Learning to detect moving shadows in dynamic environments

Research output: Contribution to journalArticle

74 Citations (Scopus)

Abstract

We propose a novel adaptive technique for detecting moving shadows and distinguishing them from moving objects in video sequences. Most methods for detecting shadows work in a static setting with significant human input. To remove these limitations, we propose a more general semi-supervised learning technique to tackle the problem. First, we exploit characteristic differences in color and edges in the video frames to come up with a set of features useful for classification. Second, we use a learning technique that employs Support Vector Machines and the Co-training algorithm, that relies on a small set of human-labeled data. We observe a surprising phenomenon that Co-training can counter the effects of changing underlying probability distributions in the feature space. From the standpoint of detecting shadows, once deployed, the proposed method can dynamically adapt to varying conditions without any manual intervention, and performs better classification than previous methods on static and dynamic environments alike. The strengths of the proposed technique are the small quantity of human labeled data required, and the ability to adapt automatically to changing scene conditions.

Original languageEnglish (US)
Pages (from-to)2055-2063
Number of pages9
JournalIEEE Transactions on Pattern Analysis and Machine Intelligence
Volume30
Issue number11
DOIs
StatePublished - Oct 9 2008

Fingerprint

Dynamic Environment
Co-training
Supervised learning
Probability distributions
Support vector machines
Adaptive Techniques
Semi-supervised Learning
Alike
Training Algorithm
Feature Space
Color
Moving Objects
Support Vector Machine
Probability Distribution
Human
Learning

Keywords

  • Co-training
  • Population drift
  • Semisupervised learning
  • Shadow detection

Cite this

Learning to detect moving shadows in dynamic environments. / Joshi, Ajay J.; Papanikolopoulos, Nikolaos P.

In: IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 30, No. 11, 09.10.2008, p. 2055-2063.

Research output: Contribution to journalArticle

@article{0dc35136e22441ca901a6f8751aa8f26,
title = "Learning to detect moving shadows in dynamic environments",
abstract = "We propose a novel adaptive technique for detecting moving shadows and distinguishing them from moving objects in video sequences. Most methods for detecting shadows work in a static setting with significant human input. To remove these limitations, we propose a more general semi-supervised learning technique to tackle the problem. First, we exploit characteristic differences in color and edges in the video frames to come up with a set of features useful for classification. Second, we use a learning technique that employs Support Vector Machines and the Co-training algorithm, that relies on a small set of human-labeled data. We observe a surprising phenomenon that Co-training can counter the effects of changing underlying probability distributions in the feature space. From the standpoint of detecting shadows, once deployed, the proposed method can dynamically adapt to varying conditions without any manual intervention, and performs better classification than previous methods on static and dynamic environments alike. The strengths of the proposed technique are the small quantity of human labeled data required, and the ability to adapt automatically to changing scene conditions.",
keywords = "Co-training, Population drift, Semisupervised learning, Shadow detection",
author = "Joshi, {Ajay J.} and Papanikolopoulos, {Nikolaos P}",
year = "2008",
month = "10",
day = "9",
doi = "10.1109/TPAMI.2008.150",
language = "English (US)",
volume = "30",
pages = "2055--2063",
journal = "IEEE Transactions on Pattern Analysis and Machine Intelligence",
issn = "0162-8828",
publisher = "IEEE Computer Society",
number = "11",

}

TY - JOUR

T1 - Learning to detect moving shadows in dynamic environments

AU - Joshi, Ajay J.

AU - Papanikolopoulos, Nikolaos P

PY - 2008/10/9

Y1 - 2008/10/9

N2 - We propose a novel adaptive technique for detecting moving shadows and distinguishing them from moving objects in video sequences. Most methods for detecting shadows work in a static setting with significant human input. To remove these limitations, we propose a more general semi-supervised learning technique to tackle the problem. First, we exploit characteristic differences in color and edges in the video frames to come up with a set of features useful for classification. Second, we use a learning technique that employs Support Vector Machines and the Co-training algorithm, that relies on a small set of human-labeled data. We observe a surprising phenomenon that Co-training can counter the effects of changing underlying probability distributions in the feature space. From the standpoint of detecting shadows, once deployed, the proposed method can dynamically adapt to varying conditions without any manual intervention, and performs better classification than previous methods on static and dynamic environments alike. The strengths of the proposed technique are the small quantity of human labeled data required, and the ability to adapt automatically to changing scene conditions.

AB - We propose a novel adaptive technique for detecting moving shadows and distinguishing them from moving objects in video sequences. Most methods for detecting shadows work in a static setting with significant human input. To remove these limitations, we propose a more general semi-supervised learning technique to tackle the problem. First, we exploit characteristic differences in color and edges in the video frames to come up with a set of features useful for classification. Second, we use a learning technique that employs Support Vector Machines and the Co-training algorithm, that relies on a small set of human-labeled data. We observe a surprising phenomenon that Co-training can counter the effects of changing underlying probability distributions in the feature space. From the standpoint of detecting shadows, once deployed, the proposed method can dynamically adapt to varying conditions without any manual intervention, and performs better classification than previous methods on static and dynamic environments alike. The strengths of the proposed technique are the small quantity of human labeled data required, and the ability to adapt automatically to changing scene conditions.

KW - Co-training

KW - Population drift

KW - Semisupervised learning

KW - Shadow detection

UR - http://www.scopus.com/inward/record.url?scp=54749131598&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=54749131598&partnerID=8YFLogxK

U2 - 10.1109/TPAMI.2008.150

DO - 10.1109/TPAMI.2008.150

M3 - Article

VL - 30

SP - 2055

EP - 2063

JO - IEEE Transactions on Pattern Analysis and Machine Intelligence

JF - IEEE Transactions on Pattern Analysis and Machine Intelligence

SN - 0162-8828

IS - 11

ER -