Computer vision applied to localisation and target detection has been a field of study in the literature for some years. Conventionally, MAVs (Micro Aerial Vehicles) used to rely solely on distance sensors and ran on simple and modest embedded devices. The huge increase of computational power made possible the use of more complex computer vision algorithms for real-time embedded applications. This work focuses on evaluating different types of computationally demanding algorithms, such as accumulator-based image transforms (Hough, Radon), mathematical morphology and Monte Carlo approaches, to process data fed by a single camera in order to aid a MAV navigation through an obstacle course. These algorithms are put through various tests that seek to emulate different scenarios for the hoop detection, each algorithm evaluated on 3 key metrics: processing time, ratio of valid detections and accuracy. The best performing algorithm is analysed on its behaviour on the challenging hoops section of the IMAV 2018 indoor challenge, where lighting is precarious and the camera capture is subject to motion and vibration caused by the MAV. Finally, suitable algorithms are implemented with navigation controller for the MAV, with the final goal of passing through a hoop autonomously. The algorithms are, then, evaluated on the time taken to execute the task and flight behaviour in scenarios similar to those previously tested.