For the important problem of moving target detection on a moving platform, an ego motion compensation algorithm is presented based on the Gabor feature descriptions. After compensating the camera ego motion, it is able to detect the relative moving targets to the camera efficiently. The interesting points are detected by the scale-space theory firstly. Then the feature description for every interesting point is created according to its character scale and character angle. And the coarse matching results are obtained by cross-match. At last, the transform model is estimated by the Random Sample Consensus (RANSAC) algorithm. Experimental results show that the new method can effectively and accurately detect the moving targets on the moving platform.