PSOPaper - AUTOMATICALLY TUNING BACKGROUND SUBTRACTION PARAMETERS USING PARTICLE SWARM OPTIMIZATION Brandyn White and Mubarak Shah University of

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

View Full Document Right Arrow Icon

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

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

Unformatted text preview: AUTOMATICALLY TUNING BACKGROUND SUBTRACTION PARAMETERS USING PARTICLE SWARM OPTIMIZATION Brandyn White and Mubarak Shah University of Central Florida School of Electrical Engineering and Computer Science { bwhite,shah } @cs.ucf.edu ABSTRACT A common trait of background subtraction algorithms is that they have learning rates, thresholds, and initial values that are hand-tuned for a scenario in order to produce the desired subtraction result; how- ever, the need to tune these parameters makes it difficult to use state- of-the-art methods, fuse multiple methods, and choose an algorithm based on the current application as it requires the end-user to be- come proficient in tuning a new parameter set. The proposed solu- tion is to automate this task by using a Particle Swarm Optimiza- tion (PSO) algorithm to maximize a fitness function compared to provided ground-truth images. The fitness function used is the F- measure, which is the harmonic mean of recall and precision. This method reduces the total pixel error of the Mixture of Gaussians background subtraction algorithm by more than 50% on the diverse Wallflower data-set. 1. INTRODUCTION The increasing amount of surveillance cameras in our society places an increasing burden on the security professionals who have to mon- itor them. The field of automated surveillance alleviates this burden by using computers to detect objects in a scene, track their motion over time, identify their type, and recognize high level actions that they perform. Human activity recognition is the most sought-after of these tasks; however, it is generally based on the information gath- ered in the prior stages which places an upper bound on the relia- bility of its decisions. Of these stages, object detection is the start- ing point, with motion based background subtraction being the most popular detection method. In background subtraction, a statistical model of the scene’s background is learned from an image sequence which is used to label pixels corresponding to foreground objects not present in the model. Diverse methods of background subtrac- tion exist[1, 2, 3]. A common trait of background subtraction algorithms is that they have learning rates, thresholds, and initial values that must be tuned in order to produce the desired subtraction result. Despite the importance of parameter selection for each algorithm, end users of- ten overlook or avoid this process. Certain scenarios are more dif- ficult for the background subtraction algorithms and accordingly re- quire more attention during parameter tuning: background occlusion (e.g., crowded scenes), stopped target objects, clutter motion (e.g., moving trees), and illumination changes. Due to this, tuning param- eters for a scene can improve performance; however, the difficulty in tuning parameters and the expert knowledge that it requires makes this an expensive proposition. Since the learning parameters are of- ten managed globally rather than at a pixel level, attention must be...
View Full Document

This note was uploaded on 06/13/2011 for the course CAP 6412 taught by Professor Staff during the Spring '08 term at University of Central Florida.

Page1 / 4

PSOPaper - AUTOMATICALLY TUNING BACKGROUND SUBTRACTION PARAMETERS USING PARTICLE SWARM OPTIMIZATION Brandyn White and Mubarak Shah University of

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