Automatic Semantic Segmentation and Classification of Remote Sensing Image Data for Flood Detection Using Novel LSTM Neural Network

Floods are one of the main threats to human life and the property of natural disasters, especially in highly populated urban areas. Fast and accurate extraction of submerged area risky to supporting emergency planning and providing damage assessment in spatial and temporal measurements. Satellite mu...

Celý popis

Uloženo v:

Podrobná bibliografie
Vydáno v:	SN computer science Ročník 5; číslo 8; s. 992
Hlavní autoři:	Sonavale, Amruta, Chakkaravarthy, Midhun, Srinivasa Rao, Surampudi, Salleh, Hishamuddin Bin M., Jadhav, Jagannath
Médium:	Journal Article
Jazyk:	angličtina
Vydáno:	Singapore Springer Nature Singapore 01.12.2024 Springer Nature B.V
Témata:	Accuracy Advances in Computational Approaches for Image Processing Algorithms Automation Back propagation Banded structure Binomial distribution Classification Cloud Applications and Network Security Computer Imaging Computer Science Computer Systems Organization and Communication Networks Data analysis Data collection Data Structures and Information Theory Datasets Deep learning Edge detection Error analysis Error detection Floods Image enhancement Image resolution Image segmentation Impact analysis Information Systems and Communication Service Internet of Things Literature reviews Methods Neural networks Original Research Pattern Recognition and Graphics Performance evaluation Pixels Propagation Remote sensing Satellite imagery Saturation (color) Semantic segmentation Semantics Sensors Software Engineering/Programming and Operating Systems Spectrum analysis Statistical methods Technology assessment Threat evaluation Time series Vision Water Wireless Networks Adaptive median filtering Flood detection Neural network algorithm LSTM Versatile Linear Edge Detection (VLED) algorithm Structural object son the RSI
ISSN:	2661-8907, 2662-995X, 2661-8907
On-line přístup:	Získat plný text
Tagy:	Přidat tag Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!

Popis
Shrnutí:	Floods are one of the main threats to human life and the property of natural disasters, especially in highly populated urban areas. Fast and accurate extraction of submerged area risky to supporting emergency planning and providing damage assessment in spatial and temporal measurements. Satellite multispectral images have limited bands, low resolution, and information that can be analyzed. The bands are merged to form a unified image that incorporates data from all bands. However, current contouring techniques are affected by chromatic aberrations. This study uses intensity-hue-saturation and higher-order statistics to enhance the spatial and spectral information of remote sensing images in combination with segmentation and classification methods. High performance can be shown by implementing edge detection technology to identify objects using structure in Remote Sensing Images (RSI). The parameters used to evaluate the implementation of the proposed edge detection method included root mean square error, correlation coefficient, structural similarity index measure, and the error associated with the mean spectral analysis. The dimensionality of multiband RSI can be reduced based on higher-order data statistics using independent component analysis. Furthermore, images can be clustered using high-resolution panchromatic RSI models. The proposed technique can convert the pixel powers into the sign powers of adjacent higher-order partitions during region segmentation. Pixel intensities can be analyzed using preprocessing techniques such as denoising, which gradually stabilizes the object’s mean value. This method is shown not to impact the original seed level. In this proposed work wiener filtering is usedfornoisesuppressionintheremote-sensingimage.Afterthesecondstageofthe process, the structural region is extracted from the preprocessing image. After detecting structural regions in the RSI, the different feature values are obtained for further classification. The proposed Long Short-Term Memory (LSTM) neural network algorithm was used to predict the flood. Implementing the proposed LSTM neural network algorithm can automatically transform large-scale data sets and implement non-linear decision-making functions through the histological characteristics of multiple layers of neurons. This work compares the accuracy of flood forecasting in the collected data set using the proposed LSTM neural network algorithm. The performance of the proposed technology is analyzed in terms of accuracy, sensitivity, and specificity.
Bibliografie:	ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 14
ISSN:	2661-8907 2662-995X 2661-8907
DOI:	10.1007/s42979-024-03336-9