The highlight number of an underwater target is one of important features for the underwater target classification and recognition. However, the highlight number feature extraction is easily affected by the target bearing and there is always a blind area near the vertical azimuth angle. By analyzing the highlight structure of an underwater target echo, a new method of highlight number extraction for the underwater target is proposed based on Gerschgorin disk and signal to noise ratio (SNR) supervising. It can solve the question of recognizing the underwater target near the vertical azimuth angle. The computer simulation result indicates that the method is also effective for all other azimuth angles. It can be used to classify and recognize between a dimensional underwater target and a decox.