Revamping land coverage analysis using aerial satellite image mapping

Abstract

Summary The customary terrestrial survey involves direct measurement of the land dimension in a particular area. The information accuracy of land dimension changes person to person in the existing method. Moreover, it consumes more time to conduct an assessment and produces a precise outcome. Hence, Remote Sensing offers an effective solution for accurate measurement using the Land Satellite Image Mapping with Geographic Positioning System (GPS) values. Therefore, the proposed model concentrates on the analysis of the object using Landsat Image through categorization based on the field cover analysis. The Geographic Information System (GIS) facilitates convenient measurement to determine the land dimension along with Georeferential tag. In addition to that, the features of each object are preprocessed to reduce the noise and the acquired information is stored in the classifier. Also, the features are trained using 2‐fold validation, which will improve efficiency during detection. Finally, the soft hyperplane discriminators of Gaussian produce better separation accuracy compared to feature classes and object prediction. Furthermore, the Bag of feature model performance is improved by working on Hue, Saturation, and Intensity color map with multifeatured extractions on the sample images. The proposed model was trained with 1800 image samples of various land covers. On testing, the classifiers aggregate 91.4% accuracy on cross‐evaluation between true class and predicted class using medium Gaussian Support Vector Machine.

MAST: An Earthquake-Triggered Landslides Extraction Method Combining Morphological Analysis Edge Recognition With Swin-Transformer Deep Learning Model

“…In recent years, the development of machine learning methods has brought new opportunities in the automatic extraction of ETLs, such as support vector machine (SVM) and random forest. SVM is a supervised learning algorithm, and in automatic ETLs extraction, SVM can be trained with labeled landslide and nonlandslide samples to learn the boundary features and classification rules of landslides [3], [4], [5], [6]. SVM exhibits superior classification performance and generalization ability in the context of landslide boundary extraction.…”

Review on remote sensing methods for landslide detection using machine and deep learning

“…However, all these methods require designing of interpretation logics or extraction of features present in the image, resulting in complex algorithm design and limited the potential of algorithm improvement 18,28 . Reference 29 developed a model focusing on object‐based analysis using Landsat Image through layout based on field cover analysis. GIS promote suitable measurement techniques for determining dimensional information of land along with Georeferential tag 29 .…”

“…Reference 29 developed a model focusing on object‐based analysis using Landsat Image through layout based on field cover analysis. GIS promote suitable measurement techniques for determining dimensional information of land along with Georeferential tag 29 . In addition, pre‐processing of object features is performed to scale‐down noise and information acquired stored in the classifier.…”

“…Model training was done by using 1800 land cover image samples having six distinct classes. Testing results revealed classifier accuracy is 91.4% while performing cross‐evaluation between predicted class and true class using Gaussian SVM 29 . Deep learning is a subdivision of machine learning which has become an important research area.…”

Multilabel classification of remote sensed satellite imagery

“…Due to the high variability in the composition of the terrain and the voluminous amount of terrain categories, optimal identification with labeling for terrain types in very high resolution (V.H.R.) Data have become challenging to solve using machine learning methods 8‐10 . This challenge has ignited interest in the domain of remote sensing 6,11 .…”