Vehicle Tracking and Data Association for Parking Lot Surveillance using Multi-Cameras

• Main Authors: Ming-Xuan Wu, 吳明宣
• Other Authors: Jun-Wei Hsieh
• Format: Others
• Language: zh-TW
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/17113772585276050322

碩士 === 國立臺灣海洋大學 === 資訊工程學系 === 99 === Surveillance system has been broadly used in many places, for example, in the parking lot, in the campus, in the residential district etc. In such kind of large-scale circumstance, it is not enough to track an object by using one camera. It has to link the correspondence among difference cameras to obtain sufficient information about the scene. Then, the surveillance system in this thesis was based on six immovable cameras and images from each of them were non-overlapping. It was easier to be failure for tracking the vehicles because of those photos taken from six network IP cameras with different capture speeds. Therefore, the research question of this paper sounded like: How to predict the vehicle parking information by using Hidden Markov Model in the case of non-overlapping cameras. Even though, it was hard to track the vehicles 100% precisely in the open air, with this system (algorithms) it was still possible to estimate where the vehicle parked in the parking lot. From the experiment result, this system was suitable for use in many different situations and had good results in tracking vehicles.