Researches On Wavelet Neural Network In Automatic Target Recognition Based On Wavelet Moment Invariants

One of the advanced functions of Modern weapon is Automatic Target recognition(ATR). Wavelet analysis, which is able to capture both spatial and frequency localization and provide global approximation information and local detail information at different scales, has become an effective tool for signal processing and image understanding. Moment invariants, with simple mathematical representation and perfect theoretical framework, are well-suited for invariant target recognition due to their invariance to general affine transforms and insensitivity to noise. The fast development of Artificial Neural Network(ANN) has provide a new approach to the research of Automatic Target Recognition, difficulties remain in ATR may be resolved by the ANN methods. Automatic Target Recognition(ATR) requires invariance in translation, rotation and proportional transformation higher cognition accuracy. All these can be done well in the way of capturing the characteristics of the image with invariants moment first and then distinguishing them through wavelet neural network. By combing the feature extracting algorithm with wavelet neural network, an image recognition system is established in this paper. This image recognition can be used to distinguish several kinds of targets. Four kinds of plane images are recognized in this paper.The research work focuses on the investigations of wavelet moment invariants, wavelet neural network and their applications to automatic target recognition community. Main points of the research involve moments in ATR, differences between Hu moment, Zernike moment which describe the global features of targets and wavelet moment which describe the local features of target are explored. A new target recognition system is established for ATR in this paper, which combines wavelet moment and wavelet neural network. Experiment has proved that this system has a good recognition ability.The principal research work are listed as follows:(1)In this paper, the author research on the moment features of images ,compare the image moment invariants of Hu moment, Zernike moment with Wavelet moment. Hu moment and Zernike moment have the ability to capture global features. Wavelet transform has the exceptional property of temporal-frequency localization. The combination of the moment and the wavelet...