Search

One of the most difficult challenges in computer vision is image segmentation. It's the process of breaking down visual data into segments to facilitate image analysis. There are a variety of image segmentation techniques. Some examples include edge detection-based methods, region based methods, cluster based methods, partial differential equation based method, waterhed based method, and neural network based method. Image segmentation and classification is the focus of this research. Remote sensing image analysis is used to observe the Earth's surface using images taken by satellites. The primary purpose of an image classification-based system is to attribute semantic labels to the captured images. These semantic labels can then be used to classify the images in a semantic order. The suggested method uses satellite photos as its input. Deep learning (DL), a rapidly developing strong tool that surpasses these current classification approaches and gives more accurate categorization, has made it possible to achieve state-of-the-art outcomes in many remote sensing applications. The K-Means clustering algorithm can be employed to segment the remote sensing data in the current instance. This clustering method yields higher quality outcomes than conventional approaches. [ABSTRACT FROM AUTHOR]

The accurate and automatic categorization of lung nodule malignancy and benign is a challenging task in modern disease diagnosis. The performance of the lung nodule predictive model gets down due to noisy raw input data and the wrong model. ML and DL based Computer Aided Diagnosis Systems show excellent performance in many healthcare applications. Uncertainty measuring in most Deep learning methods is insufficiently considered. Therefore, to deal with this gap new hierarchical helpful Multilevel feature-fusion system with an uncertainty approachable module for image categorization is created. An accurate nodule detection and early treatment purposes a 3DCNN fusion model is used. A Multi_View_Divide and Rule (MV DAR) system is used to learn from reliable and unclear annotations for lung nodule malignancy predictions on chest CT scans. For benign-malignant pulmonary nodule identification a Semi_ Supervised Deep Transfer Learning (SDTL) system is used. To divide malignant from benign nodules using limited chest CT data multi view knowledge-based collaborative (MV KBC) deep model is used. For reducing the number of false positives, a framework of Ensemble-CNN (E-CNN) is used. With different sizes of lung nodules detection, a 3D Convolutional NN (3D CNN) is used. Then, another 3D CNN is presented for the consequent false positive reduction. [ABSTRACT FROM AUTHOR]