The invention discloses a CBMeMBer multi-target tracking method based on a random hypersurface. The method is mainly used to solve a problem that the shapes of multiple extension targets can not be estimated during a multiple extension target tracking process. By using the method, algorithm complexity can be reduced and simultaneously the precision of multiple extension target tracking is increased. The method comprises the following steps of (1) carrying out input measurement; (2) carrying out measurement division; (3) using a random hypersurface to carry out modeling on a target measurementsource; (4) predicting a target state; (5) updating the target state; (6) trimming the target state; (7) extracting the target state; and (8) determining whether an iteration is the last iteration, ifthe iteration is the last iteration, ending the operation, otherwise, returning to the step (2) to carry out next tracking. In the invention, the random hypersurface is used to carry out modeling onthe target measurement source so that a target center of mass position is accurately estimated in real time and simultaneously the shapes of the targets can be estimated too; and the time complexity of an algorithm is reduced.