Abstract not found

those of the authors and should not be interpreted as representing the official policies, either expressed or implied, of the funding agencies. 4 Contents

This paper discusses an obstacle detection algorithm developed at NIST in support of the obstacle detection and rough terrain conditions. The algorithm is a hybrid of grid-based and sensor-based obstacle detection and mapping techniques. The perception and obstacle detection/mapping module is part of the integrated 4D-Realtime Control System (RCS) system [1][2]. It consists of two sections: an obstacle detection section and a mapping section. The obstacle detection section processes range data read from a Ladar sensor. The algorithm converts range data into Cartesian coordinates in the sensor coordinate frame, and uses this information to detect obstacles. The second section, the mapping module, projects obstacle points onto a grid-based map. The map is used by the 4D-RCS planner module [11] to generate a traversable path for the vehicle.We have demonstrated autonomous driving with obstacle detection and avoidance on the NIST grounds and the Nike site at speeds of up to 24 km/h

In recognizing objects in an outdoor scene, range and reflectance (or color) data provide complementary information. This paper presents the results of experiments in recognizing outdoor scenes containing roads. trees. and cars from the Navlab (Navi- gation Laboratory) The recognition program uses range and reflectance data obtained a scanning laser range as well as color data from a color camera segmentation of each image into primitive regions. models of objects are matched using various properties.

des only 2-D information whereas the vehicle moves in a 3-D world. Calibration problem 2-D Possible solution: Camera model Flat ground plane Vertical objects Example: Road following (U. Maryland) f Original image (intensity, 256x256) Edge extraction in small windows (up to 64x64) of window locations Road model . Figure The original image, along with the windows and located road boun- daries Passive depth estimation The location of a point can be computed from its projections from differents viewpoints. From different cameras (stereo), or from dif- ferent positions motion). to find good matches between images match ? (features pixels) Binocular trinocular ? How to use the vehicle's motion ? PASSIVE STEREO Y Object in world Direction from left Direction from right View from right image la- View from left image x = - Disparity y = - Finding good matches Similarity measures Epipolar constraints Ordering constraints Coarse-to-fine Small range of disparity values Overd