Moving object detection using multi-layer markov random field model
Document Type
Book Chapter
Publication Title
Pattern Recognition and Big Data
Abstract
In this work, a multi-layer Markov model based spatio-temporal seg- mentation scheme for moving object detection is proposed. Spatial seg- mentation for initial frame is obtained by multi-layer compound Markov random field (MLCMRF) followed by MAP estimation with a combina- tion of simulated annealing (SA) and iterative conditional mode (ICM) techniques. For subsequent frames, a change information based heuristic initialization is proposed for faster convergence of the MAP estimation. For temporal segmentation, label difference based change detection tech- nique is proposed. In one of them, we have considered label frame differ- ence followed by Otsu's thresholding for change detection mask (CDM) generation and in the other, we have proposed label frame difference fol- lowed by entropy associated window selection for CDM generation. This CDM is further modified with the video object planes (VOPs) from the previous frame to identify the locations of the moving objects. The re- sults obtained by the proposed technique are compared with those of the existing state-of-the-art techniques and are found to be better.
First Page
687
Last Page
711
DOI
10.1142/9789813144552_0021
Publication Date
12-15-2016
Recommended Citation
Subudhi, Badri Narayan; Ghosh, Susmita; and Ghosh, Ashish, "Moving object detection using multi-layer markov random field model" (2016). Book Chapters. 244.
https://digitalcommons.isical.ac.in/book-chapters/244
