The changes of the waterlines alongside ship hulls in model ship tests provide important information to estimate the resistance performance of the full- scale ships and to improve its hull form design. This Letter presents a systematic procedure for detecting a waterline and reconstructing its three- dimensional (3D) geometry using a stereo vision system. The proposed method utilises the information from nondistinctive geometric features and shapes characterised in the waterline around a model ship without using any specified colours or artificial landmarks. For reconstructing the 3D geometry of the detected waterline, a vertical stereo system and a direct disparity estimation method are employed along with the consideration of uncertainties in the stereo correspondence matching. Experimental results are shown to demonstrate the feasibility of the proposed method using an image dataset obtained in a model ship test in a towing tank environment.