Removal of Cloud and Shadow Influence From Remotely Sensed Images Through Landsat8/Oli/Tirs Using Minimum Distance Supervised Classification

Gound, Renuka S.; Thepade, Sudeep D.

doi:10.21817/indjcse/2021/v12i6/211206118

Cited by 3 publications

(2 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…So, the pixel is assigned to the class on the basis of the spectral signature which one is closer to it by using the Euclidean Distance measure. [16] Ground truth validation using minimum distance gives overall accuracy of 99.65%, as shown in Table . 4.…”

Section: Minimum Distancementioning

confidence: 99%

“…The most critical step of pre-processing is to recognize and remove the effect of cloud and their shadows from such images. [4] [16] [22] It is the prerequisite for existing techniques of cloud and shadow detection methods to get their location information previously and they absolutely failed to utilize the hidden information over the regions of a cloud and shadow. [5] [23] Normally, complimentary images are required to discover the ground truth under the cloud region, these images are known as reference images.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Validation of Ground Truth of Remotely Sensed Data From Sentinel-2 (Msi) Using Supervised Classification, After Combined Cloud and Shadow Effect Removal

Gound¹,

Thepade²

2022

INDJCSE

View full text Add to dashboard Cite

SENTINEL-2 is utilized for the earth observation system to acquire optical imagery of land as well as coastal water areas with a high spatial resolution (10m to 60m). Earth observation supports various services like agricultural monitoring, ground truth classification, and identification of environmental changing effects over land cover. If images acquired are under the influence of climatic noise such as the occurrence of shadow and cloud, the clarity of ground truth images may get reduced. The present article provides an integrated method of the supervised classification, and segmentation of land cover boundaries using NDVI (Normalised Difference Vegetation Index), with detection of cloud and shadow pixels using SENTINEL-2 band images. Images of places Around Lonavala, Pavana lake, Mulashi Lake, Tamhini Ghat, Changad, Lavasa, etc. of India from 2020 to 2021 were used in experimentation. The pixels under the influence of cloud and shadow regions are detected and replaced by reconstructing an image using a reference image. Cloud and shadow, free images are further classified to validate the ground truth. The proposed system uses a Maximum Likelihood Classifier (MLC), Random Forest (RF), and Minimum Distance Classifier to examine the correctness of ground truth. Overall Accuracy achieved with the Maximum Likelihood Classifier, Random Forest Classification, and Minimum Distance Classification are 98.20%, 96.65%, and 99.65% and K-hat is 0.97, 0.95, 0.98 respectively. From statistical analysis and findings, shows that the proposed system can successfully denoise images obtained through the SENTINEL-2 satellite in preprocessing step and validated the ground truth obtained.

show abstract

Section: Minimum Distancementioning

confidence: 99%