Mesoscale eddies are ubiquitous in the ocean, and they play a significant role in the transport of momentum, mass, heat, nutrients, and other seawater biogeochemical elements. Many automatic eddy identification methods have been proposed based on satellite sea surface height (SSH) data. Other variables, such as sea surface temperature (SST), chlorophyll (Chl), etc., were also used to identify mesoscale eddies automatically. However, there is still a lack of knowledge about the correlations between these variables over mesoscale eddies. The paper proposes a new methodology for automatically detecting mesoscale eddies from multi-variables (SSH, SST, and Chl) using the deep learning method. It can be summarized in the following steps: 1) constructing a sample database of mesoscale eddies integrating with multi-variables. 2) designing a deep convolution neural network (CNN) model to automatically extract features from multi-variables and training the model with our database to identify the mesoscale eddies. 3) exploring the eddy identification results to investigate the correlations between variables over mesoscale eddies, which also provides references for optimizing the model. This approach has been proven to have the following advantages: 1) the approach is effective and robust under different conditions of mesoscale eddies identification since multi-variables are considered. 2) Not only are eddies detected by the approach but the correlation between these variables over eddies are also explored simultaneously. In conclusion, the proposed method provides a new perspective to detect eddies and is an effective tool for exploring the correlations between multiple variables associated with mesoscale eddies, which helps to evaluate the effect of eddies on the biological changes in the global ocean.