Low-Rank Matrix Aided Texture Occlusion Automatic Segmentation of Building Facades

Limited by flight height and scanning angle, the building façade texture is inevitably obscured resulting in missing unmanned aerial vehicle (UAV) data. However, the current segmentation methods have the problem that the target is fixed or requires manual interaction. Therefore, this paper proposes an automatic segmentation method based on low-rank matrix decomposition for facade texture, starting from the low-rank nature of building facade images. Firstly, matrix decomposition is utilized to complete the coarse extraction of the sparse occlusion of the façade texture; then the sparse breakage obtained from decomposition is subjected to image processing such as thresholding segmentation, dilatation, and filling. The result shows that: Our segmentation is as good as the manual marking, achieving the automatic extraction of the occluded façade.