基于超像素的人工神经网络图像分类

Artificial Neural Network Image Classification Based on Super-pixel

基于人工神经网络对图像标签分类,为简化后续数据处理,先用Normalized Cut将图像分割为超像素,提取特征向量,通过输入训练样本集,对网络进行训练,在最小均方误差意义下得到网络参数,最后在Matlab的仿真实验中基于不同隐藏层节点,使用BP神经网络模型对图像超像素进行分类。

This paper describes the classification of image labeling based on artificial neural network .For simplifying the data processing , obtaining the super-pixel as a result of image segmentation , we took the extracted feature vectors as input for network , obtained the parameters of network in the sense of minimum mean square error after training the network with training samples .At last, in the simulation experiments of Matlab , we classified the super-pixel of image based on BP neural network model for all hidden neurons .