A variable precision rough set approach to the remote sensing land use/cover classification

Pan

Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci.

et al. 2017

ABSTRACT:Recent advances in remote sensing have witnessed a great amount of very high resolution (VHR) images acquired at sub-metre spatial resolution. These VHR remotely sensed data has post enormous challenges in processing, analysing and classifying them effectively due to the high spatial complexity and heterogeneity. Although many computer-aid classification methods that based on machine learning approaches have been developed over the past decades, most of them are developed toward pixel level spectral differentiation, e.g. Multi-Layer Perceptron (MLP), which are unable to exploit abundant spatial details within VHR images. This paper introduced a rough set model as a general framework to objectively characterize the uncertainty in CNN classification results, and further partition them into correctness and incorrectness on the map. The correct classification regions of CNN were trusted and maintained, whereas the misclassification areas were reclassified using a decision tree with both CNN and MLP. The effectiveness of the proposed rough set decision tree based MLP-CNN was tested using an urban area at Bournemouth, United Kingdom. The MLP-CNN, well capturing the complementarity between CNN and MLP through the rough set based decision tree, achieved the best classification performance both visually and numerically. Therefore, this research paves the way to achieve fully automatic and effective VHR image classification.

Section: Rough Set Decision Tree Based Mlp-cnnmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A Rough Set Decision Tree Based MLP-CNN for Very High Resolution Remotely Sensed Image Classification

Pan

Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci.

et al. 2017

“…Unlike other approaches that deal with vague concepts such as fuzzy set theory, rough set theory provides an objective form of analysis without any preliminary assumptions on membership association, thus, demonstrating power in information granulation [35] and uncertainty analysis [36]. In the field of remote sensing and GIS, rough set theory has been applied in rule-based feature reduction and knowledge induction [30], [37], land use spatial relationship extraction [38], spatio-temporal outlier detection [39], and land cover classification and knowledge discovery [40]. However, description of the uncertainty in remote sensing image classification results, as identified as a need and proposed here, has not been addressed through rough set theory, except for the pioneering work of Ge et al (2009) on classification accuracy assessment.…”

Section: Introductionmentioning

confidence: 99%

“…Rough set theory, as proposed by Pawlak (1982), is an extension of conventional set theory that describes and models the vagueness and uncertainty in decision making [30]. It has been applied in diverse domains such as pattern recognition [31], machine learning [32], knowledge acquisition [33], and decision support systems [34].…”

Section: Introductionmentioning

confidence: 99%

“…In this paper, a variant of rough set theory, variable precision rough set (VPRS) [30], is introduced for the first time to model and quantify the uncertainties in CNN classification of VFSR imagery with a certain level of error tolerance, which is more suitable for the remote sensing domain than standard rough set theory due to its complexity. Through the VPRS theory, these classification uncertainties are partitioned and labelled automatically into positive regions (correct classifications), negative regions (misclassifications) and boundary regions (uncertain areas), respectively.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

VPRS-Based Regional Decision Fusion of CNN and MRF Classifications for Very Fine Resolution Remotely Sensed Images

IEEE Trans. Geosci. Remote Sensing

Sargent

Pan

et al. 2018

Self Cite

A Rough Programming Model Based on the Greatest Compatible Classes and Synthesis Effect

Jin

Jing

et al. 2013

Syst. Res.

The globalization connects different parts of the world tightly, one region can be closely interacted by another region. The globalized environment can become dynamic and turbulent, thus brings uncertainties into decision making. A critical challenge in system science is to deal with the uncertainties such as fuzziness, randomness and roughness of information. In this paper, a programming model in rough sets is presented. First, the characteristics and limitations of the existing rough programming methods are analysed systematically. Second, the necessity and feasibility of developing a new rough programming model is discussed, and the model is developed on the basis of the greatest compatible classes and synthesis effect. Finally, the effectiveness and characteristics of the newly developed model are validated through a case study. The result illustrates that the new programming model is of significance in practical applications, and it makes it possible to take decision preferences into account of the decision-making processes effectively.