The current global pandemic has laid bare the importance of national food security to human survival. Many cultivated lands in the hilly, mountainous, and other marginalized areas have been abandoned on a large scale, resulting in a tremendous waste of agricultural resources, thereby threatening national food security. Here, we studied abandoned farmland in Xingning City, a mountainous area in northern Guangdong province. According to the "seeding-growing-harvesting" life cycle of cultivated plots, spatial superposition method and remote sensing change detection method were applied to identify abandoned arable land. Logistic regression model was used to reveal the influencing factors and occurrence mechanism of abandoned cropland at plot scale, and cluster analysis was used to discuss the classification and management strategies. Result showed that 16.83% of the cultivated land in the study area was severely abandoned, attributed to poor location, poor basic conditions, and fragmentation of the land. Further, the abandoned farmland was divided into output-driving type, cultivation condition-driving type, and plot-condition driving type. Based on these types, we proposed some countermeasures, such as adjusting agricultural structures, tamping agricultural infrastructures, strengthening land circulation, popularizing appropriate scale operations. These measures provide a reference to effectively curb abandoned farmland and improving the utilization efficiency of cultivated land, especially in recent years.