Recently, there was brief news about Microsoft Flight Simulator and a tower more than 200 stories tall – created by a typo. As funny as that was, it missed the larger picture. Google Earth started a trend that has continued, and the virtualization of the world has proceeded at a rapid pace. It is now to the point where real business benefit is being gained by such work, supporting the application of artificial intelligence (AI) to even more problems.
There have been smaller discussions about virtualization and augmented reality, to analysis and improve performance in stores and other smaller spaces. However, similar to how a drive for better gaming led to NVIDIA’s GPUs, which helped advance AI, the business of capturing a global image base to improve gaming can now help AI lend its skills to new areas.
What’s interesting is that the volume of geographic imaging is beginning to provide analytics to a wide range of businesses. Both businesses and governments are beginning to use the images to estimate forest conditions, crop yields, and other large scale issues. In another interesting application, analysis of buildings and other large structures is beginning to yield ROI on inspections.
One example is inspection of a type of structure called a floating oil tank. It is, as the name implies, an oil storage tank. What’s interesting is the roof floats on top of the oil, raising and lowering based on oil level. The Blackshark.ai system, which includes 200 GPUs, works with satellite imagery, the time stamp of the image, and the shadows provided by the facilities. It is then simple trigonometry to provide an estimate of oil volume. Note, this is something that is good for government oil reserve estimates and for insurance, but companies would want more detailed information.
In that example, the AI is in the computer vision component, it isn’t required for the estimate creation. However, there are examples where AI can be used for additional analysis. Imagine a government trying to estimate energy usage or tax base depending on building type. A satellite image can be analyzed by an AI system which can identify building types by what is on the roof. The size of HVAC systems, for instance, can help to identify a building’s size and use type.
“The image is the starting point for the analysis,” said Michael Putz, CEO and Co-founder, Blackshark.ai. “Semantic reconstruction is the process of adding semantic information needed for critical decisions by companies, governments, and individuals.” Past computer vision systems have only enhanced images, leaving it to people to clarify items. Artificial intelligence can do the work of identifying objects, adding the semantics necessary to speed analysis and enhance the accuracy of decision making.
In the aftermath of events such as earthquakes, floods, and other natural disaster, comparisons to previous images can quickly prepare both governments, NGOs, and insurance companies in taking both faster and more effective action.
Rendering 2D images into 3D simulations also provide other business benefits. Consider wireless signal propagation. 3G and 5G have different broadcast features. Simulating geospatial features can aid coverage range and engineering cost analysis for optimal ROI for tower placement.
Notice the “individuals” mentioned Michael Putz. Think about semantic analysis and someone’s back yard. As AI is able to identify objects and even render 2D satellite images into 3D representations, that enhances the ability for homeowners and small businesses to work together to combine AI and VR to plan for changes. The example Mr. Putz provided was adding a pool to a yard. Being able to visualize that in 3D could help owners check line of site and see if other work, such as higher fences for privacy, might be needed.
At this point, I see the technology being focused on the higher end solutions, such as those for large companies and for government agencies. As with all new product arenas, advances will drive price down and the Cloud model will mean consumer applications will become profitable – just not yet.
Geospatial image capture started off small, but has now grown to a massive scale. The addition of AI both improves computer vision and downstream analysis. This is another are where the world around us is being enhanced by artificial intelligence.