This paper addresses a particular detection problem related to the largest freshwater lake of Central and Western Europe, namely the Lake Balaton. The bed-silt of this shallow water lake (its average depth is 3.3 meters) contains several type of objects: industrial debris, historical vestiges, and in particular, an unknown quantity of unexploded ordnance from the second World War. It is important to localize and classify these objects for precise risk assessment and eventual later removal (not addressed in the paper). The bed-silt of the lake can be characterized as a mud having a light and constantly changing structure. It follows that the shallow water is almost always mixed with the mud. This admixture is constantly maintained by winds and navigating boats, hence the underwater visibility is close to zero during the most part of the year, especially in the depth range close to the bottom. Since the climate makes it possible, the authors propose a special way to explore the bed-silt of the lake, namely the use an autonomous vehicle on the frozen lake. Note that the ice prevents the wind and navigating boats to generate water movement and thus the mud is sedimented. This autonomous vehicle is equipped with a GPS based onboard localization system, with multiple sensor and recording equipment, and with a radio link to its command post. More details about the nature of objects to be detected as well as about the architecture of the detection and localization system are presented in the paper.