A comparative analysis of four object detection models YOLOv11, DETR, CenterNet, and Faster R-CNN was conducted for real-time assistive technology designed to support individuals who are blind or have low vision. Despite advances in computer vision, many assistive tools still struggle with real-time performance, particularly in dynamic, cluttered environments. Most existing studies focus on general use cases, often neglecting accessibility-specific needs. This study fills that gap by evaluating one-stage, transformer-based, anchor-free, and two-stage models using a real-world dataset. YOLOv11 achieved the best balance of speed and accuracy, with a mean Average Precision (mAP) of 0.52 and an inference time of 2.6ms, making it ideal for edge deployment. Faster R-CNN delivered the highest precision but suffered from slower inference, limiting its usability in real-time scenarios. These results underscore the importance of tailoring detection models for assistive use, balancing precision and speed to enhance accessibility solutions for the visually impaired.

Sleep disorders pose a significant challenge to worldwide health, underscoring the critical demand for accurate and prompt diagnostic methods. This study explores the use of machine learning (ML) methods to enhance and automate diagnostic procedures in healthcare systems for treating sleep disorders. A comprehensive dataset of physiological and behavioral sleep-related attributes was analyzed to evaluate and compare the performance of multiple ML algorithms, including Naive Bayes, Linear Discriminant Analysis (LDA), XGBoost Classifier, Gradient Boost Classifier etc. These models were evaluated with important metrics including accuracy, precision, recall, and F1score, and cross-validation was used to maintain reliability and strength. The analysis also considered computational efficiency and model complexity. Data preprocessing involved addressing missing values, feature scaling, and exploratory data analysis, with additional optimization through parameter tuning and feature selection. Notably, the KNN model was further optimized using the Fish Swarm Optimization technique, achieving an improved accuracy of 95.56%, surpassing its initial performance of 94.25%. This optimization underscores the novelty of the study, highlighting the potential of hybrid approaches in advancing MLdriven healthcare diagnostics.

This paper presents a comparative analysis of YOLOv11 and YOLOv12 for automated weed detection in precision agriculture. The primary objective is to assess both models' detection accuracy, generalization ability, and reliability using a custom-annotated dataset of sesame crop and weed images. YOLOv11, known for its faster inference speed, demonstrates higher mAP@0.5 in straightforward detection scenarios. However, YOLOv12 outperforms in challenging conditions due to its advanced architectural enhancements, including attention mechanisms and improved feature pyramids. This study highlights the trade-off between computational efficiency and robust detection, offering insights into choosing the optimal object detection model for real-time agricultural applications.

Facial recognition systems are widely used in various applications such as security, healthcare, and authentication, but face significant challenges in uncontrolled environments. Poor lighting conditions can obscure facial features, introduce shadows, and distort spatial relationships, while changes in pose are critical for accurate identification. Existing methods often struggle to strike a balance between accuracy, computational efficiency, and robustness. Deep learning has become popular for automatically learning features through convolution layers. This study proposes a robust framework that integrates contrast-limited adaptive histogram equalization (CLAHE) and adaptive gamma correction for illumination normalization and multi-task cascaded convolutional networks (MTCNN) for precise face detection under varying poses and lighting conditions. This study proposes a deep learning-based approach for face recognition utilising multiple models, including VGG16, VGG19, ResNet-50, ResNet-101, and MobileNetV2. For classification, an ensemble of SVM, XGBoost, and random forest classifiers is combined using weighted averaging. The approach is tested on datasets such as CASIA3D and 105PinsFace, which include variations in illumination conditions. Using deep learning for automated hierarchical feature extraction and ensemble strategies, experimental results demonstrate significant improvements in recognition accuracy and enhanced robustness against lighting and pose variations while ensuring scalability for real-world applications. The approach achieved 99.91% accuracy on the CASIA3D dataset and 98.77% on the 105PinsFace dataset, showcasing its effectiveness across challenging conditions.

Graph neural networks (GNNs) have emerged as powerful tools for analyzing graph-structured data with applications in social networks, bioinformatics, and recommender systems. However, existing GNNs struggle with (1) rigid edge weighting (e.g., GCN’s fixed normalization), (2) over-smoothing in deep layers, and (3) quadratic attention costs (e.g., GAT). MGCN introduces: (1) adaptive edge weighting to dynamically adjust neighbor influence, (2) residual connections to combat over-smoothing, and (3) a scalable attention mechanism. It also introduces a standardized evaluation framework that incorporates adaptive preprocessing techniques such as feature normalization, edge weighting, and graph augmentation. The proposed model demonstrated superior performance when compared to eight state-of-the-art GNN models such as GraphSAGE, GAT, Graph Transformer, GINConv, GCN, GraphCL, AGCN, and MGCN, across three widely used benchmark datasets: Cora, CiteSeer, and PubMed. All evaluation metrics–including Accuracy, Hit Ratio, Precision, Recall, and F1 Score–are reported as the mean ± standard deviation over 10 independent runs. The experimental results consistently demonstrate the superiority of the proposed MGCN model with approximately 2% improvement on above datasets.

This study aims to improve the segmentation of the left ventricle in cardiac magnetic resonance images, which is a crucial task for monitoring and diagnosing heart disease. We suggest an improved method based on a U-Net deep learning model that includes Grad-CAM for interpretability, a generalized dice loss to address class imbalance, and augmentation strategies specifically designed for cardiac MRI. Our approach achieves strong results on the Sunnybrook Cardiac Dataset using a pretrained model to speed up convergence and enhance segmentation performance. The results demonstrate improved model transparency and segmentation accuracy, providing a reliable and comprehensible clinical solution. This work closes a significant research gap and attempts to support clinical decision-making by focusing on the explainability and increase in cardiac magnetic resonance segmentation data.

The increasing incidence of brain tumors has underscored the critical need for accurate diagnosis and effective treatment strategies. This study explores advanced methodologies to enhance brain tumor detection and classification. We introduce an innovative convolutional model designed to significantly improve identification accuracy. The performance of several deep learning algorithms, including Inception V3, GoogLeNet, and VGG-19, is meticulously evaluated in the context of brain tumor image classification.A key novelty of this study is the implementation of decision-level fusion, a method not previously explored in this domain. By combining the classification outputs of multiple models, our approach enhances the overall decision-making process, leading to improved accuracy and robustness. This technique allows for the aggregation of diverse perspectives from different models, thereby mitigating individual model weaknesses and capitalizing on their strengths. Our results indicate that these approaches markedly enhance the accuracy (with Inception V3 reaching 98.25%, GoogLeNet 95.36%, and VGG-19 91.24%) and resilience of brain tumor detection and classification systems, laying the foundation for a reliable diagnostic tool.

The demand for food production has led to advancements in precision agriculture, aiming to enhance crop yield and quality. This study investigates the application of deep learning algorithms, including GoogLeNet, RESNET-50, MobileNet-v2, VGG-16, and ShuffleNet, for automated plant disease detection. The research utilizes a dataset comprising images of citrus diseases to train and evaluate the models. Results show promising accuracy rates, highlighting the potential of deep learning in optimizing resource utilization and facilitating timely interventions in agriculture.

Biometric template protection is important in the current situation because of the reliance on intelligent approaches for recognizing an individual. In recent years, there have been numerous high-profile frauds. Texture, color, and shape are today's most prevalent biometric features. Insufficient user data and compromised keys cast doubt on the dependability of biometric systems. The proposed system employs a novel technique to close this gap. It extracts biometric features from a person's face, iris, and palmprint. Combining biometric features increases system reliability, safety, and user privacy. We used a colorization technique to generate three separate colors Quick Response (QR) codes from a user-defined Red, Green, Blue (RGB) tuple random seed, which we then combined to create a one-way cancellable template. This work provides a biometric verification system that supports numerous cancellation mechanisms. This would reduce the dangers of biometric templates, confusing systems, user concerns, and fraud. The database saves these templates for future user-driven security key-based situations involving intra-class and inter-class verification in the changed domain. The system's remarkable performance resulted in improved accuracy and security, reaching 99.84% overall with a 0.11% crossover error rate at an optimal threshold of 5.59. Finally, the Area Under the Receiver Operating Curve (AU_ROC) was 0.97, which is closer to the optimal value of 1.

Lung cancer represents a significant contributor to global cancer-related deaths, underscoring the critical need for early detection to reduce mortality rates. Using convolutional neural networks (CNNs) and deep learning, with a specific focus on MobileNet, VGG16, GoogleNet, InceptionentV3 and ResNet50, this study delves into the integration of AI for lung cancer detection using the LC25000 dataset, encompassing a diverse range of lung pathology CT scans. By tailoring the MobileNet architecture and optimising it for CT image analysis, the research strives to enhance the model's precision in identifying lung malignancies. The customized MobileNet, InceptionNet, GoogleNet, ResNet50, VGG16 model undergoes fine-tuning via strategic adjustments and training to discern subtle patterns indicative of lung cancer. Ensembled with these models to give accurate results. within medical imaging datasets. Robust evaluation techniques are implemented to gauge the model's efficacy, incorporating metrics such as accuracy and computational efficiency, positioning it as a promising tool for advancing early lung cancer detection methodologies.

Facial image analysis and categorization have recently made great strides in computer vision. The current study, explores ways to help computers better recognize faces quickly and accurately, especially for tasks like security and entertainment. Identifying faces, emotions, and identities is crucial in Security and Surveillance, Access Control, user Authentication in Smart Devices, and Emotion Analysis in Human-Computer Interaction. Adopting the MobileNet deep learning model because it requires less memory works efficiently. To make it even more effective at recognizing faces, adjusted its parameters and tested it with two data sets, the CASIA 3D face data set and 105 pins data set. The study using MobileNetV2 achieved a very high accuracy of 98.71% on the CASIA 3D face data set and 99.29% on the 105 pins data set. The experimental results show that MobileNetV2 better understands faces in different situations.

The recommendation system is one of the most essential information services in today’s online business applications, such as Amazon, Flipkart, and YouTube. In recent days, deep learning and machine learning models have performed exceedingly well in various applications related to text processing, image processing, audio and video processing. This work aims to review recent studies that evaluated behavior of various deep learning and machine learning models in recommender systems, and summarize the various key insights related to their performance in these applications. Specifically, we focus on analysis of four types of deep learning and machine learning techniques: graph-based baselines, sequential baselines, selfsupervised sequential models, and self-supervised graph-based models. Moreover, these models are evaluated on four different types of datasets: Yelp 2018, Ml-1M, Amazon Beauty, and iFashion. Among the eleven different models employed for this analysis, the two self-supervised sequential models, CL4SRec and BERT4Rec, outperform in terms of two of the four distinct metrics (Recall and NDCG) used.

Social webs like Instagram, Twitter, and WhatsApp are full of deliberations involving sentiments, feelings, and impressions of human beings worldwide. Moreover, understanding and segregating texts based on emotions is a complex task that could be considered progressive sentiment analysis. As sentiments play a crucial role in human interaction, the skills to perceive it through textual content analysis has numerous applications in natural language processing (NLP) and human–computer interaction (HCI). This paper suggests classifying and examining tweets based on six basic emotions: happiness, fear, anger, disgust, surprise, and sadness. Language translators are used to apply data augmentation. Experimental results show that augmented data provides better results than the original data.

There are many types of fraud in our daily life. One of the frauds occurring these days is credit card fraud. When people around the globe make credit card transactions, there will also be fraudulent transactions. To avoid credit card fraud, we must know the patterns and how the fraud values differ. This paper proposed credit card fraud detection using machine learning based on the labeled data and differentiating the fraudulent and legitimate transactions. The experiment was conducted using supervised machine-learning techniques.

Communication technology is used to transmit the error correction signal to the GPS receiver from geostationary satellites to improve the accuracy of position information provided by the receiver. This study uses GPS-Aided GEO Augmented Navigation (GAGAN) Satellite-Based Augmentation System (SBAS) GPS receiver in stand-alone mode to provide a simple and cost-effective solution with improved positional accuracy to collect the ground control points. In this study, the GAGAN receiver is operated in static mode for longer durations at International GNSS Service (IGS)—HYDE reference station, whose position is known very accurately. It is found that the GAGAN receiver can achieve accuracies of 0.3–0.6 m in position and up to 1.0 m in elevation depending on the duration of the observation meeting the requirements of many survey and engineering applications. Using the GCPs obtained by this process, orthophoto having a planimetric accuracy of 1.0 m can be generated from CARTOSAT-2E High-Resolution (HR) satellite images. The typical height accuracies of the order of 2.0 m (3–4 pixels) could be achieved from the Digital Surface Model (DSM) derived from this process.

Face recognition technology is one of the everyday tasks in our daily life. But, recognising the correct face with high accuracy from large databases is a challenging task. To overcome this challenge, feature fusion of local binary pattern (LBP) with Gabor-Kernel Fisher analysis (Gabor-KFA) has proposed for face recognition. In this method, by using Gabor filter, extract Gabor features from a face image, on the other hand, extract features from LBP coded face image, then combined these extracted features generate high dimensional feature space. With this high dimensionality features, the complexity of training time and identification time may increase. To avoid this complexity, the Kernel Fisher analysis algorithm was adopted to reduce the feature vector size. Experiments were conducted separately on Gabor features and also on fused features. To test the performance of the proposed approach, the experiments were performed on the IIT Delhi database, ORL database, and FR database.

Recent years have seen increased interest in research on biometric template protection due to the widespread use of biometric authentication systems. For providing security to biometric templates, there is a procedure known as Fully Homomorphic Encryption that protects biometric templates from the malicious server environment. Users have more customization options with a biometric authentication system than a password or token. The most popular biometric modalities include fingerprints, iris scans, facial images, etc. Biometric modalities offer security on IoMT -based systems. Face recognition is one of the most often utilized biometric authentication methods in societal structure. Face recognition technology has made enormous strides in recent years. Here, we examine the viability of securing a database of iris templates using a methodology based on fully homomorphic Encryption. By directly matching templates in the encrypted domain, this framework is designed to protect confidentiality and restrict information from leaking from the templates while preserving their utility. We also investigate various classification techniques on machine learning models to achieve improved accuracy with shorter execution times. The aggregate verification vector assists in confirming the accuracy of the computed classification result, and the CKKS technique ensures confidentiality for the biometric templates. This study provides a plethora of information on fully homomorphic biometric authentication, containing a wide assortment of algorithms that satisfy homomorphic Encryption and various methods for extracting the biometric-related template.

With the enormous increase in the number of vehicles that are making use of roadways day by day, traffic congestion is one of the significant issues that are being observed. This problem can be addressed with proper traffic regulation. In this paper, proposing an automated system that will perform traffic analysis from the traffic videos which were captured from static cameras. This traffic analysis will be performed in three stages: vehicle detection, vehicle classification, into ten major categories, followed by vehicle counting under each category. The proposed work adopting the background subtraction method and vehicle classification by using a pre-trained model VGG16 as well as logistic regression (LR). Observations were made across several models ranging from the models that were built from scratch to pre-trained models as well as VGG16 + logistic regression. Experimental results show that the proposed model provides top-1 accuracy of 84.32% and top-5 accuracy of 99.77%.

This paper introduces the concept of two-dimensional (2D) color barcode, also known as color quick response (QR) pattern generation, and integration as an automatic method to produce the cancelable biometric template with improved recognition accuracy. It includes various methodologies for multi-modal based generation of biometric template, cipher conversion, diversity, irreversible property, etc. In this work, based on application of different attributes to four different biometric traits combining feature selection and fusion techniques, subsequently three templates are generated. However, in general, cancelable biometrics(CB) come with some systematic template distortion, which directly depends on input biometric characteristics to protect sensitive information. This will degrade the system performance when input deals with multiple biometric traits in the multi-biometric system. To address these issues, with the notion of color variant QR pattern analysis, dynamic constrained random key generation is introduced to generate CB templates. These templates can replace all other existing CB systems without compromising the quality metrics due to an independent transformation model for an authentication factor.

A public health emergency threat is happening due to novel coronavirus 2019 (nCoV-2019) throughout the world. nCoV-2019 is also named Severe Acute Respiratory SyndromeCoronaVirus-2 (SARS-CoV-2). COVID-19 is the disease caused by this virus. The virus originates in bats and is transmitted to humans by some unidentified intermediate animals. This virus started around December 2019 at Wuhan of China. After that, it turned into a pandemic. Even though there is no efficient vaccination, the entire world fights against the COVID-19. This article presents an overview of the scenario of the world as well as India. Some of the leading countries in the world are also affected by this virus badly. Even India is the 2nd highest population, is taking necessary precautions to protect it. With the Government of India's decisions, along with effective social distancing and hygienic measures, India is in a better position. But, in the future, COVID19 cases in India, still unpredictable. We designed an algorithm based on Convolutional Neural Network (CNN), which helps to classify COVID19+ and COVID19- persons using people's chest X-ray images automatically generated within the shortest time. The proposed method discovered that employing CT scan medical images produced more accurate results than X-ray images.

Urban building information can be effectively extracted by applying object-based image segmentation and multi-stage thresholding on High Resolution (HR) remote sensing satellite imageries. This study provides the results obtained using this method on the images of Indian remote sensing satellite, CARTOSAT-2S launched by the Indian Space Research Organization (ISRO). In this study, a method is developed to extract urban building footprints from the HR remote sensing satellite images. The first step of the process consists of generating highly dense per pixel Digital Surface Model (DSM) by using semi global matching algorithm on HR satellite stereo images and applying robust ground filtering to generate Digital Terrain Model (DTM). In the second step, multi-stage object-based approach is adopted to extract building bases using the PAN sharpened image, normalized Digital Surface Model (nDSM) derived from DSM and DTM, and Normalised Difference Vegetation Index (NDVI). The results are compared with the manual method of drawing building footprints by cartographers. An average precision of 0.930, recall of 0.917, and f-score of 0.922 are obtained. The results are found to be in a match with the method using the high resolution Airborne LiDAR DSM by providing a solution for large areas, low cost and low time.

Deep Learning is one of the machine learning area, which is widely used in recent research fields. In this, the work exhibits about working of the Convolutional Neural Networks (CNNs) for image classification. Deep learning approaches are better than the traditional learning algorithms when the data size is large because every day, a vast volume of data is accumulated everywhere. In deep learning, Convolutional Neural Network is one of the leading architecture. Convolutional Neural Network contains pre-trained models to transfer knowledge for learning the features, and such models are LeNet, AlexNet, GoogleNet, VGG16, VGG19, Resnet50, etc. These architectures are trained with a large ImageNet dataset, which contains millions of images. Moreover, these trained networks are also used to do new tasks. Among these pre-trained models, GoogleNet has less number of parameters, and this causes to reduce the computation complexity. We propose a deep network with Dimensionality Reduction Module (DRM), which works on less training data, and produce more accurate classification with minimum processing time and also a minimum number of parameters with regularization. The performance of classification, as well as training time and classification time of the proposed architecture, is measured with popular datasets such as ORL, Adience face dataset, Caltech101, and CIFAR10. The proposed architecture achieves better performance with less time when compared with the state of the work.

In this paper, a two-phase palmprint recognition approach is proposed based on statistical features and wide principal line image features through dynamic region of interest (ROI). The ROI is segmented into overlapping segments by six schemes, and the statistical features are extracted directly from the segments. The algorithm focuses on the extraction of statistical features based on standard deviation and coefficient of variation. A modified dissimilarity distance is proposed for computing the distance between two palmprints. The procedures are presented for determining the size and location of the common region of training images dynamically. Experiments are conducted by using statistical features and the combination of statistical and wide principal line image features. The results show that the correct recognition rate (CRR) of the proposed approach is better than existing methods for PolyUPalmprint database.

Brain tumors can be detected correctly by MRI images. In this paper, an advanced technique using support vector machine and fuzzy C-means classification has been proposed. Before going to apply the proposed methodology, images should be at high quality to pertain good results. The quality of the image will be enhanced using RGB to gray conversion and followed by employing median filter and binarization techniques. Then fuzzy C-means clustering has been applied to isolate the tumor portion in MRI image of brain. Local Binary Pattern (LBP) has been used to extract features of the brain image, and then SVM classification has been applied to classify the brain MRI images to know whether that tumor is normal or abnormal. The proposed technique provides an accuracy of 94.8%.

The medical field has its significance with increasing the demand of automatic diagnosis. These automated systems reduce the effort of the experts to make decisions. Our proposed system supports experts making the right decisions while predicting the cancer tumors in the lungs based on the CT image scan. This system converts RGB images into gray images, removes the noise using the median filter, and segments the CT images to avoid the unwanted part from the scanned image because of the segmented images’ discriminative features. Those features are extracted by using the Local Binary Patterns. Finally, the classification was done by the SVM kernels, such as linear, polynomial, and radial basis function. The radial basis kernel function achieved 88.76% accuracy. The proposed approach is tested on the LIDC dataset.

This paper introduces the texture alone fingerprint recognition system and uses a QR pattern to generate the cancelable biometric template with an unproved probability of error. This proposed cancelable bio-cryptosystem inherits all the advantages of texture features from fingerprint biometric traits for a template generation, cipher transformation, and non-invertible properties, etc. Here. GLCM feanire attributes are extracted from texture classified biometric images followed by feanire selection and fusion techniques. And user key-driven random transformation is carried out for the transformed domain biometric template. And for cancelable biometric. some systematic QR patterns are generated, which directly depend on the transformed template. This will not degrade the system's performance irrespective of randomizations used for nou-iuvertible transforms.

Cloud has been rising, renown, and extremely demanding innovation now a day. Cloud has wide ubiquity with its advanced features, like web access, more stockpiling, easy setup, programmed refreshes, low cost, and resource provisioning on a rent basis. Disregarding many advantages, security is viewed as increasingly significant and drew the consideration of numerous researchers. The information storage is drastically increasing, and there are many occasions that cloud doesn't ensure that data/information that has been placed in the cloud is secured from unauthorized access. Many experts are attempting to guarantee data security in the cloud, yet tragically they don't give satisfactory results. Hence we attempted to propose an effective crypto-scheme with obfuscation and cryptography for unstructured information. The scheme attempts to safeguard the secrecy of information at two phases. In the first phase, it obfuscates the file by supplanting the keywords (obfuscation), and at the subsequent phase, the obfuscated file is encoded by using the conventional RSA (Rivest Shamir Adleman) encryption algorithm for high security. Investigation results show that the proposed mechanism yields great outcomes.

A computer-based method is presented in this paper to define brain tumor using MRI images. The main classification motive is to identify a brain into a healthy brain or classify a brain with a tumor when a patient's MRI images are given. Magnetic Resonance Imaging (MRI) is an important one among the common imaging treatments, which presents more detailed brain tumor identification information and provides detailed pictures of inside your body other than computed tomography (CT). Currently, CNNs is a famous technique to deal with most of the problems with image classification as they provide greater accuracy compared to other classifiers. Hbridized CNN has been used in this work. It consists of three convolution layers and three max pooling layers which could provide outrated performance. Images from open databases such as BRATS were tested on brain MRI images. The proposed model has given the improved performance over the existing model with an accuracy of 96.15%.

In recent days, Face sketch synthesis (FSS) attracts various researchers for sketching the images to retrieve faces and in multimedia applications. The intention of FSS is to create a sketch for the image provided from a collection of sketch and photo images as the training set. Presently, the rise of deep learning (DL) models becomes useful in FSS because of its diverse benefits. As the FSS is employed in various applications, detailed experimentation to analyze the state of the art approaches methods is nontrivial. Though numerous FSS approaches are available, there is no review paper exist regarding the hierarchical classification of DL based FSS. Keeping this in mind, in this paper, we provide an extensive review of the available DL as well as conventional FSS techniques. We made a clear classification of the FSS techniques, and these are categorized into data-driven and model-driven methods. A comparative analysis of the reviewed techniques is made based on various aspects such as the objective, algorithms used, benefits, and performance measures.

Heart disease is a very deadly disease. Worldwide, the majority of people are suffering from this problem. Many Machine Learning (ML) approaches are not sufficient to forecast the disease caused by the virus. Therefore, there is a need for one system that predicts disease efficiently. The Deep Learning approach predicts the disease caused by the blocked heart. This paper proposes a Convolutional Neural Network (CNN) to predict the disease at an early stage. This paper focuses on a comparison between the traditional approaches such as Logistic Regression, K-Nearest Neighbors (KNN), Naïve Bayes (NB), Support Vector Machine (SVM), Neural Networks (NN), and the proposed prediction model of CNN. The UCI machine learning repository dataset for experimentation and Cardiovascular Disease (CVD) predictions with 94% accuracy.

Building foot prints and building count information in urban areas are very much essential for planning and monitoring developmental activities, efficient natural resource utilization, and provision of civic facilities by governments. Remote sensing data such as satellite/aerial imagery in association with digital elevation model is widely used for automatic extraction of building information. Many researchers have developed different methods for maximizing the detection percentage with minimum errors. A comparative study of different methods available in the literature is presented in this paper by analyzing the primary data sets, derived data sets, and their usage in the automated and semiautomated extraction methods. It is found that the success of the method for automatic building detection in urban areas primarily depends on using combination of high-resolution image data with digital elevation model.

Cloud has been emerging, popular and very demanding technology now a day. Cloud has got wide popularity with its sophisticated features. The primary features of cloud include internet access, more storage, easy setup, automatic updates, and low cost and resource provisioning based on “pay as you go” policy. In spite of advantages, security is considered to be more important and drew the attention of many researchers because it is not guaranteed in an open cloud. The data storage is becoming an indispensable measurement in cloud and most of the times cloud does not guarantee that data that has been stored is secured from illegitimate access. Many researchers are working to ensure data security in the cloud but unfortunately they do not provide adequate security to data. This paper is aiming to propose a secure hybrid scheme with obfuscation and cryptography to ensure the privacy of data shared in public cloud. Experimental results show that the proposed scheme yields good results.

In this paper, we present a new method for synthesizing a face sketch from a photo using deep neural networks. The face sketch has been synthesized by the framework through replicating the artists form sketch in a cascading way. Before transforming the digital image to face sketch, gamma correction is applied to enhance the contrast of the image. Next, the content image is produced which make face shape outline and key facial features. To improve the sketch details, shadings and textures are inserted. To generate a content image, fully convolutional neural network (FCNN) is employed first and then a style transfer method is applied to set up shadings and textures depending on the new projected pyramid column feature with gamma correction (PCF-G) method. The style transfer strategy preserves additional sketch details that depend on the pyramid column feature when comparing with general style transfer strategy and conventional patch-based methods. Qualitative and quantitative examinations recommend that this structure is even better when compared with the standard techniques on the applied various sample images. The presented PCF-G method exhibits superior results with a maximum Structural similarity index (SSIM)value of 0.504 on the applied test images.

In recent years, the use of biometrics for person authentication and image encryption to achieve and maintain the security of the image is extensively used. A competitive call is made for the researchers in transmission of digital data with truth of security is prioritized in image applications, in particular, Health Care Informatics (HCI). A novel method is proposed to cater to these requirements, which realizes the properties of Random Grid Visual Secret Sharing through the Quick Response Code (RGVSSQRC). RGVSSQRC provides perfectness, idealness, storage, and contrast requirements for preventing authenticating information from stolen attacks. The objective of the present research paper is to disseminate the use of Random Grid Visual Secret Sharing (RGVSS) for multi-biometric template encryption in medical applications without the use of any key for generating secret cipher shares with optimal contrast and aspect ratio for better vision through Quick Response (QR) Code.

Cancer is one of the deadliest diseases leading to innumerable deaths worldwide. Early detection of lung cancer could increase the survival rate. To detect cancer various image processing techniques have been innovated and applied like median-wiener filter in the preprocessing stage. In the classification Back Propagation model, SVM (Support Vector Machines), Forward Neural Networks, Convolution Neural Networks are used to detect whether the nodule is cancerous or not. Although, there are many such techniques which are available these days but there is still need to further develop early detection to improve accuracy leading to better survival rate.

Several cancelable biometric cryptosystems have been proposed to give security and protection to the biometric data. Even though these- techniques provide security from pre-image attacks and template protection. Developing innovative and highly robust cancelable biometric cryptosystems are vital. This paper proposes a novel cancelable biometric cryptosystem for template protection using color QR code. The proposed biometric cryptosystem is key generation based and registration free feature based multimodal biometric template of cancelable biometric method and works with conventional matcher. The proposed system has realized the properties of cancelable biometrics – revocability, diversity, non-invertible biometric encryption and pre-image attack resistant. Keywords:cancelable biometrics; biometric cryptosystems; color QR code; revocability, pre-image attack; non-invertible.

Women Deep Learning is-one of the machine learning areas, applied in recent areas. Various techniques have been proposed depends on varieties of learning, including un-supervised, semi-supervised, and supervised-learning. Some of the experimental results proved that the deep learning systems are performed well compared to conventional machine learning systems in image processing, computer vision and pattern recognition. This paper provides a brief survey, beginning with Deep Neural Network (DNN) in Deep Learning area. The survey moves on-the Convolutional Neural Network (CNN) and its architectures, such as LeNet, AlexNet, GoogleNet, VGG16, VGG19, Resnet50 etc. We have included transfer learning by using the CNN’s pre-trained architectures. These architectures are tested with large ImageNet data sets. The deep learning techniques are analyzed with the help of most popular data sets, which are freely available in web. Based on this survey, conclude the performance of the system depends on the GPU system, more number of images per class, epochs, mini batch size.

Now-a-day’s Digital Image Processing assumes an indispensable job in our day by day works too. Quality of images plays a crucial role, for example in Medical field. Medical Fundus images are used for detecting eye related diseases. Primary objective of Denoising of an image is not only to remove noise but also to preserve the image details as many as possible. In this paper, the work focuses on various image denoising techniques and their efficiency is measured through various parameters like PSNR-Peak Signal Noise Ratio and MSE-Mean Square Error.

Lung cancer is the world's leading cause of cancer death. The convolutional neural network (CNN) has been proved able to classify between malignant and benign tissues on CT scan images. In this paper, a deep neural network is designed based on GoogleNet, a pre-trained CNN. To reduce the computing cost and avoid overfitting in network learning, the densely connected architecture of the proposed network was sparsified, with 60 % of all neurons deployed on dropout layers. The performance of the proposed network was verified through a simulation on a pre-processed CT scan image dataset: The Lung Image Database Consortium (LIDC) dataset, and compared with that of several pre-trained CNNs, namely, AlexNet, GoogleNet and ResNet50. The results show that our network achieved better classification accuracy than the contrastive networks.

The primary objective of this study is to provide a methodology to generate a dense point cloud of digital surface model (DSM) and digital terrain model (DTM) from 0.6 m GSD stereo images acquired by CARTOSAT-2E satellite of the Indian Space Research Organization. These products are required for many high-resolution applications such as mapping of watersheds and watercourses; river flood modeling; analysis of flood depth, landslide, forest structure, and individual trees; design of highway and canal alignment. The proposed method consists of several processes such as orienting the stereo images, DEM point cloud extraction using the semi-global matching, and DSM to DTM filtering. The stereo model is built by performing aero triangulation and block adjustment using the ground control points. The semi-global matching algorithm is used on the epipolar images to generate the DSM in the form of dense point cloud corresponding to one height point for each pixel. The planimetric and height accuracies are evaluated using orthoimages and DSM and found to be within 3-pixel (~ 2 m) range. A method for extracting DTM by ground point filtering, to discriminate the probable ground points and the non-ground points, is provided by using discrete cosine transformation interpolation. This robust method uses a weight function to differentiate the noise points from the ground points. The overall classification efficiency kappa (κ) value averages at 0.92 for ground point classification/DTM extraction. The results of benchmarking of the GPS-aided GEO augmented navigation GPS receiver by operating it over IGS station, in static mode for collecting the checkpoints, also are presented.

In this paper, palmprint verification and identification with minimum number of features is proposed. The wide principal line extractors (WPLEs) on the region of interest (ROI) are applied to generate wide principal line images (WPLIs). The WPLI is segmented into 2 × 2, 4 × 4, 8 × 8 and 16 × 16 and the feature value is extracted directly from each segment. Experiments are conducted by using the extracted features. The results show that the equal error rate (EER), decidability index (DI) and correct recognition rate (CRR) of the proposed approach is better than existing methods for PolyUPalmprint Database.

The concept Location Based Encryption is pretty much useful in increasing information security to another level when combined with mobile applications. Sometimes data breach may happen because of these identities are misused as a result security may shutdown. When it comes to personal use and organizational use, it is crucial to check all the boxes of data security in storing data. Hence, we require a better form of encryption techniques. In this paper we focus on the notion of Location Based Data Encryption Algorithm. The Android operating system is cool and great open source [10]. Using Linux kernel, android consists of lots APIs offering location services which provides various services to obtain phones location from any location provider like GPS and algorithm is designed to decrypt data in trusted location.

Biometric template protection techniques like biometric cryptosystems and cancelable biometrics are most widely used in many large-scale biometric systems. Though generic biometric cryptosystems differ from other conventional cryptosystems, still it is insufficient to overcome the challenges ahead of identity frauds and vulnerabilities to major attacks. In recent years it's been used as promising primitives in many Internet of Things (IoT) devices and third party Intellectual Property protections with the name called cancelable biometrics where both user-defined random transformations are combined with biometric template vectors. However, protection over biometric templates (e.g., retina, iris, and palmprint) needs to be improved. In this work, the analysis presents biometric cryptosystems and cancelable biometrics with major outlook to recent prospects like obfuscation and multi-object biometric system.

In this paper proposed palmprint identification using Gabor features, Gabor and Wide Principal Line Image (WPLI) features. Extracted a fixed size ROI from palmprint images. Resize the extracted ROI into 64 x 64. Apply the Gabor filters to extract the features from the resized ROI. Dissimilarity distance is used to measure the dissimilarity between the query palmprint and database palmprint images. Experiments were conducted on Polyu Palmprint Database using Gabor features, Gabor and WPLI features. Experimental results shows that the proposed approach using Gabor and WPLI features obtains better results compared with the existing methods.

In this paper, a novel palmprint identification and verification algorithm is proposed based on wide principal lines through dynamic ROI. Region of interest (ROI) extraction is an important task for palmprint identification. Earlier reported works used fixed size ROI for the recognition of palmprints. When the fixed size ROI is used the palm area taken up for recognition is less compared to dynamic ROI extraction. The proposed algorithm focuses on extraction of maximum possible ROI. A set of wide principal line extractors are devised. Later these wide principal line extractors are used to extract the wide principal lines from dynamic ROI. A two stage palmprint identification algorithm is proposed based on wide principal lines. The experimental results demonstrate that the proposed approach extracts better ROI on the PolyUPalmprint Database when compared to the existing fixed size and dynamic size ROI extraction techniques. The experimental results for the verification and identification on PolyUPalmprint Database show that the discrimination of wide principal lines is also strong.

This paper proposes an effective resolution enhancement approach for images such as satellite images as well as normal images. In this method DT-CWT and bicubic interpolation were used. The proposed method was tested on well-known benchmark images. Finally Peak Signal to Noise Ratio (PSNR) and visual results of the proposed method out performs the state of art image resolution enhancement techniques. © 2013 IEEE.

In this paper, a novel palmprint identification and verification algorithm is proposed based on wide principal lines. A set of wide principal line extractors are devised. Later these wide principal line extractors are used to extract the wide principal lines. Morphological operators and grouping functions are used to eliminate the noise. In matching stage, a matching algorithm, based on pixel-to-pixel comparison is devised to calculate the similarity between the palmprints. In identification stage, wavelets and principal component analysis (PCA) are used for dimensionality reduction. Then Locally Discriminating Projection (LDP) is used to get the indexed list and the user is identified based on matching algorithm. The experimental results for the verification and identification on PolyU Database and Sub2D database are provided by Hong Kong Polytechnic University show that the discrimination of wide principal lines is also strong. With a minimum number of verifications, user is identified on these databases. © 2012 ACM.

Region of Interest (ROI) extraction is an important task for palmprint identification. Earlier reported works used fixed size ROI for the recognition of palmprints. When the fixed size ROI is used the palm area taken up for recognition is less compared to dynamic ROI extraction. The proposed algorithm focuses on extraction of maximum possible ROI compared to existing fixed and dynamic ROI extraction techniques [7, 19]. The experimental results demonstrate that the proposed approach extracts better ROI on three databases, 1. The PolyU Palmprint Database, 2. CASIA Palmprint Image Database and 3. IIT Delhi Palmprint Database, when compared to the existing fixed size and dynamic size ROI extraction techniques. © 2012 Springer-Verlag.

LDP is a supervised feature extraction algorithm hence it considers both class and label information for classification. A new face recognition algorithm named Modular Locally Discriminating Projection (MLDP) is presented in this paper. In the proposed method, initially the training and test face images are subdivided into smaller sub face images and then LDP is applied to each sub face images. Within this sub face images some of the local features do not vary largely corresponding to pose, lighting and facial expression of individual face images. The proposed MLDP captures most of the similarity features against pose, lighting and facial expression. This improves the classification accuracy. The experimental results on the ORL face database suggest that the proposed modular LDP has better recognition rates than Modular PCA and other conventional feature extraction methods. © 2011 IEEE.

The face recognition task involves extraction of unique features from the human face. Manifold learning methods are proposed to project the original data into a lower dimensional feature space by preserving the local neighborhood structure. LPP should be seen as an alternative to Principal Component Analysis (PCA). When the high dimensional data lies on a low dimensional manifold embedded in the ambient space, the Locality Preserving Projections are obtained by doing the optimal linear approximations to the Eigen functions of the Laplace Beltrami operator on the manifold. However, LPP is an unsupervised feature extraction method because it considers only class information. LDP is the recently proposed feature extraction method different from PCA and LDA, which aims to preserve the global Euclidean structure, LDP is the extension of LPP, which seeks to preserve the intrinsic geometry structure by learning a locality preserving submanifold. LDP is a supervised feature extraction method because it considers both class and label information. LDP performs much better than the other feature extraction methods such as PCA and Laplacian faces. In this paper LDP along with Wavelet features is proposed to enhance the class structure of the data with local and directional information. In this paper, the face Image is decomposed into different subbands using the discrete wavelet transform bior3.7, and the subbands which contain the discriminatory information are used for the feature extraction with LDP. In general the size of the face database is too high and it needs more memory and needs more time for training so that to improve time and space complexities there is a need for dimensionality reduction. It is achieved by using both biorthogonal wavelet transform and LDP the features extracted take less space and take low time for training. Experimental results on the ORL face Database suggests that LDP with DWT provides better representation and achieves lower error rates than LDP with out wavelets and has lower time complexity. The subband faces performs much better than the original image in the presence of variations in lighting, and expression and pose. This is because the subbands which contain discriminatory information for face recognition are selected for face representation and others are discarded. © 2010 IEEE.