Minimize Multiclass object detection in aerial and satellite images

Date added: 30 July 2013

Organisation: DLR

Location: Oberpfaffenhofen, Germany


The Photogrammetry and Image Analysis department of DLR has developed a system for monitoring the road traffic from aerial images in real-time (the project VABENE: and a system for monitoring ships from optical satellite images in near real-time (the project OpsServe). In both systems a crucial part is a reliable and fast visual object detector. In the next phases of the projects the existing detector needs to be extended for multiple object classes.

The proposed master thesis focuses on developing a visual object detector which detects multiple object types (e.g. vehicles, ships) on aerial and satellite images.

The main tasks are:

  • Studying the state of the art computer vision methods for object detection and classification
  • Implementing the detector C++ program using the openCV computer vision library
  • Evaluating the performance of the detector