Many modern applications have to cope with objects comprising vague and uncertain data. Example applications are location determination and proximity detection of moving objects, similarity search and pattern matching in sensor databases or personal identification and recognition systems based on video images or scanned image data. In the recent decade a lot of approaches that address the management and efficient query processing of uncertain data have been published. They mainly differ in the representation of the uncertain data, the distance measures, the types of queries, the query predicates and the representation of the result. This talk gives an overview of methods for effective and efficient similarity queries on uncertain data in feature databases. It especially emphasizes probabilistic similarity ranking methods that exploit the full information given by inexact object representations. Here, we assume the discrete uncertainty model where the uncertain point objects are represented by a set of alternative vectors which are assigned confidence values reflecting the likelihood that the point object is located at the corresponding vector position.