Geotagging and object detection using dashcam (GEODASH)

• Main Authors: Infante, Rhon Christopher H., Mendoza, Judy May M., Ambrosio, Kristal Mae Anne M., Fernandez, Justine Anne T.
• Format: text
• Language: English
• Published: Animo Repository 2022
• Subjects: Traffic cameras; Global Positioning System; Computer Sciences
• Online Access: https://animorepository.dlsu.edu.ph/etdb_comtech/2; https://animorepository.dlsu.edu.ph/cgi/viewcontent.cgi?article=1001&context=etdb_comtech
• Institution: De La Salle University

Numerous works on detecting and geotagging street objects utilize feeds from fixed traffic cameras However, these systems are limited to detecting objects passing through the traffic cameras’ field of view, and cannot be used to provide a complete picture of objects around an area. In this work, we used a smartphone dashboard camera with a built-in GPS and magnetometer sensors to acquire street view footage allowing the system to perform simultaneous detection and geotagging of target objects. Deep learning approach for object detection and depth estimation for geotagging are used as core algorithms of the system. As opposed to fixed traffic cameras, dashboard cameras allow the crowd participation in data collection. The developed system has 81.12% mAP in object detection and 1.53 m average MAE in geotagging. A visualization application was developed to showcase the potential of the system