Makine Teknolojileri Elektronik Dergisi

Son Sayı Arşiv
Article Name GPS Assisted Moving Object Detection From Unmanned Ground Vehicles by Segmented Background Modelling
Author Name Güray SONUGÜR**, Barış GÖKÇE*
Address *Afyon Kocatepe Ünivesitesi Bilgi İşlem Daire Başkanlığı, 03200 Afyonkarahisar/TÜRKİYE ** Afyon Kocatepe Ünivesitesi Teknoloji Fak. Mekatronik Müh. Böl., 03200 Afyonkarahisar/TÜRKİYE
Abstract In this study a global positioning system-aided model developed to be able to detect moving objects which located on the route of unmanned ground vehicles by using computer vision systems. Before real-time application, a reference model representing a simple background image was created on all coordinates determined on the route. 160x160 pixel blocks are used to generate the models. Each model associated with its coordinate information was stored into the application database. a number of snapshots are captured and modeled using the same method during the real time movement of the vehicle when the vehicle reached the specified coordinates. Then, these instant snapshots were compared with the reference model of the associated coordinates by using interframe differencing method. End of this process, the pixels which grey levels higher than the threshold were considered as dynamic objects. In the developed approach, moving objects can be detected by the camera on the spatially moving platform. There is no need for complex and costly process to compensate for the noise generated by the moving camera with reference model are held in memory for each coordinate on the route. In the experimental study, the Object detection accuracy in the instant background image frame of the developed system is observed between 100% and 72.7%, also It was determined that low-cost computer system can be used with 0.020 sec processing time per frame and 4% CPU load.
Published in Electronic Journal of Machine Technologies
Issue 13
Volume 2
Pages 49-63
Year 2016
Type Paper
Language Turkish
Download Click here to DOWNLOAD  İndirThis article has been downloaded 3 times.