This paper presents a research in progress towards autonomous underwater robot manipulation. Current research in underwater robotics intends to increase the autonomy of intervention operations that require physical interaction. Autonomous grasping is still a very challenging skill, especially in underwater environments, with highly unstructured scenarios, limited availability of sensors and adverse conditions that affect the robot perception and control systems in various degrees. To tackle those issues, we propose the use of vision and segmentation techniques that aim to improve the specification of grasping operations on underwater primitive shaped objects. Several sources of stereo information are used to gather 3D information in order to obtain a model of the object. Using a RANSAC primitive shape recognition algorithm, the model parameters are estimated and a set of feasible grasps are computed. This approach is validated in simulation and the quality of different 3D reconstructions from both real and virtual scenarios is analyzed.