Spontaneous combustion of coal is a critical problem encountered by mining and thermal power industries. Usually, coal is stored in open area in the form of stockpiles in the coal mines and thermal power plants. In this paper, we have focused on localization of open-cast coal mines and coal stockpiles using satellite images to automate the entire process of prediction of spontaneous combustion in the coal stockpiles. We have used USGS Landsat-8 Satellite images, collected from various coal mines and thermal plants across the world. The satellite images consist of 11 bands including Red, Blue, Green, Near Infrared (NIR), and Shortwave Infrared (SWIR). Apart from the reflectance measurements obtained from these bands, we also use standard indices including Normalized Difference Vegetation Index (NDVI), Normalized Difference Moisture Index (NDMI) and Normalized Difference Water Index (NDWI) as the features to train the models. The ground truth for the training dataset has been created by manually annotating these images for three classes: coal mines, coal stockpiles and water bodies. The Fully Convolutional Network (FCN) based U-Net architecture has been trained to develop two models to classify pixels between (A) Coal Mine and Water and (B) Coal Stockpile and Water. In this paper, we present an exhaustive experimental results to demonstrate the effective localization of coal mine and coal stockpiles using the proposed FCN based approach.
In this paper, we envision the use of satellite images coupled with GIS to obtain location specific crop type
information in order to disseminate crop specific advises to the farmers. In our ongoing mKRISHI R
project, the
accurate information about the field level crop type and acreage will help in the agro-advisory services and supply
chain planning and management. The key contribution of this paper is the field level crop classification using
multi temporal images of Landsat-8 acquired during November 2013 to April 2014. The study area chosen is Vani,
Maharashtra, India, from where the field level ground truth information for various crops such as grape, wheat,
onion, soybean, tomato, along with fodder and fallow fields has been collected using the mobile application. The
ground truth information includes crop type, crop stage and GPS location for 104 farms in the study area with
approximate area of 42 hectares. The seven multi-temporal images of the Landsat-8 were used to compute the
vegetation indices namely: Normalized Difference Vegetation Index (NDVI), Simple Ratio (SR) and Difference
Vegetation Index (DVI) for the study area. The vegetation indices values of the pixels within a field were then
averaged to obtain the field level vegetation indices. For each crop, binary classification has been carried out
using the feed forward neural network operating on the field level vegetation indices. The classification accuracy
for the individual crop was in the range of 74.5% to 97.5% and the overall classification accuracy was found to
be 88.49%.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.