IEEEXplore - Proceedings of the 2007 IEEE International...

Info iconThis preview shows pages 1–2. Sign up to view the full content.

View Full Document Right Arrow Icon
Object Tracking by introducing Stochastic Filtering into Window-Matching Techniques Flávio B. Vidal and Víctor H. Casanova Alcalde, Member, IEEE, Abstract — This paper describes the development and the application of an object tracking algorithm from a sequence of images. The algorithm is based on window-matching techniques using the sum of squared differences (SSD) as a distance- similarity measure, but adding stochastic f ltering. The algo- rithm is then applied for tracking a vehicle on an urban environment and for tracking the ball on a ping-pong game. It is concluded that incorporating the Kalman f ltering greatly improves the tracking performance. I. INTRODUCTION Visual representation and digital image analysis and processing are usual procedures in areas such as robot- ics, biomedicine, geo-processing and process control among others. Algorithms that combine digital image processing and visual servocontrol techniques are being applied to the solution of complex problems such as object tracking from a sequence of images [1]. Visual tracking can be considered an estimation process acting together with digital image processing techniques. For the estimation process a stochastic f ltering approach using Kalman f lter can be applied [2] and the particle f lter [3]. These estimation approaches can be applied to visual servocontrol in association with window- matching techniques yielding better results[4]. Methodologies for motion detection based on differential techniques can be modi f ed to perform object tracking in a sequence of images [5]. However, these techniques demand numerical calculation of derivatives that could be impracti- cable in circumstances where there is high level of noise, reduced number of frames or the effect of aliasing in the image-acquisition process. An alternative procedure for motion detection is based on window-matching techniques. Reference [6] introduced these methods based on the assessment of the degree of similarity among regions in sequential images, so that an object may be recognized and its position inferred in subsequent frames. The window matching techniques may be applied to object tracking and to other issues in computer vision. In this paper it is proposed an object tracking algorithm that combines the window-matching method of [6] and the functionality of linear stochastic f ltering - Kalman Filter [7]. The window-matching algorithm is modi f ed and a Kalman f ltering stage is coupled to improve the tracking performance. The main idea is building a structure in which a linear form of the Kalman f lter could be coupled to This work was partially supported by CAPES, brazilian organization for higher education. F. B. Vidal and V. H. Casanova Alcalde are with the Electrical Engi- neering Department, University of Brasília, 70910-900 Brasília, DF, Brazil fbvidal@ene.unb.br, casanova@ene.unb.br a window-matching algorithm. The paper discusses issues related to the window-matching procedure, it also discusses problems concerning with the Kalman f ltering initialization and evolution. Some solutions for these issues and problems
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 2
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 02/01/2010 for the course CS 2110 taught by Professor Francis during the Spring '07 term at Cornell University (Engineering School).

Page1 / 6

IEEEXplore - Proceedings of the 2007 IEEE International...

This preview shows document pages 1 - 2. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online