This thesis presents an automatic soil texture classification system using hyperspectral soil signals and wavelet-based statistical models. Previous soil texture classification systems are closely related to texture classification methods, which use images for training and testing. Although using image-based algorithms is a straightforward way to conduct soil texture classification, our research shows that it does not provide reliable and consistent results. Rather, we develop a novel system using hyperspectral soil textures, better known as hyperspectral soil signals, which provide rich information and intrinsic properties about soil textures. Hyperspectral soil textures, in their very nature, are nonstationary and time-varying. Therefore, the wavelet transform, which is proven to be successful in such applications, is incorporated. In this study, we incorporate two wavelet-domain statistical models, namely, the maximum likelihood (ML) and the hidden Markov model (HMM) for the classification task. Experimental results show that this method is reliable and robust. It is also more effective and efficient in terms of practical implementation than the traditional image-based methods.