Detection et Suivi Événementielles
David Reverter Valeiras · HAL (Le Centre pour la Communication Scientifique Directe) · 2017
Neuromorphic event-based cameras are a new type of biomimetic vision sensors, whose principle of operation is inspired by the functioning of the retina. Unlike conventional cameras, these devices do not encode visual information as a sequence of static frames, but as a stream of precisely timestamped events. Every pixel is independent and asynchronously generates events when it detects a sufficient amount of change in the luminance at its corresponding field of view. This frame-free approach avoids redundant sampling of previously known information, resulting in a drastic increase of the temporal resolution without raising the amount of data to process or the energy consumption. This new way of encoding visual information calls for new processing methods, as classical image-based algorithms do not fully exploit the potential of event-based neuromorphic cameras. The main objective of this thesis is the development of truly event-based algorithms for visual detection and tracking.In the first place two plane trackers are introduced. Firstly, a part-based shape tracking is presented. This method represents an object as a set of simple shapes linked by springs. The resulting virtual mechanical system is simulatedwith every incoming event. Next, a line and segment detection algorithm is introduced, which can be employed as an event-based low level feature. Two event-based methods for 3D pose estimation are then presented. The first of these 3D algorithms is based on the assumption that the current estimation is close to the true pose of the object, and it consequently requires a manual initialization step. The second of the 3D algorithms is designed to overcome this limitation.All the presented methods update the estimated position (2D or 3D) of the tracked object with every incoming event. This results in a series of trackers capable of estimating the position of the tracked object with microsecond precision. Experiments are provided in order to test each of the methods, comparing them against other state-of-the-art algorithms. This thesis shows that event-based vision allows to reformulate a broad set of computer vision problems, often resulting in simpler but accurate algorithms.