This paper is concerned with the development of information processing for robust detection of target maneuvers. Indeed, the price to pay for having good performance for target maneuver detection is to accept that the event that a target maneuver will be falsely detected will be not so infrequent. Classical remedies exist and perform satisfactorily like the test based on consecutive exceeds. It is proved here that this test is robust and emphasis is put on performance analysis via Markov chain analysis. It is then possible to derive general method for optimizing the test so as to satisfy operational requirements.