Recent Developments in Spatial Analysis and Computer Vision

| |

Advertisement

Several computing and analytical developments have combined to make real-time spatial analysis using computer vision techniques applicable for a variety of areas, including helping businesses improve their services and the monitoring of crops.

Computer vision offers spatial analysts the possibility to conduct a variety of analyses without using traditional spatial tools.

Spatial Analysis of Retail Space

Recently, Microsoft began using its cloud computing infrastructure, Azure, to launch a spatial analytics service, called Spatial Analysis, that takes video camera data, including multiple images, and allows stores and services to better monitor their customers’ shopping experience. The services provide the ability to quickly count the number of people in a building and the distances between them.

Other analytics can be performed and monitored, including using the Metrics Advisor service that looks for anomalies or variations in data from simply utilizing data logs.  This tool is being pitched as a service that helps companies maintain a healthy environment in their buildings and stores as the monitoring of social distancing and number of occupants can be used to see if safety guidelines are followed in the age of COVID-19.[1]

Microsoft uses computer visions with its Azure product to perform spatial analysis within a retail store to understand the number of people within it and the spacing between them.  Images: Microsoft.
Microsoft uses computer vision with its Azure product to perform spatial analysis within a retail store to understand the number of people within it and the spacing between them. Images: Microsoft.

While major companies such as Microsoft are competing for the consumer market in its use of computer vision and artificial intelligence related to spatial analysis, other areas have also practical development that can aid environmental conservation efforts.

Weed Management and Spatial Analytics

One area of development is better focusing the use of herbicides.

Trilogy Networks and the Rural Cloud Initiative have partnered with South Dakota State University to create a distributed cloud computing service that farmers can directly tap into. The approach uses computer vision techniques on imagery, such as satellite or UAV data, to identify weeds growing within fields.

This allows farmers to then target where weeds are growing to apply their herbicides. This minimizes costs for farmers while also minimizing environmental impact by only using targeted rather than broad herbicide treatment.[2]

Integrating WebAR and Computer Vision Capabilities

Another emerging area is the integration of WebAR (web augmented reality) and computer vision capabilities.

Companies can now customize websites, rather than have dedicated apps that users have to download, to integrate real visual data with augmented reality.

For instance, video or photograph data can be taken, turned into an augmented reality visualization, and then that data can give users both a visual experience and data provided from the images to inform on products or object data within a virtual scene.

Google’s Scene Viewer has become one of the more popular platforms that developers are using to embed 3D models and images used for augmented reality-based searches and product views.[3]

Google's Scene Viewer lets developers embed 3D images into scenes for an AR experience.  Images: Google
Google’s Scene Viewer lets developers embed 3D images into scenes for an AR experience. Images: Google

Edge Computing

One reason why there has been an expanded interest in computer vision in spatial analytics is the development of edge computing, which aims to bring cloud-based system analytics and data closer together to enable faster computation.

Tools such as OpenVINO aim to enable users to have much faster processing of video and other forms of visual data to enable complex computer vision tasks that can auto-detect features and provide information about objects more quickly.

Tasks such as segmentation, object detection, similarity measures between objects, action recognition, depth estimation, and many other tasks are incorporate as models that can be used without having to recreate them. Service tasks could then be accomplished more quickly since they only need training data, where they can even be configured to use Internet of Things (IoT) or other remote technologies that provide data to the service.[4]

We are now seeing the merger of new technologies that enable new sets of analytical and visual experience tools that even a few years ago would not be possible.

For the spatial community, the benefits have been relatively limited but are likely to expand in coming years as the speed and ability to get and deliver data have increased more rapidly with edge-based cloud computing and other related platforms.

Computer vision techniques can now enable the possibility to do many spatial operations without even using GIS tools. This will likely lead to more scientific use of such tools particularly as services become easier to use with new tools becoming increasingly available. 

References

[1]    For more on the new frameworks and tools released by Microsoft, see: https://www.datanami.com/2020/09/22/microsoft-launches-spatial-analytics-other-ai-services-at-ignite/.

[2]    For more on how Trilogy Networks is working with South Dakota State University on the weed detection initiative and service, see:  https://www.prnewswire.com/news-releases/trilogy-networks-joins-grand-farm-to-provide-cloud-based-computing-for-precision-agriculture-301142145.html.

[3]    For more on augmented reality and computer vision techniques to enhance web and search experiences, see: https://next.reality.news/news/googles-arcore-updates-bring-scene-viewer-for-ar-web-search-improvements-image-recognition-ambient-lighting-0197232/.

[4]    For more on related services that can be provided using computer vision libraries and integrated with cloud-based and IoT services, see: Castro-Zunti, R.D., Yépez, J., Ko, S.-B., 2020. License plate segmentation and recognition system using deep learning and OpenVINO. IET Intelligent Transport Systems 14, 119–126. DOI: 10.1049/iet-its.2019.0481

Related


Enter your email to receive the weekly GIS Lounge newsletter: