A method to efficiently identify object from its background noise for a sonar picture using fuzzy artificial neural network is introduced. The sonar images are from a high precision array sonar imaging device. The ability to set focus to a certain range of interest will produce a well lamented grayscale picture around the focus distance. The intelligent window over the digital picture based on neural network feedback is used to partition the picture into different levels of grid and the grid data is treated as the input for the network. Multiple levels of gridding are applied corresponding to the layers of neural network. Samples are provided for training to get the focus related threshold values. The method is faster and can be used for the pre-identification of object from sonar image so additional processing can be called for shape recognition. Matlab programming is used for data processing.