Morphological and triangular irregular network (TIN) ground filters require setting up different parameters to achieve high accuracy for different terrains. A proposed morphologically iterative TIN (MIT) ground filter only requires maximum building size in the processing of raw light detection and ranging (LiDAR) data. This approach applies morphological and TIN densification in an iterative way for separating ground points from off-ground ones. A radial nearest neighbor is designed to select the surrounding nearest neighbors for each point, and these neighbors are analyzed to define the parameters of a local translational 3D plane surface. Experimental results using ISPRS benchmark datasets show that MIT achieves an average total error of <4.0 % , and an average kappa coefficient of >85 % . Further experimental validation with Hong Kong LiDAR datasets reveals that MIT is effective in detecting dense ground points and robust in various terrain situations.