To facilitate urban planning and management in fast-growing metropolitan areas, it is highly necessary to detect the spatiotemporal changes of different land cover types. This study aimed at identifying Beijing’s land cover types and detecting the characteristics of their spatiotemporal changes using time series remote sensing and GIS techniques from 1978 to 2010. A total of 16 Landsat MSS/TM/ETM+ images were collected during the spring and late summer seasons. After preprocessing the dataset, artificial neural network was used to perform the land cover classification. Consequently, four maps were generated for 1978, 1992, 2000, and 2010, with six classes (agriculture, woodland, grassland, water, urban, and barren land) according to the level I classification scheme. Three transition matrices were constructed to represent all possible changes that occur in the landscape. The results showed that agriculture, barren land, and grassland had an increase in area, while urban, water, and woodland had a reduction within the study area. A total of 2,032.341 km2 agriculture was reduced and 2,359.146 km2 woodland was increased. In the three periods for 1978–1992, 1992–2000, 2000–2010, agriculture had the largest amount of transfer out primarily to urban class around central urban areas and woodland had the most transfer in mainly from barren land in mountainous areas. More importantly, the driving forces analysis including economic development, growth of population and construction areas, and institutional policies was conducted to find out the primary factors inducing the land cover change.