Abstract: ‍ ‍The accuracy of the direct positioning determination depends on the fineness of the grid division， and the traditional off-grid method alleviates the dependence on the grid division， but it still has the problems of low compensation accuracy and high algorithm complexity. In order to solve these problems， this paper proposes a dynamic grid direct position determination algorithm based on SOMP. Firstly， the initial signal was denoised by employing subspace adaptation method， and the two dimension space of interest was divided into scattered grids. Secondly， the grid quantization error was considered in this paper and the influence on direct position determination was analyzed. Lastly， the proposed method was introduced. The method was different from the traditional method of JSOMP which was also based on SOMP and settled the compensation value after all iterations. In the process of each iteration， the method employed Taylor compensation to improve the location of the grid point which had the highest correlation with the received signal. At the same time， the related parameter of the original dictionary matrix was replaced with the updated one， so as to obtain a more optimal dictionary matrix. The results of the computer simulation tests showed that the proposed algorithm could overcome the grid mismatch problem effectively and estimate position of radiation source accurately. Compared with JSOMP， OG-SBI and MUSIC-Taylor algorithms， the proposed method has higher speed of operation and lower deviation of position location.