Name | Modified | Size | Downloads / Week |
---|---|---|---|
Parent folder | |||
README.txt | 2011-04-27 | 1.5 kB | |
AutoSegment-WindowsXP.zip | 2011-04-27 | 2.1 MB | |
Totals: 2 Items | 2.1 MB | 0 |
- This project implements an enhanced version of the idea proposed in: http://www.ncc.org.in/download.php?f=NCC2005/S09-01.pdf "A NOVEL SCHEME FOR AUTOMATIC SEGMENTATION OF MOVING OBJECTS FROM VIDEO USING PRESTORED IMAGES" The aim is to learn the background of a scene (without objects), and use this learning to detect any moving objects in the scene. The method is fully automated. It does not require any parameters / variables to be input / tweaked. About the zip file: - The zip file contains a Windows executable bundled with required OpenCV2.1 dlls - Requires access to camera (mapped to Cam 0) - Tested working on Windows XP running on DELL Inspiron, and using integrated webcam - Test procedure: - Confirm PC / laptop has a working camera - Run AutoSegment.exe - The first window (Original) comes up, displaying the image captured from the camera. Ensure that the camera "looks" at a static scene with no moaving objects. Let it run for 10 seconds; this trains the algorithm. - Now press 't' - this invokes the automatic segmentation. The result of the algorithm will be displayed in a new window (Segmented). Also displayed is a live "cluster plot" for the current image. - Press 'q' to quit segmentation and go back to Original. - Press 'q' in Original to exit. Todo: - Introduce file input - Test on OpenCV2.2 - Optimize and improve frame-rate - Enhance algorithm by using more inputs from cluster plot