Applications of Machine Learning in Remote Sensing and Gis

Hello. Welcome to atom Geospatial Solutions. Thanks for spending your time to always read our content. Your contribution in reading our posts has always motivated us to write more.

Kindly like our Facebook page @ atomgeospatialsolutions to join our audience and never to miss out on nice literature.

Today we'll talk about the applications of Artificial intelligence in the geospatial Domain.

Introduction to Artificial Intelligence

Probably you have always heard about AI from internet, people, newspapers, Facebook, YouTube etc.

If you do a Google search about AI, you would get all sorts of complicated images such as shown below.

Image from Google search

Image from Google search

But what is Artificial Intelligence really?

AI is a way of making a computer, robot, or a software 

think in a way an intelligent human would. 

It is accomplished by studying how human brains think, 

learn, decide, and work while trying to solve a problem.

The outcomes of that study are used as a basis of developing 

intelligent software and systems.

It is a multi-disciplinary field that is applied in a variety of areas 
e.g. engineering, health sector, business, computer science etc.

Let's now dive into AI in the Geospatial domain.

AI in the geospatial world

Traditionally, many tasks where done by people manually, but with the Advent of AI such tasks are now automatically done by computer programs. Such applications include:

Satellite Image segmentation.

Image segmentation deals with classification of components of an image into individual classes at a pixel level.

The presence of huge amounts of earth observations data from satellites, drones, aerial survey missions etc. has led to the application of deep neural networks on such data to automatically divide the image into classes such as buildings trees, water bodies, roads, etc.

It involves designing a neural network to segment an image. Once that is done, the network is trained and evaluated using ground truth labels of the image classes. After  reaching a certain level of accuracy, it can be deployed to do the segmentation without human supervision at any time

An example of Aerial image segmentation

Object detection from aerial images.

In computer science, Object detection involves feeding a given image to a computer software, tasking the the software to determine the features present in the image and later to draw bounding boxes around the objects.

In geospatial analysis the above is used to automatically locate and count features from aerial images. However, detection stops at determining the bounding box of an object. Think of counting animals in a game pack using drone imagery.

Swimming pool detection

In the image above, a computer program was designed to automatically pick out swimming pools within a given residential area from a drone image.

Satellite image Feature extraction

This involves extraction of a particular feature from an image. It is different from segmentation in such a way that segmentation deals with classification of all objects into their prospective classes within an image while feature extraction looks at a particular feature e.g buildings and its extraction 

Automatic Building Footprint Extraction

Geospatial data analysis.

At a certain point p in time t, in your education journey as a Geospatial professional you will be required to do a thesis either at undergraduate, masters, PHD.......

To demonstrate this let us assume that your thesis  deals with estimation of temperature at a given location from sentinel 2A imagery.

You will have temperature values from the sentinel satellite and ground truth values from sampled stations on Earth.

Since temperature is continuous, this data set will contain a lot of values and thus will require advanced processing and analysis techniques.

Well thanks to AI,  machine learning algorithms can be designed to learn patterns and dynamics in our data and later model the best the  relationship between sentinel  2A temperature and ground temperature values.

Regression Analysis

Image classification

In computer science , image classification involves giving a computer an image and then tasking it to classification the whole image into a single class.

 It is mainly done to in classification of different items e.g animals, cars, vehicles e.t.c

When we come back to the Geospatial domain such can be used in many applications e.g traffic control in developed countries and busy streets.

Well how is this done?

In developed cities, CCTV cameras are  installed on every road and junction. These take pictures of the traffic situation and send them to control centers at given time periods.

A computer program is designed to classify the traffic in each image received at the control center as either low, high or medium.

Basing on that, the traffic lights at junctions can be controlled and this effectively monitoring traffic in different busy centers

thanks for reading our post Kindly like our page @ atomgeospatialsolutions   to join our Audience