基于BP神经网络的非结构化道路识别

Unstructured Road Detection Based on BP Neural Network

基于BP神经网络,针对非结构化道路的彩色图片,利用熵、对比度等纹理特征值作为BP神经网络的输入层,设隐层有12个节点,输出层有1个节点,进行网络训练。完成网络训练之后,可以得到网络的阀值,然后将图片中的每一个20×20小块的纹理特征值输入BP神经网络的输入层,经过运算判断小块是否为道路部分,直至完成整个图片的判别。

Based on BP neural network,in view of the unstructured road color images,using the texture eigenvalues as a BP neural network input layer trains the network,the hidden layer of the network is equipped with 20nodes,the output layer is equipped with one node.After finishing the network training,we can get threshold matrix and weight matrix of the network,then put the texture characteristic value of picture in each 20square area into the BP neural network input layer at a time,to determine whether the small area is road section or not,until completing the whole picture,to get a smooth road boundary curve.