Monitoring ecological environment with remote sensing technology is an effective means of environmental management and sustainable development. Based on Landsat image data and field survey data, this paper used the unweighted voting multi-classifier ensemble method to extract the ecosystem cover data of the water-land interaction area of Nansi Lake from 1987 to 2017, and analyzed the evolution characteristics of the ecosystem structure with the help of the transfer matrix and landscape pattern index. The results showed: (1) The accuracy of multi-classifier ensemble was higher than that of single classifier, with a maximum improvement of 4.17%. (2) In 2006, the change inflection point of ecosystem structure appeared in the study area, which showed that the trend of increase and decrease of ecosystem area changed, and the degree of landscape fragmentation was the lowest. The structure and type transfer of the ecosystem in the study area showed the characteristics of “ring” distribution, and the distribution and transfer of ecosystem at the junction of land and water and land were more complex. (3) Human activities under the intervention of China’s policies, climate change and ecosystem self-regulation were the main driving factors for the change of ecosystem structure in the study area. This study provided technical methods for the further exploration of the ecosystem, and provided constructive guidance for the protection and sustainable development of ecological environment.