Abstract: It is hard for traditional methods to detect targets in correlated non-Gaussian clutter backgrounds. The likelihood ratio test based on spherically invariant random process clutter could get good performance by modeling clutter accurately. However, it is very hard to get closed form of test statistical magnitude for that the probability distribution function form is usually complicated. Based on information geometry theory, the distance detector is defined by comparing the distance between the distribution estimated by observation data and two hypothetical distributions, translating detecting problem into geometry problem on statistical manifold. Simulation results show that detection performance of distance detector under correlated non-Gaussian clutter backgrounds outperforms traditional methods, simultaneously, it can be easily operated.