Mariculture is crucial for grain security and sustainable development, but identifying and counting for the different types of mariculture remains a challenge. In this research, the data of satellite remote sensing, unmanned aerial vehicle (UAV) and shore-based camera is adopted to collect high-definition pictures, and the deep learning algorithm is adopted to detect and position the marine aquaculture areas. Then the classification model of convolutional neural network is adopted to identify the aquaculture species, and the OpenCV model is adopted to assess the area of aquaculture area. The approach obviously improves the efficiency and accuracy of marine aquaculture type identification and area assessment, reduces the subjectivity and error rate of manual statistics, has higher robustness and applicability, can further optimize the algorithm in the future, and is expanded to be applied to the scope on ocean environment monitoring, fishery management and the like.