Summary: | 碩士 === 國立臺灣科技大學 === 電子工程系 === 98 === By the development of image processing and network techniques the use of IP surveillance systems to solve identification problems has become one of the main utilities. It includes the methods of precise measuring of the area objects
such as people, cars, etc, which are very important research items.
The common methods of object management are manual counting and automatic counting. Commonly, the counting of people is done manually by human observers at the entrance or exit. Gate counting can also be done by using the sensors placed above or below the entrance to identify and count incoming and outgoing objects. Network camera based object counting using virtual detection line can be done in any open space and there is no need for sensors or human observers. However, several factors may influence the accuracy of network camera based counting, such as different object speed and sequence, leading to inaccurate results. This thesis introduces a three-line counting method that decreases the errors of video based counting and enhances the accuracy of counting data.
|