Tracking objects and wing beat analysis methods of a fruit fly with the event-based silicon retina Semesterarbeit

This work is separated into two parts. The first part is about tracking objects with a Kalman filter(KF) on the top of a existing clustering algorithm which tracks moving objects. The second part is about tracking wing edges of a fruit fly(drosophila melangulaster) and doing some analysis like reading out amplitude of the wing beat and its frequency. The framework around this two applications is the same: With the silicon retina objects are recorded. The chip in the camera produces events. An event is a loss or a rise in contrast with a time stamp in the order of mi-croseconds. Through a USB2 interface events are sent to a Java framework where events can be processed. Detailed information about the transient silicon retina can be found in [2]. Movies can of course also be stored and can be played on the screen. The Java framework supports a filter chain where a new filter (in form of a class) can be registered. The framework sends then the data (the amount of data is depending on play-back speed) and informs about special situations like a reset order by the user. A filter can process the data(like filter out noise) and send it ahead the filter chain. This work is about two new filters for this framework. This documentation about the filters do not contain Java specific implementation aspects. The theory of the used mathematical methods like the Kalman filter isn't covered, but the documentation tries to describe their functioning and practical usage by example. Furthermore this documentation describes what algorithms and procedures were implemented in the Java framework but it's not a manual. All implemented options and parameters are shortly described in the appendix.