MT

Md. Touhid Islam

Hajee Mohammad Danesh Science and Technology University

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Trends

Author Order

Document Type

Co-Authors

Md. Rashedul Islam

Chittagong University of Engineering & Technology

Md Palash Uddin

Deakin University

Md. Sohrawordi

Hajee Mohammad Danesh Science and Technology University

Mohadeb Kumar

Hajee Mohammad Danesh Science and Technology University

Anwaar Ulhaq

Charles Sturt University

Faiaz Ahsan

Rajshahi University of Engineering and Technology

Ehsanul Islam Zafir

Rajshahi University of Engineering and Technology

Rahul Joysoyal

Rajshahi University of Engineering and Technology

Uzair Aslam Bhatti

Hainan University

Shekh S. Uddin

Rajshahi University of Engineering and Technology

Cooperative Institutions

Hajee Mohammad Danesh Science and Technology University

Rajshahi University of Engineering and Technology

Chattogram Veterinary and Animal Sciences University

Chittagong University of Engineering & Technology

International Islamic University Chittagong

Charles Sturt University

Monash University

University of Rajshahi

University of Technology Sydney

Bangladesh Livestock Research Institute

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Field

Segmentation-based Dimensionality Reduction for Hyperspectral Image Classification

2019 4th International Conference on Electrical Information and Communication Technology (EICT) (2023)

Tazbiul Hasan Md. Liakot Ali Liton Md Habibur Rahman Md. Touhid Islam Md. Rashedul Islam

Hyperspectral imaging (HSI) is a potent technique for capturing detailed spectral data across an extensive range of electromagnetic wavelengths with high precision. Furthermore, it is utilized in various domains like agriculture, surveillance of the environment, mineral investigation, medical studies, and defense due to its wide range of applications. It provides valuable insights by capturing and analyzing spectral information beyond the capabilities of traditional imaging techniques. However, storing, analyzing, and interpreting HSI data because of its great dimensionality presents difficulties. Therefore, specialized techniques are required to extract meaningful information from the vast amount of spectral data. In this paper, a segmentation-based approach for feature reduction in HSI data using Factor Analysis (FA) and feature selection techniques has been proposed. The Hyperspectral data was segmented based on spectral characteristics. Furthermore, FA was applied to extract the most significant spectral bands. Additionally, the extracted features were integrated, and feature selection methods based on Correlation, Mutual Information, and Maximum Relevance — Minimum Redundancy(mRMR) were applied. The proposed method aims to reduce dimensionality while preserving relevant information for accurate analysis and interpretation. The experiments on an Indian Pines dataset demonstrate that the spectral segmentation approach with mRMR-based feature selection achieves better classification results. The proposed approach surpasses traditional methods by requiring reduced computational time and memory. It enhances HSI data analysis through increased efficiency and accuracy, as evidenced by the improved classification results visualized in confusion matrices. This method marks a significant advancement in the field by optimizing both computational time and accuracy reducing memory requirements.

Feature (linguistics)

10.1109/eict61409.2023.10427661

Cite

Citations (0)

Author response for "Blockchain for sustainable city transformation: A review on Bangladesh"

Rahul Joysoyal Shekh S. Uddin Md. Touhid Islam Subrata K. Sarker Li Li

10.1002/eng2.12948/v2/response1

Cite

Citations (0)

A Deep Learning-Based Hyperspectral Object Classification Approach via Imbalanced Training Samples Handling

Remote Sensing (2023)

Md. Touhid Islam Md. Rashedul Islam Md Palash Uddin Anwaar Ulhaq

Object classification in hyperspectral images involves accurately categorizing objects based on their spectral characteristics. However, the high dimensionality of hyperspectral data and class imbalance pose significant challenges to object classification performance. To address these challenges, we propose a framework that incorporates dimensionality reduction and re-sampling as preprocessing steps for a deep learning model. Our framework employs a novel subgroup-based dimensionality reduction technique to extract and select the most informative features with minimal redundancy. Additionally, the data are resampled to achieve class balance across all categories. The reduced and balanced data are then processed through a hybrid CNN model, which combines a 3D learning block and a 2D learning block to extract spectral–spatial features and achieve satisfactory classification accuracy. By adopting this hybrid approach, we simplify the model while improving performance in the presence of noise and limited sample size. We evaluated our proposed model on the Salinas scene, Pavia University, and Kennedy Space Center benchmark hyperspectral datasets, comparing it to state-of-the-art methods. Our object classification technique achieves highly promising results, with overall accuracies of 99.98%, 99.94%, and 99.46% on the three datasets, respectively. This proposed approach offers a compelling solution to overcome the challenges of high dimensionality and class imbalance in hyperspectral object classification.

Benchmark (surveying)

10.3390/rs15143532

Cite

Citations (5)

MC-NET: Spectral-Spatial Feature Reduction for Hyperspectral Image Classification with Optimized Technique Series

Md. Touhid Islam Mohadeb Kumar Md. Rashedul Islam

The use of hyperspectral images (HSI) in a variety of practical applications has become widespread. High correlations between and among classes, the curse of dimensionality, and overfitting have drawn attention. Deep learning-based algorithms with dimensionality reduction methods have been proposed as a feasible alternative for hyperspectral image analysis owing to their effective feature extraction and high overall performance. Principal component analysis (PCA) and Sparse PCA (SPCA) are frequently employed in the fields of dimensionality reduction. Since principal component analysis (PCA) is hard to interpret because each principal component is a linear mixture of the original variables and SPCA cannot select the most important features. Dimensionality reduction via SPCA-MIFS (mutual information feature selection) is utilized to address the problem. The 2D and 3D Convolutional Neural Network (CNN) methods provide the backbone of many of the deep learning combination techniques. A fused 3D-2D CNN makes the model less complicated than a 3D-CNN alone, and it can also work well in noisy environments and when there aren't many training samples. Overfitting is addressed and classification accuracy is enhanced by using several optimization techniques. To evaluate the efficacy of this proposed strategy performance results compared highly reputed models with the less numbers of characteristics.

Overfitting

Feature (linguistics)

10.1109/icecte57896.2022.10114513

Cite

Citations (3)

Blockchain for sustainable city transformation: A review on Bangladesh

Engineering Reports (2024)

Rahul Joysoyal Shekh S. Uddin Md. Touhid Islam Subrata K. Sarker Li Li

Abstract Blockchain (BCn) revolution across the world leads the global transformation by altering the existing structure and introducing the accessible and decentralized paradigm. This technology facilitates access to resources and redefines conventional frameworks, empowering people and communities globally. Motivated by these attention‐seeking potentiality of BCn features, in this article, we examine the challenges, advancements, and emerging prospects of the BCn revolution in Bangladesh. With a specific focus on BCn implications for sustainability and smart development, this study addresses the urgent need to understand how BCn can facilitate sustainable development in rapidly evolving countries like Bangladesh. The goal of this research is to draw attention to the successful revolution of BCn in the world and use this knowledge to extract best practices and cutting‐edge strategies for sustainable transformation of Bangladesh. The contribution of this study covers the underlying concepts of the BCn technology, showing the revolutionary state of BCn across the world, applications of BCn for sustainability, assessment of its current usage, and highlighting the BCn implementation problems. Furthermore, this study serves as a gateway to a transparent, decentralized future that fosters inclusive national progress and aligns with sustainable development goals. This study also strives to inspire individuals to use this transformational technology for the collective benefit of society by highlighting BCn's potential to empower communities while driving positive change.

10.1002/eng2.12948

Cite

Citations (2)

Spectral–Spatial Feature Reduction for Hyperspectral Image Classification

Md. Touhid Islam Mohadeb Kumar Md. Rashedul Islam

Hyperspectral Image (HSI) has many narrow and continuous spectral bands. There are many problems with the original HSI. For example, the classification accuracy of the test data is affected by the curse of dimensionality problem because the image bands are highly correlated in both space and time. So, dimensionality reduction is done and improved the classification result. In this paper, a deep convolutional network is suggested to reduce the dimensionality of HSI classification by considering both spectral and spatial features. When combined factor analysis and mRMR are used, spectral features are reduced, while 2D wavelet CNN reduces spatial features. A wavelet CNN is an extension of a 2D CNN that can be used to classify high-resolution images. Wavelet CNNs also use layered wavelet transformations to pull out spectral features. A wavelet CNN is easier to calculate than a 3D CNN or a 2D-3D CNN. In the next step, the spectral features are connected to the two-dimensional CNN to get the spatial features, creating a spatial-spectral feature vector. It makes a model that can accurately classify HSI data at multiple resolutions. As part of the HSI classification, we used data sets from the Pavia University and Salinas Scene dataset to see how well the two methods work together. In two datasets, the proposed Expanded 2DNET did better than the handcrafted methods, according to the experiment.

Feature (linguistics)

Feature vector

Contextual image classification

10.1007/978-3-031-34622-4_45

Cite

Citations (2)

Improving Hyperspectral Image Classification with Compact Multi-Branch Deep Learning

Remote Sensing (2024)

Md. Rashedul Islam Md. Touhid Islam Md Palash Uddin Anwaar Ulhaq

The progress in hyperspectral image (HSI) classification owes much to the integration of various deep learning techniques. However, the inherent 3D cube structure of HSIs presents a unique challenge, necessitating an innovative approach for the efficient utilization of spectral data in classification tasks. This research focuses on HSI classification through the adoption of a recently validated deep-learning methodology. Challenges in HSI classification encompass issues related to dimensionality, data redundancy, and computational expenses, with CNN-based methods prevailing due to architectural limitations. In response to these challenges, we introduce a groundbreaking model known as “Crossover Dimensionality Reduction and Multi-branch Deep Learning” (CMD) for hyperspectral image classification. The CMD model employs a multi-branch deep learning architecture incorporating Factor Analysis and MNF for crossover feature extraction, with the selection of optimal features from each technique. Experimental findings underscore the CMD model’s superiority over existing methods, emphasizing its potential to enhance HSI classification outcomes. Notably, the CMD model exhibits exceptional performance on benchmark datasets such as Salinas Scene (SC), Pavia University (PU), Kennedy Space Center (KSC), and Indian Pines (IP), achieving impressive overall accuracy rates of 99.35% and 99.18% using only 5% of the training data.

Benchmark (surveying)

10.3390/rs16122069

Cite

Citations (3)

Enhanced Brain Tumor Classification from MRI Images Using Deep Learning Model

Asadullah Bin Rahman Md. Touhid Islam Md. Rashedul Islam Md. Sohrawordi Md. Nahid Sultan

In the realm of image classification, traditional algorithms, encompassing both machine learning and deep learning, grapple with formidable challenges arising from uneven pixel ranges and dimensionality reduction. This results in a significant impediment to achieving accurate image categorization. Numerous examples of such traditional methods, including KNN, Random Forest, SVM, DNN, CNN etc, have encountered persistent issues such as inefficient performance of feature engineering, limited accuracy, etc. In response to these challenges, this paper introduces a novel image classification method that integrates pixel mapping, DWT, and CNN for improved efficiency and reliability. By resolving irregular pixel ranges through initial pixel mapping, our method establishes uniformity as a foundation for subsequent image analysis. Subsequently, DWT is employed to dissect and reduce image dimensionality, extracting essential features while lowering computational complexity. This two-step preprocessing approach forms a robust foundation for effective data classification. Within this framework, our proposed CNN architecture plays a pivotal role, utilizing both spectral and spatial information to address image categorization challenges. The network's capacity to learn complex patterns enhances classification accuracy. In extensive evaluations, our methodology surpasses conventional classification techniques, yielding impressive results. With an Overall Accuracy (OA) of 96.9% and a Kappa statistic of 95.16%, our method showcases excellence and practical potential. These compelling achievements underscore the significance of our approach in tackling image classification challenges, paving the way for enhanced precision and efficiency across various domains.

Brain tumor

Contextual image classification

10.1109/iccit60459.2023.10441064

Cite

Citations (0)

Advancing hyperspectral image classification with dual branch networks and informative feature extraction

Journal of Spatial Science (2024)

Md. Touhid Islam Md. Rashedul Islam Abdulla Al Mamun

Hyperspectral imaging is increasingly important in academia and various professions, facing challenges like redundant features, inter-class correlations, and the curse of dimensionality. Principal Component Analysis and its variants, such as Sparse-PCA and Segmented-PCA, reduce the dimensionality of hyperspectral data but interpreting PCA results is complex. Our SPCA-mRMR technique integrates Sparse-PCA with a greedy feature selection method, mRMR. This, combined with a Dual Branch CNN model, improves hyperspectral image analysis, especially with noisy or limited data. Optimization reduces computational costs and enhances classification accuracy. Evaluations show our method's efficiency, using fewer variables without compromising accuracy, crucial for advancing HSI applications.

Feature (linguistics)

Contextual image classification

10.1080/14498596.2024.2364231

Cite

Citations (0)

Spectrally Segmented-Enhanced Neural Network for Precise Land Cover Object Classification in Hyperspectral Imagery

Remote Sensing (2024)

Md. Touhid Islam Md. Rashedul Islam Md Palash Uddin Anwaar Ulhaq

The paradigm shift brought by deep learning in land cover object classification in hyperspectral images (HSIs) is undeniable, particularly in addressing the intricate 3D cube structure inherent in HSI data. Leveraging convolutional neural networks (CNNs), despite their architectural constraints, offers a promising solution for precise spectral data classification. However, challenges persist in object classification in hyperspectral imagery or hyperspectral image classification, including the curse of dimensionality, data redundancy, overfitting, and computational costs. To tackle these hurdles, we introduce the spectrally segmented-enhanced neural network (SENN), a novel model integrating segmentation-based, multi-layer CNNs, SVM classification, and spectrally segmented dimensionality reduction. SENN adeptly integrates spectral–spatial data and is particularly crucial for agricultural land classification. By strategically fusing CNNs and support vector machines (SVMs), SENN enhances class differentiation while mitigating overfitting through dropout and early stopping techniques. Our contributions extend to effective dimensionality reduction, precise CNN-based classification, and enhanced performance via CNN-SVM fusion. SENN harnesses spectral information to surmount challenges in “hyperspectral image classification in hyperspectral imagery”, marking a significant advancement in accuracy and efficiency within this domain.

Land Cover

Object based

10.3390/rs16050807

Cite

Citations (2)