With the increase of marine activity, protecting marine infrastructures from terrorist threats has become a main concern. We propose to develop and demonstrate a novel prototype for reliable, real-time detection of diver and mines. Our system combines acoustic remote detection with verification using pattern recognition on underwater imagery. First, we analyse the acoustic reflections to localize a target that fits the pattern of a diver or a submerged mine. Then, in case the release or presence of a mine is suspected, a vehicle is dispatched to first inspect the target through its optical and sonar systems, and then to send the processed information back through underwater acoustic communication.