Marine mesoscale eddies are a common marine ocean phenomenon. Identifying mesoscale eddies and constructing their three-dimensional structural models can play an important role in shipping, military, and ocean resource development. 3D convolution algorithm was used to optimize the three-dimensional structure of mesoscale eddies. Firstly, the mesoscale eddies were identified from each layer of the global marine physics reanalysis data, and spatial matching method was used to construct a 3D structure of these eddies. Afterwards, in order to compensate for the missing data in the middle layer of the eddy, a sharpened convolution kernel was used to perform 3D convolution on the reanalysis dataset and a complete 3D structure of the mesoscale eddy was reconstructed. Finally, in order to verify the effectiveness of the method, by comparing the temperature map, salt map and current map in the fault layer, the optimized eddy structure was determined to be consistent with the actual situation, and proving that the 3D convolution algorithm has a good optimization effect for solving the "fault" problem of mesoscale eddies three-dimensional structure.