[Objective] Due to the fact that there are only three bands (red, green and blue) in most UAV(unmanned aerial vehicle)images, a method of extracting water body and monitoring water changes based on texture feature is proposed in order to provide technical support for water conservancy department.[Methods] We used eCognition software to divide the UAV images into multiple scales, and from the image object firstly. The angular second-order moment, mean, entropy and other texture features of the image object was then calculated. These texture features was combined as the image band to highlight the water body. Finally, the water body was extracted by ISO classification method, and the water body extracted from different periods was analyzed to obtain the change area of the water body.[Results] By testing in different regions, the accuracy of the water extraction and detection for water change was above 90%.[Conclusion] The proposed method can effectively improve the efficiency of the water body change monitoring, and reduce the amount of manual labor.