Following successful demonstrations in road-following, research efforts in autonomous navigation have focused on cross-country navigation. In cross-country navigation, autonomous vehicles need to detect obstacles and maneuver around them using on-board electronics. Due to the dynamic nature of the en-vironment (moving vehicle/moving obstacle), positions of obstacles need to be constantly tracked in order to avoid collisions. This requires a high speed range detection system. Furthermore, for military applications a passive ranging system is desirable in order to reduce the detectability of the vehicle. In this paper we present a pipeline architecture that performs correlation-based feature matching between two images in near real-time. Based on this architecture, a binocular stereo range detection system has been implemented using DATACUBE image processing boards. The system matches 512 x 512 pixel stereo image pairs in approximately 5 seconds using a disparity range of 128 pixels, and 256 x 256 pixel stereo image pairs in approximately one second using a disparity range of 64 pixels. The range information provided by the system is processed and displayed in different forms using a TAAC-1 graphics accelerator board attached to the SUN-4 host system. We present the feature matching algorithm, its hardware implementation, and show results obtained using the binocular stereo matching system. Finally, we discuss future directions and other possible applications of the system.