الفهرس 
chapter 1Only 14 pages are availabe for public view
 |  | from | 139 |  |  |
| | | from | 139 | | |
Abstract
The work introduced by this thesis presents two contributions in the field of crowd density calculation.
The first contribution based on using heat signature in human density calculations with adding two enhancements on previous researches. The two enhancements summarized as dividing the video frames into near and far parts and dividing the temperature of the human shield into two ranges for covered and uncovered skin.
The second contribution is based on combining a background subtraction technique for motion analysis with frame differencing to detect human motion density percentage. In addition, the threshold used in motion analysis methodology is calculated from the motion speed value. A proposed approach in this research is to use the relation between speed and density value to obtain the required speed value without applying the object tracking technique. So, the density value obtained from the heat technique used to detect the speed of moving persons. The results of that comparison showed the high accuracy of the proposed approaches and suggested enhancements for both heat signature and motion analysis crowd estimations that matched the results of manual estimation. The system is capable of estimating the density of any crowd (homogenous or non-homogeneous) also offered accurate crowd classification results that were checked by watching the videos. Using the GUI developed crowd dynamics can be observed to detect any abnormal changes in crowd state. The integration of two different approaches (heat signature and motion analysis) guarantees an accurate refinement of false positives of both.
Last, but not least, the system has the following advantages over it
. No user input is required in real time work of the system
The system does not depend on a specific direction of people motion.