Search

Machine Learning (ML) is an advanced branch of Artificial Intelligence (AI) focused on creating algorithms and statistical models that empower computer systems to learn from data and autonomously make informed decisions or accurate predictions, all without requiring explicit programming for every individual task. It enables computers to recognize patterns, relationships, and insights within the data and improve their performance through experience. Machine learning has had a significant impact on medical imaging in recent years, revolutionizing the field and enhancing healthcare practices. Machine Learning provides improved diagnostic accuracy, faster image analysis, reduced errors and variabilities and detection of anomalies and lesions. In this paper we applied various ML algorithms like Random Forest (RF), Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Logistic Regression (LR), Naïve Bayes (NB) on HAM (Human Against Machine) 10000 skin cancer data set. Since the data set is huge, we used dimensionality reduction and feature extracting techniques like Principal Component Analysis (PCA) and Gabor filters. The data set has 10015 images of seven classes of skin cancer. Our findings reveal that Random Forest when used with PCA produced an accuracy of 89% and when it is used with Gabor feature extraction it produced an accuracy of 84%. The SVM classifier with PCA produced an accuracy of 82% and when used with the Gabor feature extraction SVM produced an accuracy of 84%. RF produced an increased accuracy of 92% when the data samples for each class is increased. [ABSTRACT FROM AUTHOR]

Currently, most detection and analysis are done based on the deep neural network to get better results. At the same time, issues such as lower accuracy, higher error rate, and higher system complexities arose in detection. So, a novel Ant lion-based Elman Neural Network (ALbENN) was proposed to overcome the abovementioned problems. Moreover, the proposed model detected the occlusion images and found the authentication user based on the input images. Ear data sets were used for this research to find the occlusion and predict the authenticated user. Subsequently, the input image contained both the normal and the occlusion images, so the system pre-processed the image data sets to find the occlusion. After the pre-processing phase, the features were extracted, and the removed featured images were used for further detection. The proposed model was implemented with the help of Python software, and the results were measured. A case study was established to explain the working presentation of the proposed model. Consequently, a comparative analysis was made to compare the outcomes of the proposed approach with the existing occlusion detection models. Moreover, the comparison analysis proved that the proposed model attains better results than the current models. [ABSTRACT FROM AUTHOR]

The classification of targets is one of the challenging tasks in the field of modern radar systems. Manual feature extraction with high level computer vision algorithms or neural networks needs knowledge of the subject domain. In this explore a pretrained Recurrent neural network (RNN) with a bidirectional LSTM model for radar target classification where a separate feature extraction is not required for this network. The canonical models for simple geometric shapes, namely sphere, cylinder, disc and frustum and complex geometric shapes, namely complex sphere, complex cylinder, complex frustum and complex disc are developed using dedicated phased backscatterd algorithms. In terms of performance metrics, Bi-LSTM network gives an accuracy of around 99.77% for simple targets and 99.63% for complex targets, far better than the machine learning models. The experimental study on target classification of radar sequence data using Bidirectional Long short-term memory (Bi-LSTM) of RNN is presented. [ABSTRACT FROM AUTHOR]

This article introduces the hybrid encryption (HE)-based CIW using combined integer wavelet transform (IWT) with discrete cosine transform (DCT), where the HE performs on watermark image to encrypt the data securely before embedding it into cover image while both IWT, and DCT are used to extract the features from cover RGB image. In addition, the grey-wolf optimization (GWO) is employed for extracting the enhanced features obtained from IWTDCT. Then, LSB approach is employed for embedding the encrypted watermark image into the cover image, here after the proposed method is denoted as OHE-based IWT-DCT with LSB. Finally, the watermarked image is obtained using inverse transforms of IWT, and DCT with postprocessing operations. The simulations carried out on standard test images discloses the superiority of proposed OHE-based IWT-DCT with LSB approach as compared to stateof-art CIW approaches in terms of peak signal-to-noise ratio (PSNR), structural similarity (SSIM) index, mean square error (MSE), normalized cross correlation (NCC) and unified averaged changed intensity (UACI) values. [ABSTRACT FROM AUTHOR]

Road feature extraction from the remote sensing images is an arduous task and has a significant role in various applications of urban planning, updating the maps, traffic management, etc. In this paper, a new band combination (8652) to form a road index (RJ) from OLI multispectral bands based on the spectral reflectance of asphalt, is presented for road feature extraction. The 8652 is converted to road index by normalization. The morphological operators (Top-hat or Bottom-hat) uses on RI to enhance the roads. To sharpen the edges and for better discrimination of features, shock square tilter (SSF), is proposed. Then, an iterative adaptive threshold (lAT) based online search with variational min-max and markov random fields (MRF) model are used on the SSF image to segment the roads and non-roads. The roads are extracting by using the rules based on the connected component analysis. IAT and MRF model segmentation methods prove the proposed index (RJ) able to extract road features productively. The proposed methodology is a combination of saturation based adaptive thresholding and morphology (SATM), and saturation based MRF (SMRF), applied to OLI images of several urban cities of India, producing the satisfactory results. The experimental results with the quantitative analysis presented in the paper. [ABSTRACT FROM AUTHOR]

Segmentation of the iris is a crucial stage in an automated iris‐based recognition system. The performance of any biometric system primarily relies on how effectively the iris is extracted from the unwanted parts of an iris image. The process of iris segmentation is mainly affected by the noise artefacts such as eyelid/eyelashes occlusions, specular reflections, intensity inhomogeneities, and non‐circularity of the iris boundary. A novel and an efficient method has been proposed in this work to segment noisy and non‐circular iris boundaries. The mathematical modelling of morphological reconstruct fuzzy C‐means clustering (MRFCM) has been presented. The MRFCM based on improved particle swarm optimisation has been implemented before the segmentation in the recognition framework. The resultant images are then segmented by employing geodesic active contours incorporated by a new stopping function. The effect of the proposed segmentation method on iris recognition is observed through matching score distribution. The popular and publicly available datasets such as UBIRISv1, CASIA‐v3‐Interval, MMU1, and Mobile Iris Challenge Evaluation databases are considered for the evaluation of the proposed method. Recognition accuracy is validated and compared with the well‐existing methods. [ABSTRACT FROM AUTHOR]

Biometric systems performing identity recognition based upon extracted informative data from an individual are vital for security applications. The vital characteristics of an ECG signal depend upon its Characteristic points’ P, Q, R, S and T. In this paper, an effective feature extraction method is proposed, in which for each record of ECG, the best 6-PQRST fragments are extracted according to priority basis and their positions are normalized. A total of 72 different features are calculated, finally the performance of feature set is examined and compared using ANN. The proposed algorithm is tested for MIT-BIH ECG ID database signals. [ABSTRACT FROM AUTHOR]