Smart India Hackathon 2020

ISRO Problem Statements and Data

All Querys related to ISRO problem statements may be sent to hackathon2020@sac.isro.gov.in

NM370-ISRO
Air/Water turbidity estimation mobile app:

Development of App for measurement of Air and water turbidity using smartphone.

Air Quality: It would involve imaging sun by placing appropriate light diffusing mechanism in mobile, developing App for image processing taking care of different sun viewing geometry to estimate the atmospheric optical thickness.

Water Quality: would involve imaging water through mobile and developing App for image processing to estimate turbidity of water.

Dataset:

This application does not require a priori dataset. You have to develop app to read the R G B values from mobile photo convert into some index and relate it with low medium and high turbid condition. You can also make a small Secchi disk(read literature) to measure the Secchi depth which is one of the quantification of water turbidity.


NM371-ISRO
Automated mapping of trees/farms in satellite image:

Design and Implement algorithm for automated mapping of trees/plantation and farmland from 5m multispectral multi-temporal (LISS IV) data.

Participants need to develop algorithm for automatically classifying vegetation areas as either farm or tree/plantation by analysing seasonal changes apparent in multi-temporal satellite imagery.

Dataset:
Bareily.tif download
Gondal.tif download
Junagadh.tif download
Moradabad.tif download
Rajkot.tif download
Satna.tif download

About Data: Information about datasets

NM372-ISRO
Extraction of crop cycle parameters from multi-temporal data:

For a given set of multispectral multi-temporal data with timestamp of one year or more, develop and implement an algorithm for extracting crop cycle parameters.

Participants need to develop a high-performance algorithm to analyse multi-temporal data at each pixel to extracting parameters such as date of sowing, date of harvesting and number of harvests based on temporal profile.

Dataset:
Clipped_NDVI.zip download

NM373-ISRO
Detecting clouds and predicting their movement from INSAT imagery:

INSAT-3D and INSAT-3DR are two geostationary meteorological satellites of India having 6 channel Imager and 19 channel Sounder payloads. The imagery captured in the visible range by INSAT satellite can be used to detect clouds.

Participants need to develop and implement algorithm to detect clouds in INSAT satellite images and predict the location of clouds in subsequent images.

Dataset:
INSAT3D_TIR1_India download
INSAT3D_VIS_India download

NM374-ISRO
Nowcasting of Meteorological Satellite Images using AI/ML techniques:

INSAT-3D and INSAT-3DR are two geostationary meteorological satellites of India having 6 channel Imager and 19 channel Sounder payloads. INSAT-3D and INSAT-3DR Imager acquire images over its footprint every 30 Min. In order to have better revisit these satellites are programmed to acquire images in staggered mode, which provides images over Indian region every 15 Min.

Nowcasting refers to forecasting for a shorter duration (3 to 6 Hrs.). Nowcasting of meteorological images will help in forecasting images with for next 3 hrs.

These Nowcasted images help synoptic meteorologists, administrators and common man for better interpretation and decision support during extreme weather events.

Machine learning (ML) provides techniques for predicting the new outcomes based on previously known results and Artificial Intelligence (AI) helps in decision making.

Develop an AI/ML based software to generate nowcasted satellite images and its animation loop for next 3 Hrs. at an interval of 30 Min. using data form INSAT-3D and INSAT-3DR

Dataset:

Please contact Ms. Shivani Shah (shivanishah@sac.isro.gov.in)


NM377-ISRO
Web-map data visualization using augmented reality globe

Develop an app to visualize globe in augmented reality on any surface. (Participants can use any physical marker or placeholder).

When viewing the physical marker/placeholder from the phone camera, an augmented reality globe must appear over the placeholder. Moving the camera

It should be possible to overlay any OGC Web Map Service (WMS) on the globe. The application must provide interface for selecting the data to be overlaid on the globe.

Dataset:

Please visit https://cesium.com/docs/tutorials/terrain/ for tutorial.

For free terrain data see https://www.maptiler.com/blog/2018/08/free-terrain-tiles-for-cesium.html .

For free map imagery please visit official bing maps website.


NM378-ISRO
AI based crop identification mobile app.

Develop a mobile application that can identify crop using only field photo of a crop. The team must target at-least 10 different crops for demonstration.

The application will allow the user to take photos and automatically identify the crop. The photo and crop information along with geolocation information should be stored in an internal database which can be exported/emailed.

Dataset:

Use google image search to obtain field photos of different crops.


NM380-ISRO
App for recording and playing geotagged videos.

Develop a mobile application for recording and playing geotagged videos.

Unlike photos in which geotag data is of a single point and orientation pair, for videos geotag data is a sequence of point and orientation pairs.

The mobile application should have two views. In one view the recorded video should play while simultaneously plotting field-of-view (orientation) cone and marker on an interactive map in the other view in a synchronized manner.

The position shown on the map should match the play position of the video.

The geotagged locations should be exportable in KML format with time tag information.

Dataset:

NM381-ISRO
App for identification of sky regions in a photo:

Generating local sky horizon has important applications for analysis of solar energy potential in an urban setting.

Develop a mobile application for automatically detecting sky pixels in a photograph. The application should generate a mask image consisting of sky pixels marked in white colour in the image and other pixels marked in black colour.

Further, using information about camera optics, the application should give angle of elevation of the lowest sky pixel for all pixel columns in the mask image.

Dataset:

Please use dataset available at the following link http://cs.uky.edu/~jacobs/datasets/skyfinder/


NM383-ISRO
Drone route planning:

Develop an application for automatically planning a route (for shortest time to cover area) and schedule of drones for mapping a given area. The input provided will be.

  1. Map of area to be covered. (Shapefile)
  2. Number of drones.
  3. Range (in km based on battery life)
  4. Top speed
  5. Location of automatic charging station

The software must have features for visualizing a simulated animation of the plan.

Dataset:

Use simulated (self-generated) dataset.


NM385-ISRO
Virtual reality based Earth/Moon explorer:

Develop application to visualize Earth/Moon globe in virtual reality using a wearable mobile based headset (using Google Cardboard VR or similar technology).

User should be able to select a location (e.g. peak of mount Everest) and explore the area in virtual reality in 3D as if he/she was present there.

Dataset:

Please visit https://cesium.com/docs/tutorials/terrain/ for tutorial.

For free terrain data see https://www.maptiler.com/blog/2018/08/free-terrain-tiles-for-cesium.html.

For free map imagery please visit official bing maps website.

For moon terrain and imagery use the following links


NM386-ISRO
Automated land use classification using AI/ML:

Develop a deep-learning based software for automatically classifying land-use from multi-temporal multi-spectral high-resolution satellite imagery.

The developed model should be scalable/efficient to allow rapid mapping of incoming datasets and must incorporate a web-based viewer for visualizing input as well as classified output. The viewer interface must also allow the user to visualize changes that have occurred within a given timeframe.

Dataset:

Data is available on Open Data archive of ISRO-BHUVAN. One such example of data is given below

L3_SAT_8B_V1_73E18.75N_E43H01_01Feb13 download
L3_SAT_8B_V1_73E18.75N_E43H01_03Nov11 download
L3_SAT_8B_V1_73E18.75N_E43H01_07Feb12 download
L3_SAT_8B_V1_73E18.75N_E43H01_13Oct08 download
L3_SAT_10B_V1_73E18.75N_E43H01_04Apr15 download
L3_SAT_10B_V1_73E18.75N_E43H01_06Nov15 download
L3_SAT_10B_V1_73E18.75N_E43H01_09Apr14 download
L3_SAT_10B_V1_73E18.75N_E43H01_17Jan16 download
L3_SAT_10B_V1_73E18.75N_E43H01_18Oct14 download
L3_SAT_10B_V1_73E18.75N_E43H01_20Feb14 download
L3_SAT_10B_V1_73E18.75N_E43H01_24Dec15 download

For queries please contact:

Mr. Prasun Kumar Gupta (prasun@iirs.gov.in) & Mr. Prabhakar Alok Verma (prabhakar@iirs.gov.in)


NM387-ISRO
GNSS Reflected Signal Coverage Simulator:

Apart from determining location, GNSS can also be used for estimating geophysical parameters using its reflected signal. Based on the satellite altitude/position, the receivers receive reflected signals from a particular region. The reflected signals get changed with change in the surface and terrain characteristics.

In this problem, the teams need to develop a software suite that will provide the following features.

  • Run dynamic simulations to identify the reflected signal area coverage of a particular satellite with its time and position.
  • Tool to analyse the effect of the obstacle in the identified coverage area.

The teams will be provided the following input

  • Orbital parameters of all GNSS satellites
  • Digital elevation model of the study area
  • On ground locations of the GNSS receivers
For queries please contact:

Dr. Bhaskar R. Nikam (bhaskarnikam@iirs.gov.in), Mr. Raghavendra Sara (raghav@iirs.gov.in) & Dr. Vaibhav Garg (vaibhav@iirs.gov.in)


NM388-ISRO
Hyperspectral image analysis tool:

Hyperspectral Remote Sensing data provides data in large contiguous number more than 100 wavelength bands. In hyperspectral data the feature and its abundance is measured using its spectral response across these bands. Each feature behaves differently in different wavelength band and sensitive to particular wavelength region.

Develop a tool to analyse the spectrum of each pixel in a hyperspectral remote sensing data cube for absorption dip, width of absorption and other characteristics at different percentile of the absorption depth. The developed tool must also provide functionality to prepare map of each of these properties and its visualization.

For queries please contact:

Dr. Vaibhav Garg (vaibhav@iirs.gov.in) and Dr. Bhaskar R. Nikam (bhaskarnikam@iirs.gov.in)


NM389-ISRO
Web based volume rendering and 3D/4D visualization of Model Forecast:

The numerical weather models are used for generating forecast, these model forecast contains 4D information (i.e., Latitude, Longitude, Height/pressure and Time). These model forecasts are gridded at a defined sampling interval, and are very useful for planning and decision supports.

WebGL is a javascript API for rendering high-performance interactive 3D and 2D graphics.

The participants must develop a web-based tool for 3D/4D visualization of model forecast.

Dataset:

Please contact Ms. Shivani Shah (shivanishah@sac.isro.gov.in)


NM390-ISRO
AI/ML based system for deriving value added parameters using satellite surface observations:

Sub-surface information of the ocean like the mixed layer depth (MLD), sonic layer depth (SLD), tropical cyclone heat potential (TCHP) etc. is very useful in wide variety of applications such as air-sea interaction studies, Naval applications, cyclone genesis and forecasting etc. This information is available either from in situ measurements or from Numerical models.

In situ measurements in the ocean are quite sparse while the numerical models have inherent errors in simulating this information. Satellites provide wide spatial coverage of the oceans; however, they only give ocean surface information. Assuming that the derived quantities of the deeper ocean such as MLD, SLD, TCHP etc have signatures in the ocean surface as well, an artificial intelligence based technique linking the satellite surface observations of the ocean with these sub-surface quantities, may be envisaged.

Satellite observations of sea surface temperature (SST), sea surface salinity (SSS), sea surface height anomaly (SSHA) and sea surface wind speed (WS) collocated with MLD, SLD, TCHP derived from in situ measurements (ARGO floats/moored buoys) can be used as training data set where we try develop a relationship between SST, SSS, SSHA, WS and the derived quantities MLD, TCHP and SLD.

In this problem the participants must develop a software to generate MLD, SLD and TCHP using satellite data (SST, SSS, SSHA, WS)

Dataset: download data

NM391-ISRO
Reconstruction of missing data in Satellite Imagery:

Short Wave Infra-Red(SWIR) detectors used in satellite imaging cameras suffer from drop outs in pixel and line direction in raw data.

Develop software to reconstruct missing parts of a satellite image so that observers are unable to identify regions that have undergone reconstruction. Study shall also compare the performance of the proposed with existing state of art technique results.

Dataset: will be uploaded shortly

NM392-ISRO
Depth Estimation of Valles Marineris using ISRO’s Mars Color Camera (MCC) images:

Valles Marineris is Grand Canyon system present along the equator of Mars. The Valles Marineris is a large tectonic crack present on the Martian crust running up-to a length of around 4000 km. Mars Color Camera (MCC) captured multiple images of Valles Marineris at varying spatial resolution which can be used to estimate the depth of the canyon system.

Develop software to generate depth map of Valles Marineris using image captured by MCC.

Dataset: download dataset

NM393-ISRO
Field Data Analysis and Automated feature validation from crowd sourced field photos;

Bhuvan is widely used for crowd sourcing and millions of photos are uploaded everyday pertaining to different subjects of interest like crops, dams, water bodies, prominent locations, grievances, infrastructure etc. Also, this has become an important platform for most of the ministerial activities, where in accountability and financial sanctions etc are also linked. Such important activities need more focused and automated validation mechanisms for understanding the photographs and relevance for the purpose. Hence, extracting information from the photograph is essential.

Develop software to automatically locate and extract text from field photographs. For example, given and image containing a traffic sign, the traffic sign should be identified and its text must be extracted.

Dataset: https://sid.erda.dk/public/archives/daaeac0d7ce1152aea9b61d9f1e19370/published-archive.html

NM394-ISRO
Air pollution hot spots detection and identifying the source trajectories using ML/AI techniques.

At present, air pollution is a global problem. India is also a big sufferer of this problem. India signed COP21 agreement for cutting the carbon emissions from 2025. Hence a study identifying the hot spots of pollutants and their transport namely carbon monoxide (CO), sulphur dioxide (SO2) and oxides of nitrogen (NO+NO2) using advanced data analysis techniques.

Satellite provides columnar concentration of these pollutants which are 90% representation of surface concentrations. Pollution sources are mainly from the land surface activities.

Satellite provides these observations on daily basis with different spatial resolutions. Challenges involved in the current statement is mining the datasets from different satellites parameters and providing the final output with moderate spatial resolution on pollution information. Hence information will be useful for change detection analysis. Identification of source pathways.

Participants must develop AI/ML based software/algorithm to identify/analyse

  1. Location of hot spots.
  2. Long-term occurrence of hot spots and changes.
Dataset:

Satellite based data (freely available),


NM395-ISRO
Identify shifting cultivation locations in dense temporal stacks

Shifting cultivation is an agricultural system involving the clearing and burning of natural vegetation, followed by the cultivation of new fields for a few years. This is followed by a period of fallow during which the vegetation regenerates, after which the cycle begins all over again. This is typically followed in the hilly regions of the country. The cycle is likely to be over 7 to 20 years.

Using multi-temporal satellite imagery. Develop AI/ML based software/algorithm to map/visualize/analyse

  1. Areas exhibiting shifting cultivation.
  2. Long-term changes.
Dataset:

Imagery from Google Earth Engine (West Garo Hills, Meghalaya 25◦47' to 26◦10'N latitude and 89◦45'to 92◦47'E longitude


NM396-ISRO
Sentiment Analysis from text feedback:

Webportals like Bhuvan get vast amount of feedback from the users. To go through all the feedbacks can be a tedious job.

Develop software to categorize opinions expressed in feedback forums. This can be utilized for feedback management system. The software must provide the following output.

  1. Classification of individual comments/reviews.
  2. Determining overall rating based on individual comments/reviews.
Dataset:

The Multi-Domain Sentiment Dataset contains product reviews taken from Amazon.com from many product types (domains).

http://jmcauley.ucsd.edu/data/amazon/

NM397-ISRO
Voice command driven Web-GIS Applications (mobile/ desktop):

As on date, Bhuvan is driven by GUI based features. Develop software for voice based navigation of Bhuvan portal and/or applications listed on Bhuvan portal.

Solution can be provided for at-least 2 languages-

  1. Language-1: Hindi or any other Indian regional language. (Participant may have to prepare the data-sets for desired language, and may also use online datasets, If available)
  2. Language-2: English (Participant may use online available datasets).

Specific focus should be on providing voice-based navigation for Web-GIS applications.

Dataset:

NM398-ISRO
Change detection and extraction of information / features of interest in RS images using time series information:

Bhuvan - The EOS visualisation platform has large amounts of satellite data along with derived information from various sensors and collateral data. Users of Bhuvan are widely interested in downloading data and carrying out change detection analysis for each of their areas of interest as on date. Utilities for change detection on Bhuvan platform, directly using knowledge base / Deep Learning based algorithms for identifying the changes in given / selected time series data and extracting features of interest changed over time will be a very good value addition for water resources, urban dynamics, infrastructure monitoring and disaster management. This is possible only by identification / detection of features on the fly from RS images, which are more amenable for noise from different implicit processes in atmosphere and data processing algorithms.

Given time series tiles of Bhuvan Imagery / RS image ( 2 dates ) the system should

  1. Identify changes and highlight areas of change.
  2. Extract features that are evident from image 2 and not in image1, categorise in known lists i.e. water, roads, buildings, parks, trees
  3. Find features that are missing in image2 w.r.t. image 1.
Dataset:

Datasets are available in Bhuvan of different times and sensors Also sample datasets of other areas / sensors can be provided as per subjective requirement from - LISSIV, LISSIII, AWIFS or data as permitted by data policy from NRSC.

SAC website for data on SIH-2020 ( LISSIV / LISS III Images )


NM399-ISRO
Efficient Communication scheme for Human Space Flight Programme

To increase the transmission reliability and better service quality, connectivity through multiple links or networks, i.e., multi-homing, is considered which provides robustness and offers concurrent data transfer over multiple paths. The real-time data streaming requires low delay, jitter free transmission medium. For real-time multiplexed streams, comprising audio, video and data, the transmission reliability requirement is different for each type of stream. So, there is a need of an efficient transmission technique for such multiplexed streams with different reliability parameter settings for each stream over a multi-homed networking environment. It is also required to have a suitable handoff mechanism in case of link failover with minimum handoff latency.

The participants must design implement and demonstrate transmission protocol with point-to-point real-time Audio, video and data conferencing application over multi-homed network.

DATA : Audio, video stream may be generated through webcam and mic. attached to a laptop/desktop machine for testing and evaluation.

Dataset: will be uploaded shortly

NM400-ISRO
Link Adaptive Speech Compression scheme for Human Space Flight Programme:

The objective is to design light weight end to end communication protocol which facilitates satellite link adaptive speech compression schemes. When the satcom link becomes erroneous and information rate needs to be reduced to maintain the link quality, the speech encoding scheme needs to be changed to provide lower rates. The information on type of encoding used needs to be made available to the receiver as part of end to end protocol to support decoding.

Desired Outcome: The participants must design, implement and demonstrate end to end communication protocol with point-to-point real-time speech communication (voice call) under variable link conditions using different available open source speech compression schemes like G729, CELP, MELP etc. The rate adaptation with link condition to be demonstrated. Participants are encouraged to provide novel ways of determining link quality (self-learning) as part of end to end communication protocol.

DATA: Live speech stream may be generated through laptop mic. And satellite link quality (BER) can be simulated using available tools like MATLAB/C for testing and evaluation.

Dataset: will be uploaded shortly

NM401-ISRO
Blind Scrambling Code Identification:

Satellite Link Monitoring and Analysis System requires identification of transmitted waveform parameters. Scrambler polynomial is one of the parameters. The objective is to identify the scrambler polynomial and initial seed (if any) used for scrambling on the dataset. The scrambling polynomials can be from any of the given standards namely DVB-S, CCSDS, DVBS2, V.35.

Desired Outcome: The challenge is in developing an algorithm that can identify the scrambler polynomial from the scrambled dataset given with the problem. The scrambler used in generating the data can be either self–synchronous or synchronous. It is expected that the output of the developed algorithm should be polynomial estimation of the scrambler used in generating the waveform. Reasonable assumptions for the algorithm design can be taken.

DATA: Matlab generated scrambled dataset

Dataset: will be uploaded shortly

NM402-ISRO
Efficient Header Compression technique for IP based communication over Satellite network:

SATCOM networks operate under bandwidth & power limited scenarios. In order to effectively utilize the network resources, the overheads are required to be reduced (specially TCP/IP Headers) in the user traffic. The IP header compression scheme should accept TCP/IP traffic from users & provide header compressed data over satellite link. The header compression system should support all protocols & provide seamless communication among multiple users by minimizing the TCP/IP headers through elimination/caching static fields.

The compression algorithm should be capable of maximizing Header compression factor for applications like VoIP, FTP, HTTP etc. & handling static/dynamic fields of TCP/IP headers during communication between multiple users.

Desired Outcome: Complete algorithm development for header compression technique, system development and demonstration with different IP based services like – VoIP, FTP etc. Microcontroller based solution will be cheered

DATA: Real Time VoIP calling, FTP between server & client etc.

Dataset: will be uploaded shortly

NM403-ISRO
Customized Web-based animation for data cube, volume rendering and time series data visualization:

Develop software for animating multi-spectral, multi-temporal data (in the form of orthorectified data cube). User should be able to select parameters such as the band to visualize, the time-range or the time-step and animation speed parameters e.t.c.

The software should be web-based and support combining data from various satellite imaging sensors. The software system should be able to handle large volumes of data.

Dataset:
Hyperion_Ahmedabad_1 download
Hyperion_Ahmedabad_2 download
Landsat-8_Ahmedabad download
Sentinel-2_Ahmedabad download

NM404-ISRO
Size Invariant Ship detection from SAR Images:

SAR satellites provide useful information for object detection. A methodology needs to be developed to detect ships at ocean area using SAR data and estimate size of the detected ships (size invariant). The methodology should be applicable for different resolution SAR data of same bands.

Develop software with following two features-

  1. Land water discrimination using SAR imagery,
  2. Output detected ships as a vector file, with an estimate of the size of ship
Dataset:
Kandla download
Karachi download

NM405-ISRO
Processing, visualization and application development of raw GNSS data on Android Smartphones:

These days almost every android smartphone includes a GNSS sensor. Increasingly this GNSS sensor supports hosts of GNSS constellation like GPS, Galileo, Beidou, Glonass and QZSS. In near future it will support NavIC also. Until now, the smartphone GNSS chipsets were operating on single frequency at L1 only. However, some of the advanced smartphone chipsets now supports L5 band also. Along with this a major development has happened wherein the android has enabled access to raw GNSS data also. This opens up enormous opportunity to develop newer smartphone based GNSS applications enabling accuracy and integrity, which was not possible earlier. This problem statement deals with the processing and visualization of such raw GNSS observable from the android devices.

Desired Outcome: teams are free to bring out novel processing strategies and their implementation on android, utilizing the raw GNSS observables. Detection of interference, interference localization, spoofing detection, enhancing accuracy, fusion with other smartphone sensors are the capabilities to name a few.

Dataset:

Raw GNSS observable from a device running latest android release. Google provides data logging and desktop-based analysis tools.


NM406-ISRO
Ethical hacking for extraction of restricted GNSS signal features

The Global navigation Satellites transmit navigation signals for both civilian as well as strategic and military users. All the necessary signal details are available and well documented regarding signals for civilian applications. However, it is not available for the signals for strategic users. All the GNSS service provides be it, GPS, GLONASS, Galileo or Beidou is transmitting such signals for their strategic users. These signals are broadcasted from their respective satellites; however, the necessary details of the signals is not available in public domain. This problem deals with the subject of ethical hacking of strategic GNSS signals. The transmitted signals from GNSS satellites in L1 band (≈1.5 GHz) can be digitized and stored for subsequent processing. Primarily the idea would be to process the recorded GNSS signals in order to extract the important signal feature such as, data rates, data frame lengths, correlation among data frames, spreading code structures, code lengths, overlay codes, spectral signatures etc. The requisite skillset for this activity includes (but not limited to) signal processing, pattern recognition, etc.

Desired Outcome: Algorithms and processing software to extract the signal features from the digitally recorded GNSS signals. Any one test case among GPS M-code signal, Galileo PRS signal or Beiodu PRS signal can be considered.

DATA: The digitized GNSS signal for sufficient duration will be recorded using SAC/ISRO high gain L band facility. This data set alongiwth with the metadata information will be provided to the teams.

Dataset: will be uploaded shortly