Traffice Signs Recognition Research Based On Wavelet Neural Network

As we all known, image recognition is an important branch of pattern recognition. Through few decades, it has been applied successfully in the military, space exploration, medical science and post, etc. So it has great importance and practical value.The purpose of this thesis is to utilize the predominance of the wavelet neural network for the recognition of images. This paper revolves around the central task of image identification. It is mainly about collecting and preprocessing the original data of target images, methods of invariable feature extraction and the identification technology of the wavelet neural network.At the image preprocessing part, we smooth and denoise the target image firstly, then we detect its edge and buildup it, finally, in order to eliminate the effect of the translation, scaling, skewing and rotation on the recognition result. We proposed a method to normalize the target image, which involved in the feature extraction part.During the feature extraction part, we use moment algorithm and discuss normal moment, wavelet moment in detail. The traditional moment invariants have the defect: these moments are the whole features calculated from the whole image space, which are apt to disturbed by noise. Aim at the above defect, a new moment invariant wavelet moment is presented, which apply wavelet analysis to moment invariant. Thus wavelet moment possess the image objects invariant to translation, scaling and rotation. By using wavelet moment invariant, not only the local feature of image object is obtained, but also the description ability for the fine feature of image construct is improved. Thus the higher recognition rate is obtained, especial to similar images.In the recognition part, wavelet moments are used as the image features, then the features abstracted are optimized and finally the features optimized are combined with BP neural network and wavelet neural network classifier to make object image recognition. In order to ensure veracity of the recognition result, this paper use classes of traffic signs for training and testing. The experimental results demonstrate that compared with popular BP neural network, the wavelet neural net work is of much more practical value in self-traffic-navigation because of its high speed and precision.