We propose a novel learning strategy called Global-Local Motion Pattern Classification (GLMPC) to localize pedestrian-like motion patterns in videos. Instead of modeling such patterns as a single class that alone can lead to high intra-class variability, three meaningful partitions are considered - left, right and frontal motion. An AdaBoost classifier based on the most discriminative eigenflow weak classifiers is learnt for each of these subsets separately. Furthermore, a linear three-class SVM classifier is trained to estimate the global motion direction. To detect pedestrians in a given image sequence, the candidate optical flow sub-windows are tested by estimating the global motion direction followed by feeding to the matched AdaBoost classifier. The comparison with two baseline algorithms including the degenerate case of a single motion class shows an improvement of 37% in false positive rate.
Pedestrian Detection Using Global-Local Motion Patterns
Asian Conference on Computer Vision ; 2007 ; Tokyo, Japan November 18, 2007 - November 22, 2007
2007-01-01
10 pages
Article/Chapter (Book)
Electronic Resource
English
Support Vector Machine , Motion Pattern , Global Motion , Frontal Motion , Linear Support Vector Machine Computer Science , Image Processing and Computer Vision , Computer Imaging, Vision, Pattern Recognition and Graphics , Pattern Recognition , Artificial Intelligence , Biometrics , Algorithm Analysis and Problem Complexity
Can appearance patterns improve pedestrian detection?
IEEE | 2015
|Pedestrian Detection via Periodic Motion Analysis
British Library Online Contents | 2007
|Pedestrian detection from non-smooth motion
IEEE | 2015
|