Scribd
Upload
49 views3 pages

Water Volume Estimation - v1

The document proposes a pilot project to develop an AI-based system for near real-time monitoring and estimation of water volumes in surface water tanks using drone and satellite imagery. The methodology includes: 1) Generating a digital terrain model from drone images, 2) Automatically downloading satellite imagery every 12 days and using deep learning to map water spread, 3) Estimating water volumes by combining satellite water spread maps and the drone terrain model, 4) Developing a web GIS application to host the system and visualize/analyze results. The pilot will cover 134 tanks over 45 days with a budget of INR 33.1 lakh.

Uploaded by

judgeraja
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
49 views3 pages

Water Volume Estimation - v1

The document proposes a pilot project to develop an AI-based system for near real-time monitoring and estimation of water volumes in surface water tanks using drone and satellite imagery. The methodology includes: 1) Generating a digital terrain model from drone images, 2) Automatically downloading satellite imagery every 12 days and using deep learning to map water spread, 3) Estimating water volumes by combining satellite water spread maps and the drone terrain model, 4) Developing a web GIS application to host the system and visualize/analyze results. The pilot will cover 134 tanks over 45 days with a budget of INR 33.1 lakh.

Uploaded by

judgeraja
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 3

Near real time water spread monitoring and volume estimation of surface water tanks

using AI and geospatial technologies

Background

Continuous observation of surface water parameters such as water spread, level and volume
is a functional requirement for effective management of ecological and hydrological
resources. Frequent collection of these parameters by manual approach is prohibitively
expensive. Alternatively, remote sensing-based earth observation satellites are recognized as
an optimal means to monitor water bodies over larger areas. Tamil Nadu is a tropical region
where most of the areas are usually under severe cloud cover for a major part of the year.
Hence, the satellites with special sensors like microwave which are resistant to cloud cover is
highly desirable. But unfortunately, these microwave images are quite expensive for a
developing economy like India. Since 2016, the European Space Agency (ESA) has been
providing us with microwave images for free of cost from its satellite called Sentinel-1,
which makes statewide satellite-based water mapping economically feasible. With these
satellites, it is feasible to map the water spread area. However, to estimate the volume, the
elevation (terrain) information is required in addition to water spread area. But satellite
images (e.g., Stereo images from space borne platforms) are not sufficient enough to generate
terrain information for very shallow water bodies like irrigation tanks. In recent years, drone
technology is revolutionizing the geo-information field and clearly becoming a desirable
alternative to the conventional satellite and manned aerial data collection formats used in
many remote sensing applications. Particularly, its capability to acquire high resolution
overlapping images could be of help here to generate a highly accurate digital terrain model
through a photogrammetric approach. To this end, we hypothesis that it is feasible to
periodically estimate the volume by integrating the periodic water spread information from
temporal satellite images (e.g., every 12 days) with digital terrain model generated via one-
time drone survey. Also, we propose for a pilot project to demonstrate the hypothesis by
developing an end-to-end AI based water volume estimation system using satellite and drone
images. The study area, methodologies, budget and timeline for the proposed pilot is
provided in the sections below.

Study area
134 tanks and 1 reservoir in lower Vellar sub-basin
Methodology
The development process of proposed AI based tank volume estimation system for the pilot
includes the following four key tasks:
1) Digital Terrain model via Drone images: The drone images will be captured by
preparing appropriate flight plans. Once the images are captured and uploaded to
our server, it will undergo data processing chain to generate the required DTM: (a)
alignment between overlapping images, (b) optimize the camera position and
orientation, (c) dense point cloud generation of a 3D mesh, (d) ground filtering with
or without above ground object points, (e) eliminating or keeping all-natural
(vegetation) above-ground objects from the dense point cloud, (f) if the above objects
are eliminated a mesh, a DTM is created, and (g) if the above objects are kept in the
dense point cloud, a DSM and orthomosaic are created. The overall workflow is
shown below:

2) Automated system for periodic estimation of water spread area from satellite
images: Manual downloading of satellite imageries in large volume and frequency is a
tedious, labor and time intensive process. However, most of the satellite image providers offer
an API service for automated download. By utilizing this facility, a module will be developed
to automate this data download process. This module will automatically download the
satellite image for every 12 days for the specified region. Subsequently, the downloaded data
will be automatically pre-processed and converted into an analysis-ready data. A deep-
learning based LSTM module will be used to generate the water spread map by classifying
the analysis-ready data. The spread maps will be provided in both raster and vector formats.

3) Volume estimation by integrating satellite image-based water spread and drone


image-based terrain model: This module will consume both waters spread map from
satellite images and terrain model from drone images. Subsequently, it will estimate
the volume by fitting the area-volume-elevation curve. The water volume information
will be provided in terms of 2D and 3D maps in GIS ready format and textual
information in database file format.

4) Customized open-source based Web-GIS Application: Design and Develop


GIS based web application for water resource department with access governance to
host the above modules 2 & 3. The application will 1) allow the user to upload DTM
generated from drone images, 2) specify the area for continuous monitoring either by
selecting the area in map view by drawing a polygon or uploading the shape files, 3)
triggering system to generate the surface area and volume for a given region and time-
interval and 4) Cater to the viewing, analysing and downloading of maps and reports.

Timeline, budget and deliverables:


S.N Components Features & Deliverables Budget (INR) Timeline (in days)
O
Start End
1 Products using Contour map, DEM, DSM, 15,60000 0 20
Drone images: Orthomosaic for an area of (INR 19,500 per
80 sq.km covering 134 tanks sq.km inclusive
and 1 reservoir in lower of image
Vellar sub-basin capturing,
processing and
travel & boarding
charges for
pilots)
2 Automated 1. Module for automated 7,50000 0 30
system for data download and pre-
periodic processing
estimation of 2. Deep learning module to
water spread automatically classify the
area from image and map the water
satellite spread area
images 3. Water spread map in
raster and vector format
3 Tank volume 1. Module for volume 1,50000 20 30
estimation estimation using water
spread from satellite image
and DTM from drone image
2. 2D and 3D maps of water
volume and reports in
textual formats
4 Customized 1. An user-interface to host 8,50000 0 45
open-source and invoke the above
based Web- modules
GIS 2. Data governance to
Application receive, store, process and
retrieve data
3. To visualize, analyse and
download the maps and
reports in specific formats
Total 33,10000 0 45

You might also like