the simplified architecture of convolutional neural network. To address the issue that PCANet cannot process image samples in real-time during the convolutional kernel extraction process, this article proposes an incremental sequential row-column 2DPCA network (IRC2DPCANet). This method can process training samples on time in the process of filter training, which can improve the efficiency of network training. The experiments on three typical face datasets, which is PIE, AR and Yale, indicate that this method has good classification performance. Finally, the influence of the filter number and filter size on classification rate is also investigated.