Region-based classification of PolSAR data using radial basis kernel functions with stochastic distances

Abstract

Region-based classification of PolSAR data can be effectively performed by seeking for the assignment that minimizes a distance between prototypes and segments. Silva et al. (2013) used stochastic distances between complex multivariate Wishart models which, differently from other measures, are computationally tractable. In this work we assess the robustness of such approach with respect to errors in the training stage, and propose an extension that alleviates such problems. We introduce robustness in the process by incorporating a combination of radial basis kernel functions and stochastic distances with Support Vector Machines (SVM). We consider several stochastic distances between Wishart: Bhatacharyya, Kullback-Leibler, Chi-Square, Rényi, and Hellinger. We perform two case studies with PolSAR images, both simulated and from actual sensors, and different classification scenarios to compare the performance of Minimum Distance and SVM classification frameworks. With this, we model the situation of imperfect training samples. We show that SVM with the proposed kernel functions achieves better performance with respect to Minimum Distance, at the expense of more computational resources and the need of parameter tuning. Code and data are provided for reproducibility.

Revisiting the effect of spatial resolution on information content based on classification results

“…However, the presence of speckle in SAR images makes classification a non-linearly separable problem; cf. Aghababaee, Amini, and Tzeng (2013); Negri et al (2018), and Fig. 1.…”

Spectral–Spatial-Aware Unsupervised Change Detection With Stochastic Distances and Support Vector Machines

“…The more recent use of stochastic distances has supported several remote sensing applications, including image classification [17]- [19], speckle filtering [20], and change detection [21]. Bhattacharya, Kullback-Leibler, Hellinger, Harmonic, and Triangular are the examples of such stochastic distances.…”

“…Consequently, the test statistic S h φ ( θ, θ k ) is given by (6). The rationale behind (19) is that the probability Pr(χ 2 M > S h φ ( θ , θ k )) > α states that the null hypothesis (i.e., H 0 : θ = θ k ) should not be rejected with significance 1 − α. As a result, if α → 1, the similarity between θ and θ k will be high in order to avoid rejecting H 0 .…”

Spectral-Spatial Aware Unsupervised Change Detection with Stochastic Distances and Support Vector Machines

“…The Bhattacharya distance and test. The more recent use of stochastic distances has supported several Remote Sensing applications, including image classification [12]- [14], speckle filtering [15] and change detection [16]. Bhattacharya, Kullback-Leibler, Hellinger, Harmonic, and Triangular, are examples of such stochastic distances.…”