Land surface water is an important part of the water cycle and is invaluable for human survival. Timely monitoring of surface water and the delivery of data on the dynamics of surface water are essential for policy and decision-making processes. The rapid, accurate, and automated extraction of aquaculture water is significant for assessing its role in fishery informatization and scientific management. Remote sensing technology has the advantages of macroscopic, real-time, dynamic access to land-surface information, and can be used to obtain accurate spatial and temporal distribution and dynamic changes of aquaculture water.Commonly used spectral index-and threshold-based approaches are highly efficient, but they require carefully selected threshold values that vary depending on the region being imaged and on the atmospheric conditions. Moreover, these indexes easily mistake other targets with similar spectral characteristics for surface water, such as shadow. Here, we developed an integrated classification model for aquaculture water extraction, which combines Random Forest (RF), Support Vector Machine (SVM), and Back Propagation Neural Network (BP), and the result was voted by above three methods. The input for this model was spectral features and texture features from the domestic GF-1 and ZY-3 high-resolution remote sensing image, calculated by Normalized Difference Water Index (NDWI) and Gray-Level Co-occurrence Matrix (GLCM). Moreover, the shadow detection method, Enhanced Shadow Water Index (ESWI), was proposed for removing shadows from mountains and buildings.We tested the accuracy of the new model using GF-1 images in Chaohu City. The results indicate that the integrated classification model performed significantly better than other methods with total accuracy by 97.4%, Kappa by 0.87, omission error by 3.7% and commission error by 6.4%, respectively. In addition, the details showed that this algorithm can effectively distinguish shadows of high buildings and mountains from water bodies to improve the overall accuracy. Moreover, this new algorithm may also be suitable for monitoring the changes of aquaculture water. Spatiotemporal changes of aquaculture water in the experimental area from 2013 to 2016 were evaluated using ZY-3 and GF-1 images. The aquaculture water area was 423.9 km² from GF-1 imagery in 2016 compared with 396.7 km² from ZY-3 imagery in 2013, and the water area increased 6.9% for 3 years.The main purpose of this study was to devise a model that improves water extraction accuracy, particularly in areas with shadows, which is often a major cause of low classification accuracy. It is believed that this algorithm, which combines an integrated classification model with a shadow detection method, can significantly improve the accuracy of aquaculture water detection, especially in mountainous and urban areas where deep shadow caused by the terrain and buildings is an important source of error. This algorithm also provides a foundation for the automatic renewal of a larger range of aquaculture water and should promote the integration of high-resolution remote sensing imagery in hydrological applications.