In order to flexibly monitor the underwater structure in the complex underwater environments, a small multi-thrusters AUV (Autonomous Underwater vehicle) has been designed with the attitude fusions of monocular vision and IMU (Inertial Measurement Unit). The tasks for detecting the surface faults of the underwater structures in designated area, such as large vessel propellers and offshore oil rig, determine the functional requirements of underwater image processing, flexible attitude control and autonomous navigation for AUV. In this paper, we present a novel design scheme of AUV, including the AUV hardware platform, the corresponding hydrodynamic model for “SAIL”, and the modified attitude fusion algorithm based on monocular vision and IMU. Finally, the experiments are carried out to evaluate the performances of the AUV.