Unsupervised Land Cover Classification Of Hybrid And Dual Polarized Images Using Deep Convolutional Neural Network
Unsupervised Land Cover Classification Of Hybrid And Dual Polarized Images Using Deep Convolutional Neural Network
Embedded Files
We propose an unsupervised learning algorithm to cluster hybrid polarimetric SAR images, and dual polarized SAR images using deep framework. We use feature extraction layers of VGG16 model with batch normalization, which is trained with small patches derived from the hybrid polarimetric SAR images. It uses an entropy based loss function, and an adaptive learning rate optimization algorithm, for training. Broadly, the patches are segmented into three classes, namely, surface, volume, and double-bounce, which are defined with reference to the SAR scattering characteristics. Further, we classify volume into dense forest region and agricultural crop fields. We also observe mixed classes between volume and double-bounce, mainly covering the settlements surrounded by areas covered by tall trees. Furthermore, we use transfer learning for generating the labels for dual polarized images by using the learned weights of hybrid polarized image model.
We propose an unsupervised learning algorithm to cluster hybrid polarimetric SAR images, and dual polarized SAR images using deep framework. We use feature extraction layers of VGG16 model with batch normalization, which is trained with small patches derived from the hybrid polarimetric SAR images. It uses an entropy based loss function, and an adaptive learning rate optimization algorithm, for training. Broadly, the patches are segmented into three classes, namely, surface, volume, and double-bounce, which are defined with reference to the SAR scattering characteristics. Further, we classify volume into dense forest region and agricultural crop fields. We also observe mixed classes between volume and double-bounce, mainly covering the settlements surrounded by areas covered by tall trees. Furthermore, we use transfer learning for generating the labels for dual polarized images by using the learned weights of hybrid polarized image model.
Semi-supervised Classification of Paddy Fields from Dual Polarized Synthetic Aperture Radar (SAR) images using Deep Learning
Semi-supervised Classification of Paddy Fields from Dual Polarized Synthetic Aperture Radar (SAR) images using Deep Learning
The semi-supervised algorithm to detect paddy fields is divided into two parts, unsupervised and supervised. Feature extraction layers of Visual Geometry Group (VGG16) model is used to segregate patches into five clusters using an entropy based loss function. Depending on the availability of the ground truth, data is sampled (stratified sampling), and only a certain amount of data is further used for the supervised section. Clusters showing higher similarity (Jaccard/Tanimoto test) with the paddy or non-paddy classes are connected to a three layer neural network used for supervised method.
The semi-supervised algorithm to detect paddy fields is divided into two parts, unsupervised and supervised. Feature extraction layers of Visual Geometry Group (VGG16) model is used to segregate patches into five clusters using an entropy based loss function. Depending on the availability of the ground truth, data is sampled (stratified sampling), and only a certain amount of data is further used for the supervised section. Clusters showing higher similarity (Jaccard/Tanimoto test) with the paddy or non-paddy classes are connected to a three layer neural network used for supervised method.
Building Cluster-Class Association for Detecting paddy fields under Semi-Supervised deep learning framework
Building Cluster-Class Association for Detecting paddy fields under Semi-Supervised deep learning framework
An orchestration of two learning methods is proposed by exploiting cluster-class associations. The architecture has two parts, in the front-end, a CNN is trained to perform clustering. The response of clustering becomes input to the classifier, back-end of the architecture. The classifier is a selectively connected neural network, where every node in the hidden layer represents a class. The whole architecture is then trained with a limited set of labeled data. During training, the weights of the front-end architecture are not updated.
An orchestration of two learning methods is proposed by exploiting cluster-class associations. The architecture has two parts, in the front-end, a CNN is trained to perform clustering. The response of clustering becomes input to the classifier, back-end of the architecture. The classifier is a selectively connected neural network, where every node in the hidden layer represents a class. The whole architecture is then trained with a limited set of labeled data. During training, the weights of the front-end architecture are not updated.
Unsupervised Land Cover Classification Of Hybrid PolSAR Images Using Deep Network
Unsupervised Land Cover Classification Of Hybrid PolSAR Images Using Deep Network
We devise an unsupervised patch based learning method to cluster hybrid polarimetric SAR images. Small patches are extracted from the image data set for training using an entropy based loss function. Each patch represents a particular class (based on the scattering observed) and hence, the patches are selected based on the resolution of the image to make sure that the scattering does not change within a particular patch.
We devise an unsupervised patch based learning method to cluster hybrid polarimetric SAR images. Small patches are extracted from the image data set for training using an entropy based loss function. Each patch represents a particular class (based on the scattering observed) and hence, the patches are selected based on the resolution of the image to make sure that the scattering does not change within a particular patch.
Page updated
Google Sites
Report abuse