Object detection for autonomous search and rescue quadrotor application

Abstract

Indonesia is located in the ring of fire zone. It causes a high risk of disaster threats. Volcanoes and earthquakes are unpredictable sources of natural disaster in a massive size, which many times caused an inaccessible area. We need to prepare how to overcome this situation. Quadrotor is an aerial vehicle that can access the disaster location easily and passing through many obstacles. Its autonomous ability can increase searching efficiency. Object detection will be a primary parameter which responsible to move the quadrotor. In this paper, we proposed to use cascaded Canny Edge with SURF clarified with SIFT to get more robust detection in real-time application. Our method is using image detection in guiding the quadrotor movements. First, we detect the object by cascaded Canny Edge with SURF and ORB, both of them checks whether the target is true or not. When both methods approved that target have been detected, then the process uses only the cascaded Canny Edge with SURF methods to locate the victims and to decrease the image processing time. Output of this image detection method applies into our quadrotor control movements, by using simple Proportional Integral Derivative (PID) control. This visual feedback control guides the quadrotor flies to the target and note the victim position and reported in its log. This paper will also provide data comparison between our method with the only SURF and the only ORB method to do the same task.