A Landmark Detection and Recognition Algorithm for UAV Autonomous Pitching

Landmark detection and recognition algorithm is a very important technology for vision-based Unmanned Aerial Vehicles (UAVs) autonomous pitching. The deformation and rotation of landmarks and the background distraction will be the challenges for detection and recognition. Based on Support Vector Machine (SVM) and the appearance features of landmarks, a landmark detection and recognition algorithm is proposed in this paper. The algorithm presents a landmark detection scheme based on ellipse detection which forms ellipses by optimized arcs and estimates parameters in a decomposed space using Hough transform. To get better edge features, a segmentation is designed to reduce the background noise. Due to the lack of direction information of landmarks in detection procedure, a SVM classifier with a multi-direction voting mechanism is presented for recognition. We expand the training sample set through the affine transformation and make a vote on classification results from multiple directions to achieve accurate landmark recognition. Experimental results show that our landmark detection and recognition algorithm is effective on the UAV platform and the adaptability to the environment is strong.