Stereo obstacle detection in challenging environments: The VIAC experience

Abstract

Abstract-Obstacle detection by means of stereo-vision is a fundamental task in computer vision, which has spurred a lot of research over the years, especially in the field of vehicular robotics. The information provided by this class of algorithms is used both in driving assistance systems and in autonomous vehicles, so the quality of the results and the processing times become critical, as detection failures or delays can have serious consequences. The obstacle detection system presented in this paper has been extensively tested during VIAC, the VisLab Intercontinental Autonomous Challenge [1], [2], which has offered a unique chance to face a number of different scenarios along the roads of two continents, in a variety of conditions; data collected during the expedition has also become a reference benchmark for further algorithm improvements.

Real-time obstacle detection using range images: processing dynamically-sized sliding windows on a GPU

“…A similar approach was used in ref. [8], where the image is divided into horizontal segments and sub-sampled according to the distance represented by such segments. In, 9 a set of discrete parameters was employed.…”

“…In practice, the method is robust and yields great results even when the "method assumption" is violated, as the results from refs. [7], [8], [10], and [12] suggest. We also have the benefit of the projection assumption but, as the original authors, rather than projecting a triangle or trapezoid we employ a rectangular region (see Fig.…”

“…[8] and [10]. It is also worth mentioning that by reading the original authors' papers, 7,12 one may infer that by increasing the projection size the method assumption would be somehow alleviated.…”

“…It was successfully employed in a long autonomous navigation experiment, in which few geometric assumptions could be made. 8 In practice, its only drawback is its high computational cost; it may take the method seconds to detect obstacles in a practical-sized range image. Researchers usually rely on approximations, [8][9][10] so it can perform within a reasonable time setting.…”

Performance analysis of stereo reconstruction algorithms

“…Instead of using mutual information as the pixel-wise matching function, as it is done in the original work [9], the Hamming distance of the Census transform of a 5 x 5 window cropped around each pixel has been computed, since it provides similar results [11] while reducing the overall processing burden [12]. Each position C(p, d) of the cost volume is then initialized with the number of differing bits between the corresponding transformed values of the left and right images.…”