Camera-based auto-positioning drone

The Indonesian Flying Robot Contest (KRTI) requires a drone to fly autonomously to a designated object. In order to fly autonomously, the robot can use both GPS and camera for navigation. The use of GPS is not always accurate because data from GPS depends on the satellites obtained. Therefore, it is necessary to have a backup system, namely camera-based vision system. The vision methods used in this research are colour segmentation and haar cascade. Colour segmentation has the advantage of good computation by manually classifying images, whereas haar cascade has the advantage of being able to classify images automatically. The results of the design of the navigation method have been validated through simulations and experiments with quadcopter drone hardware. The results show that the navigation performance can work well, both through simulation and hardware experiments. Apart from that, testing object detection with the colour segmentation method and haar cascade can produce a good image processing but still needs to be developed for the level of accuracy.