Research on extraction method of non-food cropland based on temporal remote sensing and continuous change detection

Under the background of rapid urbanization, the food security is threatened by the problem of democratization of cultivated land, and it is the basis for formulating related policies on cultivated land protection to accurately identify the degree and change law of cultivated land non-grain cultivation. In this paper, the long time series (1991-2021) multisource remote sensing data was used to obtain the typical temporal trajectory of cultivated land change. The continuous change detection and classification (CCDC) algorithm was used to identify the temporal trajectory breakpoints of each pixel, and the non-grain distribution and spatial-temporal evolution of regional cultivated land were obtained. The method was verified by taking Pujiang County of Chengdu as the research area. The results show that the democratization of cultivated land in Pujiang County mainly consists of forest and fruit land and water surface of pits and ponds, which presents a radiative development. From 2015 to 2018, the non-grain increased at the fastest rate, reaching a peak of 2662.11 hectares in 2017, and decreasing from 2019 to 2021, which is closely related to the stage economic and social conditions. This method can accurately identify the time nodes of cultivated land democratization, reduce the misjudgment caused by seasonal changes, and is suitable for large-scale and long time series change analysis, and can provide countermeasures and suggestions for curbing cultivated land democratization.