The invention discloses a convolution retrieval method based on the truncated singular value decomposition theory. According to an echo model of a scanning radar, scanning radar azimuth echoes are built into convolutions of an antenna directional diagram and a target scattering coefficient. Scanning radar corner super-resolution imaging is achieved by using the convolution retrieval method. According to noise amplification in the convolution retrieval process, firstly, the noise amplification reason is analyzed from the algebraical angle; secondly, singular value decomposition is carried out on a convolution matrix, and singular value components causing noise amplification are removed; finally, convolution retrieval and scanning radar corner super-resolution imaging are achieved through the adoption of an algebra inversion method. By means of the method, the problem of the noise amplification caused by a traditional deconvolution method in the convolution process is effectively solved, and the accuracy of scanning radar corner super-resolution imaging achieved according to the deconvolution method is accordingly improved.