This document discusses motion detection techniques for real-time applications. It describes using GPUs for parallel processing to meet frame rate requirements of 30 fps or more. Methodologies discussed include object detection, tracking objects across video frames by changes in location relative to background, and morphology-based tracking using contour registration, feature vectors, and segmentation. Applications mentioned are traffic surveillance and mobile robots. Limitations include dynamic environments, abrupt motions, and computational expense.
1 of 13
Download to read offline
More Related Content
Real Time Motion Object Tracking Using GPU
1. Author: Lubomir Riha and Hoda El-Sayed
Resource: https://ieeexplore.Ieee.Org/document/6126628
2. WHAT IS MOTION DETECTION?
Motion detection is an important computer vision problem that has been
used in different applications including surveillance. Most of the applications
require fast processing due to their real time nature. The GPU (graphic
processing unit) is used as a cost-efficient tool that provides great
opportunities for parallel processing.
3. HARDWARE IN REAL-TIME TRACKING
• Memory : tracking system encountering limited memory problems
• Frame rate : 30 fps
• Processors :
i. Allow saturated arithmetic operation
ii. Powerful operation ability
iii. Can do several memory accesses in a single instruction
4. METHODOLOGIES
• Object detection and tracking
• In a video sequence an object is said to be in motion, if it is changing
its location with respect to its background
• The motion tracking is actually the process of keeping tracks of that
moving object in video sequence I.E. Position of moving object at
certain time etc.
9. • Background estimation: Image differencing and thresholding
• Object registration :
i. Contours are registered
ii. Width, height and histogram are recorded for each contour
• Feature vector : each object represented by a feature vector
• Combine with different features, color features and texture information
• Use segmentation for tracking object
Morphology based object tracking