Vehicle detection and tracking using homography-based plane rectification and particle filtering

Abstract

Abstract-This paper presents a full system for vehicle detection and tracking in non-stationary settings based on computer vision. The method proposed for vehicle detection exploits the geometrical relations between the elements in the scene so that moving objects (i.e., vehicles) can be detected by analyzing motion parallax. Namely, the homography of the road plane between successive images is computed. Most remarkably, a novel probabilistic framework based on Kalman filtering is presented for reliable and accurate homography estimation. The estimated homography is used for image alignment, which in turn allows to detect the moving vehicles in the image. Tracking of vehicles is performed on the basis of a multidimensional particle filter, which also manages the exit and entries of objects. The filter involves a mixture likelihood model that allows a better adaptation of the particles to the observed measurements. The system is specially designed for highway environments, where it has been proven to yield excellent results.I. INTRODUCTION Nowadays advanced driver assistance systems receive increasing interest both commercially and from the scientific community. In particular, much work has been devoted to the research on techniques for the detection of vehicles in traffic scenarios based on monocular computer vision, due to its low cost and good performance. Aside from traditional knowledge or feature-based approaches, which are dependent on the specific conditions (e.g., weather, illumination), use of motion information is in the basis of many recent works in the field, as this is inherent to the vehicles regardless of the conditions.Usually movement on a planar surface (i.e., the road) is assumed. Nonetheless, environments using non-stationary camera settings, such as the considered traffic environment, feature motion both of the own vehicle (ego-motion) and of the surrounding vehicles. Many methods have been presented that aim at computing ego-motion in order to decouple these motions [1][2]. However, reliable and efficient ego motion estimation in still and open issue.As for vehicle tracking, statistical approaches are adopted in many recent works. In particular, particle filters have emerged as a very powerful tool to perform tracking in a wide variety of applications [3] [4]. Many of these works assume a given initial detection and perform tracking using likelihood models based on appearance templates. However such detection-by-tracking approaches carry the danger of drifting away from the correct targets [5]. Additionally, tracking is often limited to a constant number of objects.

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

“…A similar concept, i.e., dynamic visual maps, was developed in [88] for detecting vehicles and identifying unusual maneuvering in the scene. In [89], a homography matrix was computed between adjacent video frames; image regions that did not cleanly map between frames were assumed to be vehicles. This method seems likely to return many false alarms, but quantitative performance analysis was not included.…”

Vision-based Vehicle Detection Survey

“…Geometrical relations between the elements in the scene so that moving objects were exploited for vehicle detection. In [28], between the successive images homography matrix was computed; for reliable and accurate homography estimation Kalman filtering based probabilistic framework is presented, which in turn allows to detect the moving vehicles in the image. This method seems likely to return many false alarms, but quantitative performance analysis was not included.…”

An Improved ORB Algorithm of Extracting and Matching Features

“…Finding homography matrix [10] between two images and it is a 3 * 3 matrix. Eliminating false matching points with a threshold to keep good matching points in this paper.…”