[Paper Review] Supervised Classification Performance of Multispectral Images

K. Perumal, R. Bhaskaran|arXiv (Cornell University)|Feb 22, 2010

Remote-Sensing Image Classification9 references153 citations

TL;DR

This paper evaluates supervised classification performance of multispectral images using various algorithms, focusing on the impact of increasing spatiotemporal data dimensions. It finds that the Mahalanobis classifier outperforms other methods in classification accuracy, demonstrating its effectiveness for remote sensing data analysis under growing data complexity.

ABSTRACT

Nowadays government and private agencies use remote sensing imagery for a wide range of applications from military applications to farm development. The images may be a panchromatic, multispectral, hyperspectral or even ultraspectral of terra bytes. Remote sensing image classification is one amongst the most significant application worlds for remote sensing. A few number of image classification algorithms have proved good precision in classifying remote sensing data. But, of late, due to the increasing spatiotemporal dimensions of the remote sensing data, traditional classification algorithms have exposed weaknesses necessitating further research in the field of remote sensing image classification. So an efficient classifier is needed to classify the remote sensing images to extract information. We are experimenting with both supervised and unsupervised classification. Here we compare the different classification methods and their performances. It is found that Mahalanobis classifier performed the best in our classification.

Motivation & Objective

• To address the limitations of traditional classification algorithms in handling the increasing spatiotemporal dimensions of modern remote sensing data.
• To evaluate and compare the performance of multiple supervised classification algorithms on multispectral imagery.
• To identify the most effective classifier for accurate remote sensing image classification in data-intensive scenarios.
• To support the development of efficient, high-precision classification systems for applications in agriculture, defense, and environmental monitoring.

Proposed method

• The study employs a range of supervised classification algorithms on multispectral remote sensing images.
• It compares performance using standard evaluation metrics, with a focus on classification accuracy.
• The Mahalanobis classifier is applied based on its ability to account for covariance structure in multivariate data.
• The analysis includes both supervised and unsupervised classification for comparative evaluation.
• Data preprocessing and feature representation are assumed to be standard for multispectral image classification.
• Performance is assessed using empirical results from real multispectral image datasets.

Experimental results

Research questions

• RQ1 Which supervised classification algorithm performs best on multispectral images under increasing data complexity?
• RQ2 How do traditional classification methods fare against modern demands of high-dimensional remote sensing data?
• RQ3 What role does covariance structure play in improving classification accuracy for multispectral data?
• RQ4 Can the Mahalanobis classifier effectively handle the spatiotemporal dimensions of modern remote sensing imagery?

Key findings

• The Mahalanobis classifier achieved the highest classification accuracy among all tested methods.
• Traditional classification algorithms show reduced effectiveness as the spatiotemporal dimensions of remote sensing data increase.
• Supervised classification methods outperform unsupervised methods in terms of precision and reliability for this dataset.
• The performance gap between classifiers becomes more pronounced with higher-dimensional data.