The detection of plane landmarks from airborne Lidar points cloud data is considered the key technology for mapping and guidance. Traditional plane detection algorithm based on 3D Hough transform using polar coordinate suffers inconsistent space segmentation problem which causes apex phenomenon when normal of plane is vertical to ground. The new 3D Hough transform is presented with dual space segmentation, utilizing two orthogonal polar coordinates to segment half unit sphere into Gaussian grid. This algorithm can avoid space inconsistent problem. Simulation and real data processing result indicate that the new algorithm can detect plane landmarks effectively without increasing calculation amount.