An Efficient and Robust Human Classification Algorithm
Abstract
This paper describes an object classification algorithm for infrared videos. Given a detected and tracked object, the goal is to analyze the periodic signature of its motion pattern. We propose an efficient and robust solution similar to frequency estimation techniques in speech processing. Periodic reference functions are correlated with the video signal. In order to capture the frequency response at a given set of period, we explore a local version of DFT. By estimating the periodicity at every pixel, we obtain the overall response for the object, which helps us to make decision robustly. Experimental results for both infrared and visible videos acquired by ground-based as well as airborne moving sensors are presented.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 2004
- Accession Number
- ADA432371
Entities
People
- Isaac Weiss
- Larry S. Davis
- Qinfen Zheng
- Yang Ran
Organizations
- University of Maryland