Information Geometry, a new subject originated from differential geometry, studies the statistics problem in manifold and metric space, and serves as a new solution method for the problem involved in radar signal processing and data processing. In the current study , radar echoes were modeled as the complex multivariate Gaussian distribution, and a kind of CFAR detector was proposed based on the Riemannian geometric structure of the manifold, and the false alarm probability and detection probability was analyzed. Simulations show that the detection performance of the information geometry method was better than the traditional coherent accumulation method for the small pulse number case. The findings of the research will be helpful to improve the detection performance of mechanical scan radar system at low PRF work mode.