The invention relates to a beam-forming method for ultrasonic imaging by adoption of generalized sidelobe cancellation based on a feature structure. The method comprises following steps: firstly, delay-time and forward-smoothing processing of sampling signals of receiving array elements; obtaining a forward spatial covariance matrix estimation; loading forward spatial covariance matrixes; combining the forward spatial covariance matrix estimation with direction vectors; calculating optimal self-adaptive weighting vectors based on the minimum variance principle; configuring a generalized sidelobe canceller with minimum variance in the equivalent solution in order to obtain two parts of self-adaptive and non-adaptive weighting vectors; decomposing characteristics of the forward spatial covariance matrix estimation of receipt signals and configuring characteristic threshold signal subspace; projecting integral weighting vectors to left singular vector space of the characteristic threshold signal subspace in order to obtain self-adaptive beam weighting vectors; and utilizing weighting vectors obtained from projection and receipt data of ultrasonic array elements to undergo the point multiplication in order to obtain ultrasonic beam synthesized data. The beam-forming method for ultrasonic imaging by adoption of generalized sidelobe cancellation based on the feature structure has following beneficial effects: characteristic vectors with great interference can be eliminated such that a higher anti-noise capacity of ultrasonic images and better image resolution are acquired.