An evolutionary optimized vehicle tracker in collaboration with a detection system

Abstract

In this work a learning algorithm for visual object tracking is presented. As object representation a fast computable set of Haar-like features is used and a weighted correlation is applied for the matching process. The object tracker utilizes the same set of features that is already calculated for object detection and thus it is possible to reuse features for detection and tracking. The feature's weight values are optimized for the tracking purpose by means of evolutionary strategies.Different tests of the object tracker on real-world sequences are presented using vehicles as example objects. Additionally, an object detection system and the integration of the object tracker into that system is described. Besides the system is based on a cascade of boosted classifiers, Haar and Triangle features, an adaptive sliding window and finally a Kalman filter.

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

“…Side profiles of vehicles have been also detected using Haar-like features [22], by detecting the front and rear wheels. Haar-like features have been also used to track vehicles in the image plane [50]. In [51], Haar features were used to detect parts of vehicles.…”

“…Tracking facilitates estimation of motion and prediction of vehicle position in the image plane. Second, tracking enforces temporal coherence, which helps to maintain awareness of previously detected vehicles that were not detected in a given frame [50], while filtering out spurious false positives [49].…”

“…Vision-based tracking is taken one step further using feature-based tracking [145]. In [50], vehicles were detected using Haar-like features and AdaBoost cascade classification. Candidate vehicles' locations were predicted using Kalman filtering in the image plane.…”

“…While early works in monocular on-road vehicle tracking used template matching [24], recursive Bayesian filtering approaches, such as Kalman filtering [50] and particle filtering [75], have become the norm. Estimation of a tracked vehicle's position in the image plane can be augmented by using optical flow, as vision-based vehicle detectors can fluctuate in their pixel locations from frame to frame, even for a static object.…”

Integrated Lane and Vehicle Detection, Localization, and Tracking: A Synergistic Approach

“…Various subsequent studies have applied this classification framework to vehicle detection [33], [34], using Adaboost [35]. Rectangular features and Adaboost were also used in [14], integrated in an active learning framework for improved on-road performance.…”