Stix-Fusion: A Probabilistic Stixel Integration Technique

Abstract

In summer 2013, a Mercedes S-Class drove completely autonomously for about 100 km from Mannheim to Pforzheim, Germany, using only close-to-production sensors. In this project, called Mercedes Benz Intelligent Drive, stereo vision was one of the main sensing components. For the representation of free space and obstacles we relied on the so called Stixel World, a generic 3D intermediate representation which is computed from dense disparity images.In spite of the high performance of the Stixel World in most common traffic scenes, the availability of this technique is limited. For instance under adverse weather, rain or even spray water on the windshield results in erroneous disparity images which generate false Stixel results. This can lead to undesired behavior of autonomous vehicles.Our goal is to use the Stixel World for a robust free space estimation and a reliable obstacle detection even during difficult weather conditions.In this paper, we meet this challenge and fuse the Stixels incrementally into a reference grid map. Our new approach is formulated in a Bayesian manner and is based on existence estimation methods. We evaluate our new technique on a manually labeled database with emphasis on bad weather scenarios. The number of structures which are detected mistakenly within free space areas is reduced by a factor of two whereas the detection rate of obstacles increases at the same time.

Slanted Stixels: A Way to Represent Steep Streets

“…First, we will comment on works proposing different road scene models. Occupancy grid maps are models used to represent the surroundings of the vehicle (Dhiman et al 2014; Muffert et al 2014;Nuss et al 2015;Thrun 2002). Typically, a grid in bird's eye perspective is defined and used to detect occupied grid cells and then, from this information, to extract the obstacles, drivable area, and unobservable areas from range data.…”

“…Finally, there are some works proposing fast implementations for Stixel computation. The FPGA implementation from Muffert et al (2014) runs at 25 Hz with a Stixel width of 5 pixels, but the authors do not indicate the image resolution. Hernandez-Juarez et al (2017a) present a GPU-accelerated implementation that runs at 26 Hz for a Stixel width of 5 pixels and image resolution of 1024 × 440 pixels, computed using a Semi-Global Matching (SGM) (Hirschmüller 2008) stereo algorithm.…”

“…The Stixel model has been successfully used for automotive vision applications either to decrease parsing time, increase accuracy or both. We can find examples of works using the Stixel representation in different topics such as object recognition (Benenson et al 2012;Li et al 2016), building a grid map over time (Muffert et al 2014) and for autonomous driving (Ziegler et al 2014b). Specifically, for motion planning in the context of autonomous driving, the Stixel model has been used c.f.…”

Slanted Stixels: Representing San Francisco's Steepest Streets

“…The first group is comprised by road scene models. In most cases, occupancy grid maps are used to represent the surrounding of the vehicle [7,17,18,24]. Typically a grid in bird's eye perspective is defined and used to detect occupied grid cells.…”

“…A recent work [3] uses edge-based disparity maps to compute Stixels: this method is also fast but gives inferior accuracy compared to the original Stixels [21]. The FPGA implementation from [17] runs at 25 Hz. Finally, [11] present an embedded GPU implementation that runs at 26 Hz for Stixel widths of 5 px computed using an SGM stereo algorithm also implemented on GPU [10].…”