Cloud Detection From RGB Color Remote Sensing Images With Deep Pyramid Networks

TL;DR

This paper presents a deep pyramid network approach for accurate cloud detection in RGB remote sensing images, addressing the challenge of no distinct spectral cloud patterns and achieving superior pixel-level segmentation.

Contribution

The study adapts a deep pyramid network with pre-trained encoder parameters for improved cloud detection in RGB images, a novel approach for this challenging task.

Findings

Outperforms baseline methods in cloud detection accuracy.

Effective even in difficult scenarios like snowy mountains.

Achieves precise pixel-level segmentation and classification.

Abstract

Cloud detection from remotely observed data is a critical pre-processing step for various remote sensing applications. In particular, this problem becomes even harder for RGB color images, since there is no distinct spectral pattern for clouds, which is directly separable from the Earth surface. In this paper, we adapt a deep pyramid network (DPN) to tackle this problem. For this purpose, the network is enhanced with a pre-trained parameter model at the encoder layer. Moreover, the method is able to obtain accurate pixel-level segmentation and classification results from a set of noisy labeled RGB color images. In order to demonstrate the superiority of the method, we collect and label data with the corresponding cloud/non-cloudy masks acquired from low-orbit Gokturk-2 and RASAT satellites. The experimental results validates that the proposed method outperforms several baselines even for hard cases (e.g. snowy mountains) that are perceptually difficult to distinguish by human eyes.