Hyperspectral Remote Sensing Image Classification Using Active Learning

Shrivastava, Vimal K.; Pradhan, Monoj K.

doi:10.1007/978-3-030-50641-4_8

Cited by 3 publications

(8 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The disagreement between all the predictions for a given instance is a common measure for uncertainty, although computationally inefficient [11,14]. It is calculated using the set of classifications over a single instance, given by the number of votes assigned to the most frequent class [35]. This method was implemented successfully for complex applications such as deep active learning [11].…”

Section: Ensemble-based Selection Criteriamentioning

confidence: 99%

“…Therefore, it can be seen as a bootstrap aggregation (bagging) ensemble disagreement method. It is represented by the maximum disagreement score out of a set of disagreements calculated for each view [35]. A lower value for this metric means a higher classification uncertainty.…”

Section: Ensemble-based Selection Criteriamentioning

confidence: 99%

“…EQB has been observed to be an efficient selection criterion. Specifically, Shrivastava and Pradhan [35] applied EQB on hyperspectral remote sensing imagery using Support Vector Machines (SVM) and Extreme Learning Machines (ELM) as choosers, achieving optimal results when combining EQB with ELM. Another study successfully implemented this method on an active deep learning application [12].…”

Section: Entropy-based Criteriamentioning

confidence: 99%

“…Margin Sampling is a SVM-specific criterion, based on the distance of a given point to the SVM's decision boundary [35]. This method is less popular than the remaining methods because it is limited to one type of chooser (SVMs).…”

Section: Other Relevant Criteriamentioning

confidence: 99%

“…Selecting an efficient selection criterion is particularly important to find the instances closest to the decision border (i.e., instances difficult to classify) [35]. Therefore, many AL related studies focus on the design of the query/acquisition function [13].…”

mentioning

confidence: 99%

See 4 more Smart Citations

Increasing the Effectiveness of Active Learning: Introducing Artificial Data Generation in Active Learning for Land Use/Land Cover Classification

2021

View full text Add to dashboard Cite

In remote sensing, Active Learning (AL) has become an important technique to collect informative ground truth data “on-demand” for supervised classification tasks. Despite its effectiveness, it is still significantly reliant on user interaction, which makes it both expensive and time consuming to implement. Most of the current literature focuses on the optimization of AL by modifying the selection criteria and the classifiers used. Although improvements in these areas will result in more effective data collection, the use of artificial data sources to reduce human–computer interaction remains unexplored. In this paper, we introduce a new component to the typical AL framework, the data generator, a source of artificial data to reduce the amount of user-labeled data required in AL. The implementation of the proposed AL framework is done using Geometric SMOTE as the data generator. We compare the new AL framework to the original one using similar acquisition functions and classifiers over three AL-specific performance metrics in seven benchmark datasets. We show that this modification of the AL framework significantly reduces cost and time requirements for a successful AL implementation in all of the datasets used in the experiment.

show abstract

Section: Ensemble-based Selection Criteriamentioning

confidence: 99%

Section: Ensemble-based Selection Criteriamentioning

confidence: 99%

Section: Entropy-based Criteriamentioning

confidence: 99%

Section: Other Relevant Criteriamentioning

confidence: 99%

mentioning

confidence: 99%

See 3 more Smart Citations

Increasing the Effectiveness of Active Learning: Introducing Artificial Data Generation in Active Learning for Land Use/Land Cover Classification

2021

View full text Add to dashboard Cite

show abstract

A hybrid approach to remote sensing image classification based on pixel density distribution and adaptive neural networks

Singh

2024

IJ-AI

View full text Add to dashboard Cite

<span lang="EN-US">The proposed study employed a method for identifying the main contents (category/class) that a Remote Sensing Image (RSI) belongs to, as well as the percentage contribution if the image comprises a significant number of different content types. Histogram based approach has been used to extract the pixel density distribution and its normalized form helps to make solution independent from image physical characteristics. A multilayer feedforward Artificial Neural Network (ANN) design has been used to address the classification problem. The architecture included an adaptive form of transfer function, whose slope characteristics changes along with weights as learning progresses. The approach of solution design is computation efficient because it doesn’t apply extensive pre-processing.</span>

show abstract

Change Detection in Remote Sensing Image Data Comparing Algebraic and Machine Learning Methods

et al. 2022

View full text Add to dashboard Cite

Remote sensing technology has penetrated all the natural resource segments as it provides precise information in an image mode. Remote sensing satellites are currently the fastest-growing source of geographic area information. With the continuous change in the earth’s surface and the wide application of remote sensing, change detection is very useful for monitoring environmental and human needs. So, it is necessary to develop automatic change detection techniques to improve the quality and reduce the time required by manual image analysis. This work focuses on the improvement of the classification accuracy of the machine learning techniques by reviewing the training samples and comparing the post-classification comparison with the image differencing in the algebraic technique. Landsat data are medium spatial resolution data; that is why pixel-wise computation has been applied. Two change detection techniques have been studied by applying a decision tree algorithm using a separability matrix and image differencing. The first change detection, e.g., the separability matrix, is a post-classification comparison in which individual images are classified by a decision tree algorithm. The second change detection is, e.g., the image differencing change detection technique in which changed and unchanged pixels are determined by applying the corner method to calculate the threshold on the changing image. The performance of the machine learning algorithm has been validated by 10-fold cross-validation. The experimental results show that the change detection using the post-classification method produced better results when compared to the image differencing of the algebraic change detection technique.

show abstract

Hyperspectral Remote Sensing Image Classification Using Active Learning

Cited by 3 publications

References 31 publications

Increasing the Effectiveness of Active Learning: Introducing Artificial Data Generation in Active Learning for Land Use/Land Cover Classification

Increasing the Effectiveness of Active Learning: Introducing Artificial Data Generation in Active Learning for Land Use/Land Cover Classification

A hybrid approach to remote sensing image classification based on pixel density distribution and adaptive neural networks

Change Detection in Remote Sensing Image Data Comparing Algebraic and Machine Learning Methods

Contact Info

Product

Resources

About