Combining rotation-invariance images and neural networks for road scene understanding
Zhigang Zhu, Haojun Xi, Guangyou Xu · 2002
In this paper we present the results of training and testing backpropagation networks for the outdoor road scene understanding. Both the road orientations used for vehicle heading and the road categories used for vehicle localization are determined by the integrated system. The main features of the work are as follows. (1) The comprehensive image analysis techniques are combined with the adaptive neural networks. (2) An omni-view image sensor is used to extract image samples. The rotation-invariance image features are obtained for the classification network, and the results are used to select the orientation-estimation networks. (3) The internal representation, especially the number of the hidden units, is analyzed. Experimental results with real scene images are given.