Underwater target classification is mainly based on the analysis of geometrical properties of acoustic shadows. The new generation of imaging sonar provides a more accurate description of the acoustic wave scattered by targets. Therefore, combining the analysis of shadows and echoes is a promising way to improve automated target classification. Efficient and reliable schemes for automated target classification rely on a learning-based model instead of only using experimental samples of target acoustic signature to train the classifier. With this approach, a good performance level in classification can be obtained if the modeling of the target acoustic signature is accurate. This paper mainly focuses on the first phase of classification method consisting in precisely modeling the acoustic signature of the targets. As imaging sonars operate at high or very high frequency, from 400 kHz to 2 MHz, typically, the model core is based on acoustical ray-tracing. Several phenomena have been considered to increase the realism of acoustic signature (multipath propagation, interaction with the surrounding seabed, edge diffraction, etc.). To assess the consistency of the results, spherical and cylindrical objects have been modeled. Then, to train the classifier, a data base is created using manmade objects of more complex shapes and natural objects. These modeled acoustic signatures are then compared with high resolution images provided by actual sonars, to demonstrate the quality of the model.