GeoAI

GeoAI is the integration of artificial intelligence (AI) with spatial data, science, and geospatial technology to increase understanding and solve spatial problems. GeoAI includes the application of traditional AI techniques to generate spatial data through the extraction, classification, and detection of information from structured and unstructured data. GeoAI is also the use of spatially explicit AI techniques that are designed to solve spatial problems through the analysis of spatial data, and includes techniques for detecting patterns, making predictions, spatiotemporal forecasting, and more.

Key concepts

The following terms will help you understand GeoAI within ArcGIS Pro.

• Spatial analysis—The process of examining the locations, attributes, and relationships in spatial data through a series of techniques from simple overlays to advanced spatial statistics and other analytical techniques.
• Artificial Intelligence (AI)—The ability of a machine (for example, computer) to perform tasks that traditionally require human intelligence, such as perception, reasoning, and learning. We see AI applications everywhere in our daily lives—in smart assistants on our phones, in recommendations on our social media feeds, and in self-driving cars and robots. AI encompasses both machine learning and deep learning.
• Machine learning (ML)—A subset of AI referring to a set of techniques that allow computers to learn patterns within data and acquire knowledge without being explicitly programmed. ML techniques typically come in the form of statistical methods or data-driven algorithms that solve classification, clustering, and prediction (for example, regression/forecasting) problems. Think of ML as an approach to achieve AI.
• Deep learning (DL)—A subset of ML that uses trainable and learned algorithms in the form of artificial neural networks. The multilayered architecture of these networks is inspired by how the human brain operates—humans learn about and understand the world around them as a nested hierarchy of concepts. You can think of DL algorithms as functioning like the human brain, in which the computer learns complex patterns and concepts by piecing together simpler concepts. Raw input data is analyzed through different layers of the network, with each successive layer learning and capturing the definitions of more complex and specific features and patterns in the data.

GeoAI is embedded throughout ArcGIS across a wide variety of geoprocessing and exploratory analysis tools. Machine learning algorithms in ArcGIS are used in the analysis of spatial data to perform clustering, prediction (classification and regression), and spatiotemporal forecasting. Deep learning is used in ArcGIS for the generation of geospatial information from sensor data (including imagery and point clouds) using techniques and tools for pixel classification and image segmentation, detecting objects and extracting features, object tracking, change detection, and image simulation. Deep learning is also used to generate geospatial data from unstructured text using a variety of natural language processing (NLP) techniques. Deep learning can also be used for the analysis of spatial data to make predictions and forecasts. Many of our most challenging problems, however, require bringing together GeoAI and other powerful spatial analysis techniques to both understand and effectively address these challenges.

Problem-solving

GeoAI can play a critical role in spatial problem-solving in a wide range of application areas.

An important aspect of GeoAI is the application of traditional AI techniques in the generation of spatial data through extraction, classification, and detection of information from structured and unstructured data. This data includes tabular data, remotely sensed data (including raster, imagery, lidar point clouds, video, and more), and even text data. This spatial data generation includes applications such as finding and cataloging objects in imagery, creating 3D data from lidar, or extracting location information from unstructured text for subsequent geocoding. ArcGIS also includes a set of pretrained deep learning models that reduce some of the most time-consuming and resource-intensive aspects of the training process. The use of deep learning to automate these previously tedious processes of spatial data extraction and creation is valuable in many workflows and can lead to significant time and resource savings. This spatial data can also become valuable input to downstream workflows, for everything from spatial data management to advanced spatial analysis of patterns and relationships.

See an example of deep learning used to automate feature extraction from imagery

The other key aspect of GeoAI is the application of machine learning and deep learning techniques, including spatially explicit statistical and machine learning techniques, to the analysis of spatial data for applications such as detecting spatial patterns and making predictions and spatiotemporal forecasts. The use of emerging machine learning and deep learning tools with spatial data gives practitioners new alternatives to explore difficult problem spaces. The use of machine learning methods on spatial data, as well as the incorporation of spatially explicit models that incorporate some aspect of geography (location, shape, proximity, and more) directly into the algorithm can not only make models more efficient, but also often more accurate and representative of the reality that we are aiming to model. These techniques can be used to allocate resources based on meaningful spatial patterns, to find trends and anomalies in space and time, and to incorporate spatial relationships into predictions and forecasts.

See an example of machine learning methods applied to spatial data

Ultimately, GeoAI concepts are used in tools that work best in the hands of thoughtful analysts and data scientists. Like other tools, the analyst must have a thorough understanding of the problem, a drive to iterate and refine an analysis, and transparency in the process when providing the information product to the decision maker. The use of GeoAI does not change these principles, and in fact makes the need for conscientious analysis more important than ever. ArcGIS Pro provides a powerful experience to run GeoAI tools, evaluate their results, and communicate with stakeholders effectively and responsibly.

Learn more

See the following resources for more information.

• Spatial data generation and information extraction
  • Introduction to Deep Learning
  • Deep learning in ArcGIS Pro
  • ArcGIS pretrained models
  • Classification and Pattern Recognition toolset (Image Analyst toolbox)
  • Deep Learning toolset (Image Analyst toolbox)
  • Point Cloud toolset (3D Analyst toolbox)
  • Text Analysis toolset (GeoAI toolbox)
  • Interactive object detection basics
• Spatial problem-solving
  • Clustering and pattern detection
    • Mapping Clusters toolset (Spatial Statistics toolbox)
    • Space Time Pattern Analysis toolset (Space Time Pattern Mining toolbox)
    • Multidimensional Analysis toolset (Image Analyst toolbox)
  • Prediction
    • Modeling Spatial Relationships toolset (Spatial Statistics toolbox)
    • Feature and Tabular Analysis toolset (GeoAI toolbox)
    • Multidimensional Analysis toolset (Image Analyst toolbox)
  • Forecasting
    • Time Series Forecasting toolset (Space Time Pattern Mining toolbox)
    • Time Series AI toolset (GeoAI toolbox)
    • Multidimensional Analysis toolset (Image Analyst toolbox)