Feature Extraction from Image Sensing (Remote) : Image Segmentation

In this research paper, an overview of computer methods for segmenting continuous-tone images into meaningful parts and characterizing these parts with “features” is presented. Image segmentation is an essential pre-processing step in computer vision and image processing softwares. This paper discusses various techniques and algorithms used for image segmentation, including both supervised and unsupervised methods. The performance of these techniques is evaluated on different test images, and future research directions are proposed. Feature extraction from image sensing is a critical step in multimedia processing. This paper explores the process of feature extraction from remotely sensed images. With the increasing use of remote sensing technology, it is crucial to have effective and accurate methods for extracting features from remotely sensed data, which are utilized in various applications such as land-use mapping, environmental proctoring, and in natural-disaster management. This paper discusses several techniques for feature extraction, including pixel-based, object-based, and machine learning-based approaches. Additionally, it highlights the difficulties associated with feature extraction from remote sensing data, such as atmospheric effects, radiometric and geometric distortions, and the presence of noise. The paper concludes by discussing future directions in remote sensing feature extraction, including the integration of deep learning methods and the development of more robust methods has the ability of handling large-scale datasets. With the help of deep learning methods and algorithms, especially CNN’s, has tremendously improved the accuracy of image scene classification. In this study, a CNN model which is already trained was utilized to fetch-out deep features from images at various network layers, including the average pooling layer.