Automatic feature extraction and matching modelling for highly noise near-equatorial satellite images

Hayder Dibs,Hashim Ali Hasab,Hussein Sabah Jaber,Nadhir Al-Ansari

doi:10.1007/s41062-021-00598-7

Abstract

Feature extraction plays an important role in pattern recognition because band-to-band registration and geometric correction from different satellite images have linear image distortion. However, new near-equatorial orbital satellite system (NEqO) images is different because they have nonlinear distortion. Conventional techniques cannot overcome this type of distortion and lead to the extraction of false features and incorrect image matching. This research presents a new method by improving the performance of the Scale-Invariant Feature Transformation (SIFT) with a significantly higher rate of true extracted features and their correct matching. The data in this study were obtained from the RazakSAT satellite covering a part of Penang state, Malaysia. The method consists of many stages: image band selection, image band compression, image sharpening, automatic feature extraction, and applying the sum of absolute difference algorithm with an experimental and empirical threshold. We evaluate a refined features scenario by comparing the result of the original extracted SIFT features with corresponding features of the proposed method. The result indicates accurate and precise performance of the proposed method from removing false SIFT extracted features of satellite images and remain only true SIFT extracted features, that leads to reduce the extracted feature from using three frame size: (1) from 2000 to 750, 552 and 92 for the green and red bands image, (2) from 678 extracted control points to be 193, 228 and 73 between the green and blue bands, and (3) from 1995 extracted CPs to be 656, 733, and 556 between the green and near-infrared bands, respectively.

Highlights

Feature extraction from a remotely sensed dataset is an important step for different types of remote sensing applications such as band-to-band registration and geometric correction, image normalization
In this study, the sum of absolute difference (SAD) algorithm was used to refine the extracted Scale-Invariant Feature Transformation (SIFT) control points (CPs). In some respects this is different from previous studies because the weakness of both SIFT and SAD in this kind of image is overcome by removing the incorrectly matched CPs when the SIFT algorithm produces CPs
The automatic extracting CPs by SIFT is not adequate for the new optical satellite generation of near-equatorial orbital satellite system (NEqO) and multi-sensor images captured from different viewpoints and at different times and illumination

Summary

Introduction

Feature extraction from a remotely sensed dataset is an important step for different types of remote sensing applications such as band-to-band registration and geometric correction, image normalization. It has received considerable attention [1]. A considerable number of studies have proposed several local feature extraction techniques, such as the SIFT algorithm [1, 2]. These methods involve a Speeded Up Robust Features (SURF) by Bay et al [2] and Gradient Location Orientation Histogram referenced by Mikolajczyk and Schmid [3]. Lowe [4] described imagery features that have properties make them

Methods

Results

Conclusion