In recent years, automated underwater robots for infrastructure inspection attract attention. In particular, the underwater robot for quay wall inspections is required to observe a specific object while moving parallel to the wall. The proposed method contributes to robot self-localization by revealing the robot’s motion parallel to the wall based on object detection results and depth information from an assistant acoustic system. Moreover, image enhancement is also included in the proposed flow to suppress reductions in object detection rate due to underwater-specific video degradation. Simulation results show that the proposed method can complete the three steps of video enhancement, object detection, and self-localization at 25 fps for images with 1,280x720 pixels.