Remote Sensing Image Classification Using Attribute Filters Defined Over the Tree of Shapes

Abstract

Abstract-Remotely sensed images with very high spatial resolution provide a detailed representation of the surveyed scene with a geometrical resolution that at the present can be up to 30 cm (WorldView-3). A set of powerful image processing operators have been defined in the mathematical morphology framework. Among those, connected operators (e.g., attribute filters) have proven their effectiveness in processing very high resolution images. Attribute filters are based on attributes which can be efficiently implemented on tree-based image representations. In this work, we considered the definition of min, max, direct and subtractive filter rules for the computation of attribute filters over the tree of shapes representation. We study their performance on the classification of remotely sensed images. We compare the classification results over the tree of shapes with the results obtained when the same rules are applied on the component trees. The random forest is used as a baseline classifier and the experiments are conducted using multispectral data sets acquired by QuickBird and IKONOS sensors over urban areas.

Local Feature-Based Attribute Profiles for Optical Remote Sensing Image Classification

“…And secondly, by using only one tree of shapes to replace both min-tree and max-tree [6], the feature dimension of SDAPs is reduced to half of that of APs. Consequently, SDAPs have been proved to be more efficient than APs in many research studies [13], [19], [20].…”

“…For attribute filtering, we exploited two attributes including the area and the moment of inertia. Ten area thresholds were adopted for the Reykjavik data as proposed by several papers [20], [39], [40]. For the Pavia University data, fourteen thresholds were automatically computed according to [24].…”

Classification of VHR remote sensing images using local feature-based attribute profiles

“…Any other local features can be extracted to tackle more complex VHR image scenes in future work. Also, the concept of LFAPs can be applied to the extended APs [6] for hyperspectral image classification as well as to the recently proposed selfdual APs [5] which has been proved to outperform APs.…”

“…To construct the APs, three common attributes were considered including the area, the standard deviation and the moment of inertia. The threshold values were set to be similar to previous studies on the same data [5]. For classification stage, the Random Forest was employed by setting the number of trees equal to 200.…”

A Novel Semi-supervised Classification Method Based on Class Certainty of Samples

“…The traditional RS images classification technologies are mainly based on unsupervised learning [4]and supervised learning [5]. Unsupervised methods cluster the similar samples together by exploiting the characteristic distribution of samples.…”