Cultivated Land Information Extraction and Spatial Pattern Analysis in Beijing Based on Deep Learning

Abstract. Cultivated land is the basic resource and material condition for human survival, providing the necessary material basis for agricultural development and agricultural modernization (Tian and Shi, 2024). For this reason, this paper takes Beijing as an experimental area to study the cultivated land extraction method and analyze the distribution pattern of cultivated land and the degree of fragmentation. The results show that (1) the extraction results are evaluated by using confusion matrix and Kappa coefficient, and the Kappa coefficient is obtained to be 0.8358.(2) The cultivated land in Beijing is mainly distributed in the southeast as well as the northwest of the local area of gentle terrain, and the The range of the extremely high value of cultivated land kernel density in 2017–2022 is significantly reduced, the slope of the area with the smallest reduction in the proportion of cultivated land is less than 5°, the reduction in the area of cultivated land in water resource-rich areas is small, and the reduction in the proportion of cultivated land in road-intensive areas is large. (3) Due to various natural factors and human activities, the degree of cultivated land fragmentation in Beijing is increasing, the boundary shape of cultivated land patches tends to be irregular.