Overtaking is one of the most dangerous operations in driving. The rear-view mirror is sometimes not consulted by the driver or is momentarily useless because of the blind spot. This paper describes a simple FPGA system based on motion detectors of the fly and rigid-body detection that is able to efficiently segment overtaking cars using a sparse map of features from the visual field of the rear-view mirror. FPGA implementation allows real-time image processing on an embedded system.