Moving Object Segmentation Using Optical Flow and Depth Information

Abstract

This paper discusses the detection of moving objects (being a crucial part of driver assistance systems) using monocular or stereoscopic computer vision. In both cases, object detection is based on motion analysis of individually tracked image points (optical flow), providing a motion metric which corresponds to the likelihood that the tracked point is moving. Based on this metric, points are segmented into objects by employing a globally optimal graph-cut algorithm. Both approaches are comparatively evaluated using real-world vehicle image sequences.

3D Visual Information for Dynamic Objects Detection and Tracking During Mobile Robot Navigation

“…To overcome those constraints, the proposed solution consist in using one or more cameras as the only source of information in the system (Talukder & Matthies, 2004), (Williamson, 1998). Essentially vision sensors provide enough information for localization and mapping (Sola et al, 2007) or for describing static and moving objects on the environment (Klappstein et al, 2008). The stereo-vision is one of the most used techniques for reconstructing the 3D (depth) information of a scene from two images, called left and right.…”

Intelligent Traffic Information System a Real-Time Traffic Information System on the Shiraz Bypass

“…Recently, the term optical flow is also used by roboticists, encompassing related techniques from image processing and control of navigation including motion detection, object segmentation, time-to-contact information and motion compensated encoding [4], [5]. Various methods are proposed to compute optical flow and motion detection [6] - [9].…”

User-assisted sparse stereo-video segmentation

“…A key technology to enable the development of such tools is video segmentation, which has recently found many applications such as matting for cinema post-production [4], unsupervised learning [5], navigation for autonomous systems [1] and driver assistance systems [14].…”

“…For applications to stereo vision [1,14] or stereo-3D cinema postproduction, coherence of the clustering across views should be enforced as well. Indeed, a cluster should correspond to a physical object that is imaged on both views.…”

“…The main differences between views arise in binocular half-occlusion areas which are hidden in only one view and should be segmented coherently with respect to the rest of the pixels. Besides, depth information that can be extracted from stereo pairs offers additional cues which can be combined with position, motion and colour to improve the quality of segmentation techniques [7,14]. A general issue with segmentation is that it is an under-constrained problem, but the addition of depth or disparity-based constraints opens new possibilities that have not been explored that much in the literature so far.…”