A method for generating a predictive image to enable remote control of an unmanned ground vehicle (UGV) under conditions of long communications latency. An original digital image of the UGV environment (from a UGV-mounted camera) is projected forward such that a lower portion falls on a ground plane defined to approximate the UGV's travel surface and an upper portion falls on a far plane defined approximately normal to the ground plane and forward of the UGV. A perspective transformation technique used to generate the predicted image separates the ground plane image from the far plane image. Two separate mathematical perspective transforms are defined and applied, one for the lower (ground plane) part of the image, and one for the upper (far plane) part of the image. The resulting predicted upper and lower image are rejoined to form the complete predicted image, which may then be presented on a video screen.