Machine Learning in GIS and Remote Sensing
Machine Learning (ML) is a field of Artificial Intelligence that uses statistical techniques to give computer systems the ability to learn from data, without being explicitly programmed. It has the special ability to progressively improve its accuracy and efficiency throughtout the course of its training. Machine Learning algorithms specialise in improving their performance using training data.
Common terminology involved in Machine Learning (GIS)
Labeled data: Data consisting of a set of training examples, where each example is a pair consisting of an input and a desired output value ( contains both attributes and its labels ).
Classification: The goal of many ML projects is to predict discrete values, for example, True or False, 1 or 0, Forest or Not Forest and so on.
Regression: The goal of this type of ML projects is to predict continuous values, for example, percentage of land cover
There are three types of Machine Learning:
Supervised Learning: This type of learning uses training data which is labeled to produce an output from the previous experince. It helps us to solve various real-world problems. It is typically done in the context of classification.
Nearest neighbours, Logistic Regression, Decision Trees, Linear Regression are all examples of Supervised Learning.
Unsupervised Learning: This type of learning involves training data without labels and is used in pattern detection and descriptive modelling. It solves various types of real-world problems and is primarily used for mining rules and detecting patterns.
K Means Clustering, ISOData, Association Rules are all examples of Unsupervised Learning.
Downloading Satellite images in QGIS
There are a couple of ways in which satellite images can be downloaded and used in QGIS:
Semiautomatic Classification Plugin: After opening QGIS, click on the first icon of the toolbar that is shown below.
After clicking on the first icon, navigate to Download Products and click on Login Data. There are three websites mentioned in the tabs below which can be used to access data : ‘https://ers.cr.usgs.gov/’, ‘https://urs.earthdata.nasa.gov/’, ‘https://scihub.copernicus.eu/apihub’. Go to all the above websites and create your account to get your account credentials. Then paste them in the Login Data section in the QGIS.
After filling in the details, navigate to the Search tab beside Login Data. Enter the coordinates for which you want to get the satellite imagery. Refer to the below image to get an idea on how to fill the Search tab.
EO Browser: Go to the website ‘https://apps.sentinel-hub.com/eo-browser/’. Select “Sentinel-2” and turn on Advanced Search and change maximum cloud coverage to 0%. Change the timeframe for which you want to narrow down the search and Search. Multiple maps may be displayed if any maps are found. When clicked on any map, different types of maps such as Moisture Index, True Colour, False Colours etc can be viewed. These maps can be useful as we can export them and import them into our QGIS software.
Regression Analysis in GIS
Geographic Information Systems (GIS) is a comprehensive set of statistical techniques that enable the estimation of relationships between a dependent variable, often referred to as the outcome variable, and one or more independent variables, also known as predictors, covariates, or features. By employing regression analysis in GIS, researchers and practitioners gain valuable insights into the intricate correlations between these variables, facilitating the identification of patterns and the prediction of unknown values.
The primary objective of regression analysis in GIS is to elucidate the underlying relationships between the dependent and independent variables, allowing for a deeper understanding of the key factors that influence the phenomenon under investigation. Through this analytical approach, GIS professionals can uncover meaningful associations, discern trends, and make informed predictions based on the available data.
One of the key advantages of using regression analysis in GIS is its ability to explore complex spatial relationships. By incorporating geographic information into the analysis, researchers can capture the spatial variability of the variables involved, providing a more comprehensive understanding of the phenomenon being studied. This spatial perspective allows for the detection of spatial autocorrelation, where nearby locations exhibit similar values, and the identification of spatial patterns that may not be apparent through traditional statistical methods.
Moreover, regression analysis in GIS enables the creation of predictive models, which can be utilized to estimate unknown values based on the relationships discovered during the analysis. These models can prove invaluable in various applications, such as urban planning, environmental management, and resource allocation. By leveraging the power of machine learning algorithms, GIS professionals can develop accurate and reliable predictive models that assist in decision-making processes.
Conclusion
In conclusion, Machine Learning encompasses both supervised and unsupervised learning techniques, with applications ranging from classification to regression. The integration of satellite imagery into GIS, facilitated by tools like the Semiautomatic Classification Plugin and EO Browser, opens up new possibilities for geospatial analysis. Additionally, the utilization of regression analysis in GIS amplifies our understanding of complex spatial relationships, enabling predictive modeling and informed decision-making across various domains. As technology continues to advance, the synergy between ML and GIS promises to drive innovation and solve real-world problems more effectively.
