Taking a district in Changsha City as an empirical case study, and GIS spatial analysis and emerging space-time analysis are used to study the divergent patterns and environmental characteristics of theft crimes in a three-dimensional perspective combining space and space-time. Kernel density analysis and Dbscan spatial clustering analysis were used to intuitively identify crime hotspots.The spatio-temporal cube model was constructed to analyze the space-time variation characteristics of crime hotspots, eliminate the "pseudo-hotspots" that survive for a short period of time, and identify the stable space-time crime hotspots. The study shows that the spatial distribution of theft crimes in the study area has significant spatial variation and spatial shift, and the crime hotspots are concentrated in the northern region, showing the characteristics of decaying distance from the urban center; the spatio-temporal distribution has a high degree of agglomeration, with only 2.11% of the blocks in the study area gathering 62.4% of theft crimes; combined with the spatial environment characteristics of stable space-time hotspots, the hotspots can be divided into commercial aggregation type, shanty town type and school periphery type, suggesting that crime prevention and control should fully consider the influence of negative factors in the built-up urban environment on criminal behavior. These studies can lay the foundation for the subsequent research on the spatial influence factors of crime, and also provide a realistic basis for the precise prevention and control of crime.