Corse-to-Fine Road Extraction Based on Local Dirichlet Mixture Models and Multiscale-High-Order Deep Learning.
Ziyi ChenWentao FanBineng ZhongJonathan LiJixiang DuCheng WangPublished in: IEEE Trans. Intell. Transp. Syst. (2020)
Keyphrases
- high order
- mixture model
- deep learning
- road extraction
- multiscale
- unsupervised learning
- aerial images
- remotely sensed images
- satellite imagery
- supervised classification
- em algorithm
- higher order
- dirichlet distribution
- generative model
- dirichlet prior
- gaussian mixture model
- expectation maximization
- supervised learning
- probabilistic model
- image representation
- machine learning
- change detection
- image processing
- natural images
- model selection
- image segmentation
- pairwise
- superpixels
- maximum likelihood
- weakly supervised
- mental models
- semi supervised
- dimensionality reduction
- language model
- markov random field
- pattern recognition
- d objects
- prior knowledge
- high dimensional
- text classification
- feature selection