Exploring FSCNN + Focal Loss: A Faster Alternative for Road Marking Classification on Mobile LIDAR Sparse Point Cloud Derived Images.
Miguel Luis R. LagahitMasashi MatsuokaPublished in: IGARSS (2022)
Keyphrases
- point cloud
- laser scanner
- image classification
- real world objects
- dense reconstruction
- multi view stereo
- image data
- lidar data
- urban scenes
- aerial images
- three dimensional
- point sets
- rigid registration
- input image
- surface reconstruction
- structure from motion
- point cloud data
- image features
- feature space
- image regions
- satellite images
- feature points
- geometric constraints
- urban areas
- image registration
- training set
- object recognition
- feature extraction
- feature selection
- machine learning
- sparse representation
- icp algorithm