Learning from label proportions for SAR image classification

Wenxian Yu,Yongke Ding,Yuanxiang Li

doi:10.1186/s13634-017-0478-8

Abstract

Synthetic aperture radar (SAR) image classification plays a key role in SAR interpretation. Due to the cost and difficulty of truth labeling for SAR images, the newly labeled samples available for image classification are very limited. This paper focuses on defining a new sample labeling method to solve the problem of truth acquisition for training data in SAR image classification. An efficient classification framework for high-resolution SAR images is presented in this paper, which is built on learning from uncertain labels. We use grid labeling for rapid training data acquisition by assigning a label to a group of neighboring pixels at a time. A novel SVM-based learning model is proposed to optimize the uncertain training data within the constraints of label proportions in each group and then predict the label of each sample for the test data based on the optimized training set. This work intends to explore a rapid labeling method called grid labeling for efficient training set definition and apply it to large-scale SAR image classification. The model demonstrates good performance in both accuracy and efficiency for scene interpretation of high-resolution SAR images.

Highlights

Synthetic aperture radar (SAR) image classification or land use/land cover (LULC) mapping plays an important role in many and diverse SAR applications [1]
Inspired by the idea of learning from label proportions, we firstly introduce grid labeling to get the truth of training data more efficiently by giving the proportions of the labels in each group
4 Learning from grid labeling we focus on exploring SAR image classification method by learning from majorclass Grid labeling (GL)

Summary

Introduction

Synthetic aperture radar (SAR) image classification or land use/land cover (LULC) mapping plays an important role in many and diverse SAR applications [1]. Supervised learning is one of the most popular methods for SAR image classification. TL-based classification methods are designed to achieve high classification accuracy with relatively small number of labeled samples from the new image (target domain) by efficient reuse of the training data from the previous different but relative images (source domain) [5,6,7]. AL-based classification methods focus on reducing the number of samples to be labeled by a human expert through iteratively choosing the most informative (i.e., uncertain and diverse) samples from the target domain [6,7,8]. The abovementioned approaches mainly focus on effective and efficient sample selection from the same and/or different sources of images to reduce the labeling cost.

Objectives

Methods

Findings

Conclusion