The invention discloses a maximum-probability time domain processing method for background extraction and offset in target radar cross section (RCS) measurement. Based on a maximum probability principle and sliding window average processing for extracting fixed background zero-Doppler clutter (ZDC), the method utilizes original RCS amplitude and phase measurement data of a type of auxiliary measuring body, such as a low-scattering end cap, or a measured target to obtain a maximum probability time-domain ZDC estimation by data domain-to-time domain transformation and maximum probability threshold processing; on the basis, a maximum probability likelihood estimation of data-domain ZDC is then obtained by time domain-to-data domain transformation, and thereby the adverse affection of the residual components of signals of an auxiliary measuring body on a ZDC estimation in a conventional method is eliminated. By time-domain maximum probability statistics threshold processing, the residual components of scattered signals of the auxiliary measuring body in the ZDC estimation are eliminated, and thereby the precision of ZDC estimation and the effectiveness of subsequent background offset processing are increased.