The 2016 Kumamoto earthquake was a series of earthquake events, including the moment-magnitude (Mw) 7.0 mainshock on April 16, 2016 and the Mw 6.2 foreshock on April 14. Due to the strong shaking, more than 8,000 buildings were collapsed and about 30,000 buildings were severely damaged. Geospatial Information Authority of Japan (GSI) acquired high density (5.93 point/m2 ) Lidar data on May 8, 2016, three weeks after the earthquakes. In this study, the pre- and postevent Lidar data were used to detect the collapsed buildings in Mashiki town, Kumamoto Prefecture, Japan, which was one of the most severely affected regions. The pre-event Lidar data were taken on May 15, 2006 with the 0.72 point/m2 density. A report of building damage grades obtained by the field surveys of the Architectural Institute of Japan (AIJ) was introduced as the reference. First, the statistics of height differences within each building outline were calculated. Then the characteristics of the different damage grades were investigated. As a result, the average values of the height differences were adopted to extract collapsed buildings. 618 buildings were extracted as collapsed from 3,408 buildings existed in 2006. Comparing with the reference, 91% collapsed buildings were detected successfully, and the F-score was 0.88.