[1] we propose an improved instance segmentation method to extract slender urban road facilities from UAV images.

- a dense anchor ratio with an IoU-balanced sampling strategy (DASS) for proposal generation.

- the balanced finegrained features (BFGF) to improve the boundary segmentation results.

- work accepted by the journal JAG (2023).

[2] SURF 3D modeling via template matching.

- generate 3D points of SRUF by triangulation and predict the location and orientation via least-squares fitting.

- retrieve SURF 3D template from sketch model and texture database.

- work accepted by the journal RS (2021) .

[1] we propose an ensemble method to segment glass façades in oblique aerial images.

- a strategy is adopted to preserve the spatial relationship between glass façade panels.

- edge features are extracted to improve the detector’s performance.

- work accepted by the journal GRSL (2022).

[2] glass façades extraction and repair in photogrammetric 3D models.

- a projection-based method to extract glass façades in photogrammetric 3D building models.

- a mesh denoising method to remedy the geometric deformation and apply texture mapping to correct the distortion.

- applying multiple constraints guarantees the proposed pipeline will not worsen the 3D buidling models.

- work accepted by the journal JAG (2023).

Other research programs

• Vehicle detection in UAV images and moving vehicle removal from photogrammetric 3D urban models (2021.1 - 2021.4)
• Fragments assembly methods to virtually restore the Buddha statue in Yungang Grottoes (2022.4 - 2022.8)
• Assembly-based 3D traffic facility modeling (2022.9-now, ongoing)