Abstract
With the increase of resolution, effective characterization of synthetic aperture radar (SAR) image becomes one of the most critical problems in many earth observation applications. Inspired by deep learning and probability mixture models, a generalized Gamma deep belief network (g Γ-DBN) is proposed for SAR image statistical modeling and land-cover classification in this work. Specifically, a generalized Gamma-Bernoulli restricted Boltzmann machine (gΓB-RBM) is proposed to capture high-order statistical characterizes from SAR images after introducing the generalized Gamma distribution. After stacking the g Γ B-RBM and several standard binary RBMs in a hierarchical manner, a gΓ-DBN is constructed to learn high-level representation of different SAR land-covers. Finally, a discriminative neural network is constructed by adding an additional predict layer for different land-covers over the constructed deep structure. Performance of the proposed approach is evaluated via several experiments on some high-resolution SAR image patch sets and two large-scale scenes which are captured by ALOS PALSAR-2 and COSMO-SkyMed satellites respectively.
Highlights
As a result of its all-weather and timeless imaging capacity, the synthetic aperture radar (SAR) has become one of the most critical techniques in earth observation, such as land cover and crop classification [1], disaster evaluation [2], urban extraction [3] and so on
Because the deep belief network (DBN) can be used to learn the statistical dependencies among each units of the observed variables, a generalized Gamma distribution based DBN is proposed for high-resolution SAR images
By stacking the restricted Boltzmann machine (RBM) in a hierarchical manner, a gΓ-DBN is proposed to learn the discriminative information from high-resolution SAR images
Summary
As a result of its all-weather and timeless imaging capacity, the synthetic aperture radar (SAR) has become one of the most critical techniques in earth observation, such as land cover and crop classification [1], disaster evaluation [2], urban extraction [3] and so on. With the increasing number of SAR sensors, many images are produced in high-quality to provide precise information on the observed land covers. To obtain a solid comprehension of SAR images, one of the most critical problems is how to effectively distinguish different land-covers. A variety of approaches has been proposed to characterize SAR images in recent years, including backscattering coefficients [4], statistical features [5,6], texture descriptors [7], bag-of-words [8], sparse representation [9] and so on. It is difficult to apply these features on SAR images which are captured by different sensors. Much more efficient statistical distributions should be developed to characterize such a complex scene
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
