Marine mobile edge computing (MMEC) networks play a crucial role in collecting and analyzing oceanic data. However, the limited energy storage capacity and inconvenient power delivery of underwater sensor nodes (USNs) and surface sink nodes (SSNs) necessitate energy-saving operations. Moreover, the exhaustion of energy in individual USN results in a diminished lifespan of the underwater sensor networks, thereby significantly affecting the accuracy and reliability of oceanic monitoring data. In this paper, we propose a computation offloading scheme that utilizes collaborative underwater sensor data transmission via acoustic and radio frequency links. The main objective is to minimize the energy consumption of unmanned aerial vehicle (UAV)-aided MMEC networks through the joint optimization of USN clustering, SSN computation offloading decisions, and UAV trajectory planning. To address this joint problem, we design a new clustering algorithm followed by the alternating direction method of multipliers (ADMM) optimization algorithm sequentially. Numerical results show that, compared to no clustering and no computation offloading methods, the total energy consumption for the proposed method is reduced by 23.3% and 10.1%, respectively.