Space Apps Challenge 2017

The issue

Early warning. Information on forest fires in the developing countries is most frequently supplied via the official news feeds after the fact; in case of large fires, notifications during the news can occur. Local fires usually get no coverage at all.

The situation in Western countries is better due to a more advanced regional media system, although it also depends on the scale of the natural disaster. There is no efficient early warning system for forest fires, as the appearance thereof in free access is connected with either the official media or specialized sources (resources of emergency services etc.), which don’t have a wide audience and/or a means of distributing information.

The main goals are:

1. Assistance in localization and scale assessment. Currently, a person who has discovered a fire can only inform the emergency services by calling 103(911) and giving a verbal description of the problem. There is no possibility to convey the magnitude of the fire and inform the people in the immediate vicinity.
2. Projection and evaluation of forest fire trends and patterns. Currently, projections for the spread rate of fires and the probability of the occurrence thereof is done only in case of large-scale disasters and is not carried out until the situation starts to become a serious threat to human life. One of the main reasons is the diversity of data for analysis. Examples of this data include:
   • NASA data;
   • Local weather service data;
   • Local emergency service data;
   • Official news bulletins.

At this time, it is a rare occurrence when more than 1 or 2 sources are used during the formation of projections or forecasts. This is due to the complications of data acquisition and aggregation. But even if forecasts are made, it is most frequently done within the academic field and the results are not made available to the public.

The solution

As a solution to the issues stated above we propose an information network designed to collect and spread information for analysis and warning about forest fires.

Currently the solution consists of the following components:

• The network itself; a blockchain network to store the history of fire notifications
• A NASA analytics module to supply data to the network;
• An FB chat bot fulfilling 2 functions:
  • Fire notifications;
  • Gathering of messages and photographic materials from eyewitnesses;
• A module to analyze the photographic materials obtained by the chat bot;
• A website with visualization of active forest fires.

Tech stack and foundation

The main component of the system is a distributed network based on blockchain technology with the possibility of connecting an unlimited number of new elements, each of which has a verified SSL certificate. The network performs the functions of aggregation, accumulation and exchange of data and notifications.

Why blockchain?

• Distribution – there is no need for centralized storage and a source of notifications. This, in turn, ensures good scalability and stable working capacity;
• Permanence of the notification history within the network, an idea within the idea itself
• Verification of new data is carried out taking into account the encryption results of the previous data. This allows to avoid data loss and tampering. If necessary, it’s possible to locate the source of misinformation, even if measures have been taken to cover one’s tracks.

The network data is not raw material (for instance, satellite photographs, eyewitness photographs etc.), but a result of analytics with references to the source data used for analysis.

The data (currently) has 2 levels of veracity:

• Official news bulletins;
• Unofficial data.

The former have a higher level of credibility than information from eyewitnesses of results of satellite photo analysis, although the latter are posted on the Web significantly faster.

Within the hackathon we created 2 “analytical centers”:

• A NodeJS application to process NASA data and send information on the global state of affairs regarding fires;
• A Java application that carries out filtration of photographs that are posted by the users to the chat bot. Integration with the chat bot is done in order to obtain eyewitness photos, and the Google Vision service in order to detect forest fires via photographs with a certain degree of probability

Why Facebook Chatbot?

The FB chat bot was created for work with eyewitnesses and the target group. Why a chat bot in FB?

Year after year statistics show that fewer users still apply classic web-based clients and more people consume content via social media, where FB is the undisputed leader.

The chat bot presents the following possibilities for promotion:

• A wide audience;
• Users from different countries and continents;
• Familiar management and user experience interfaces;
• Higher degree of trust to an application within a network than a standalone product;
• Additional possibility for official sources to enter Web 2.0;

Fire Explorer

A web-application with an Angular 2.0 basis for visualization of the data on fires by means of an interactive map (Cesium open source framework) provides an interface familiar to users of Google Maps with additional interactivity (3D, 2.5D, 2D).

Next steps

Prospects:

1. The number of information sources, same as the number of consumers on the Web, can be infinite. The more the better, especially as the network starts to include local data providers (eyewitnesses, official services), which can supply the network with local data alongside global forest fire information providers (NASA, for instance);
2. The object can be different -- not just forest fires, but any other natural disaster (tornados or, to turn to a different field, street riots). We didn’t just create a prototype. We have proposed an approach to obtaining and spreading information about phenomena that pose danger to society.
3. There can be various types of consumer applications - websites, mobile applications, forums, other social networks etc.

Adjustments and planned features:

1. A move from Google Cloud to an in-house development based on TensorFlow for a more ”narrow focus” in photo identification;Expansion of the NASA data analytics application’s functionality;
2. Development of an “analytical center” for projection of fire patterns and the probability of fire occurrence based on NASA data;
3. Inclusion of conventional forest fire prediction sources into the network;
4. Develop a cross-platform mobile application with the same functionality as chat bot has and supported offline mode(postponed communication data network).

Enterprise customers in nearest future:

• Unofficial, non-commercial organizations(ecology centers, volunteers etc) and the general population;
• Local official services (emergency services, forestries etc.);
• Commercial organizations(private land monitoring services, insurance companies, timber companies etc);

• Fire Alarm System (FIRMS) Github page
• Fire Image Detector Github page
• Fire Explorer Web Application Github page
• Facebook ChatBot Github page
• Facebook ChatBot Adapter Github page

Data used

• Fire Information for Resource Management System (FIRMS)

Frameworks and APIs

• Bing Maps API And DataSets
• OpenStreetMap API And DataSets
• ESRI World Imagery
• Cesium.js