📄 Related Paper: [DOI: 10.5194/isprs-archives-XLVIII-2-W8-2024-125-2024]
Developed an automated pipeline for detecting, classifying, and modeling city furniture objects
(traffic lights, bus stops, lampposts, and signs) using Mobile Mapping System (MMS) data.
- Programming: Python
- Machine Learning: YOLOv8
- Point Cloud Processing: KPConv
- Computer Vision: OpenCV
- 3D Data Formats: CityJSON
- Sensor Fusion: LiDAR-Camera Fusion
✔ Implemented YOLOv8 for object detection and classification.
✔ Developed an image-based positioning algorithm using Line of Bearing (LoB).
✔ Used KPConv for point cloud segmentation and Fast Global Registration (FGR) for lamppost classification.
✔ Built a 3D modeling pipeline, integrating CityJSON for urban object representation.
✅ Achieved a positioning accuracy of 0.32m RMSE.
✅ Successfully modeled urban furniture at Level of Detail 4 (LoD4).