TerraMax Vision at the Urban Challenge 2007

Abstract

This paper presents the TerraMax vision systems used during the 2007 DARPA Urban Challenge. First, a description of the different vision systems is provided, focusing on their hardware configuration, calibration method, and tasks. Then, each component is described in detail, focusing on the algorithms and sensor fusion opportunities: obstacle detection, road marking detection, and vehicle detection. The conclusions summarize the lesson learned from the developing of the passive sensing suite and its successful fielding in the Urban Challenge

Looking at Vehicles on the Road: A Survey of Vision-Based Vehicle Detection, Tracking, and Behavior Analysis

“…Using curve-fitting techniques, such as the Hough transform [115] or the RANdom SAmple Consensus (RANSAC) [116], disparity can be modeled as a function of the v coordinate of the disparity map, and pixel locations can be classified as belonging to the ground surface if they fit this model [113]. The v-disparity has been widely used in stereo vision for intelligent vehicles [83], [113], [117]- [122].…”

Robust monocular lane detection in urban environments

“…yellow or blue). This can achieved by converting the image to an appropriate color space (such as HSV), and then isolating the desired component, or performing an approximate color selection (as it is done in [9]). The resulting images (one for each color to handle) can then be separately processed, combining resulting lane markings in a single map, along with information about their color.…”

“…The low-level processing stage derives from the one used during the 2007 DARPA Urban Challenge, described in [9]. First DLD and DLDLD transitions (with the former corresponding to single lane marking, and the latter to double ones) are extracted from the IPM and stored in separate buffers; this operation is fast since the filtering kernel is of constant size (5 and 11 pixels respectively).…”

Day and night-time drive analysis using stereo vision for naturalistic driving studies

“…In [28], [29], the so-called v-disparity is generated and used for separating objects from the ground/road surface. The v-disparity examines the vertical coordinates in a (u,v) image coordinate system and is constructed using a disparity map from, e.g.…”

“…The tracked objects are being classified using prior knowledge of vehicle shapes. Alternatively, objects can be classified using clustering in the disparity map, as seen in [28]. In [32], [33], temporal and scene priors from good conditions are used with the purpose of improving the disparity map in adverse weather conditions, such as night, rain, and snow.…”