Hybridization of data is a key strategy for increasing detection confidence and reducing error. Each data type has given expectations for ambiguity based on its source. When these source-specific qualities are considered through weighted hybridization, outputs are improved, says Chantz Thomas, Solutions Engineering Manager, BlackSky, in an exclusive interview with Geospatial World.
BlackSky fuses satellite imagery with real-time data which provides highly accurate insights to users in multiple fields. How did you come up with the idea of merging the two and any tangible difference in insights that you have observed?
Data fusion is the key to maximizing the accuracy and detail of detections. This applies to many existing non-satellite scenarios like market analysis, risk analysis, and weather predictions. BlackSky is combining its unique systems of automated, real-time data processing and automated, real-time image collection to enhance the quality and value of detections for customers. Hybridization of sources includes:
- The use of BlackSky’s rapid revisit of images to characterize behavior seen in occasional mapping data (i.e., either of extremely high or low resolution – think occasional updates via Google Earth or Landsat).
- The use of news or social media data to enrich the view of a ground situation monitored by BlackSky satellites.
- The use of synthetic aperture radar (SAR) sensors to image areas when clouds prevent visible imaging.
- The use of an earthquake alert or violence alert to trigger BlackSky images to better identify the extent of damage or risk.
BlackSky gathers data from a wide array of sources and then stores and processes it. Is the process tedious and do you think the use of AI-based programs and Machine Learning predictive algorithms will completely overhaul data processing?
AI advancements are at the core of BlackSky’s systems for open-source event detection and image analytics. While the use of AI does necessarily reduce data volume or processing time, it is better suited for handling complex, ambiguous, and changing content. With BlackSky’s event detection service, we can evaluate and make known significant events via machine learning and estimate the event’s location. AI gives us the ability to seek events thematically (i.e., conflict, disaster, disease, etc.) and more readily handle the complexity and ambiguity of written language. These elements, along with varying seasons and lighting conditions make AI a critical solution to capture the subtle signatures of objects and phenomena that may not be evident to a human eye or reproducible over time.
High-resolution images provided by BlackSky can be used for myriad purposes like tracking illegal activities, mitigating disasters or facilitating business. Since you have a repository of millions of images, how do the users search the most relevant image from your catalog?
BlackSky customers are able to search by image parameters including location, capture time, cloud cover, and resolution (which depends on satellite angle when imaging). While catalog data will grow increasingly interesting as we develop a rich time series of information for a location, we perceive that customers will continue to focus heavily on the most defining aspects of our system, which include rapid revisit, low latency, and imaging time diversity through real-time tasking of satellites.
What do you think would be the impact of Blockchain and IoT on data storing, processing and analysis?
BlackSky has consistently prioritized the privacy of customer data. We envision blockchain and other emerging technologies will become important elements in verifying the chain of custody for images, insights, and other outputs.
IoT is already influencing BlackSky detections through systems that measure radio frequency emissions like AIS beacons and communication transmissions. An enormous number of ground and space-based IoT systems are emerging, which will certainly impact our open-source detections and possibly even make their way into satellite operations and data distribution.
While gathering data from multiple streams and sources, how do you ensure that there is very little ambiguity and nullify any margin of error or discrepancy?
Hybridization of data is a key strategy for increasing detection confidence and reducing error. Each data type has given expectations for ambiguity based on its source (space-based sensor detection, written report, ground photo, etc.). When these source-specific qualities are considered through weighted hybridization, outputs are improved. No detections or images from any company will ever be free of error or discrepancy, whether the focus is on interpretations of sentences in report or small objects in an image, but characterizing confidence allows users to properly judge BlackSky outputs and act accordingly.
BlackSky intends to operate a constellation of 60 satellites by the end of 2019. How many satellites are there currently in orbit and has this led to a drastic change in the quality of data accumulated and ensured cost-effectiveness as well?
BlackSky is planning to have eight satellites in orbit by the end of 2019, 16 satellites by the end of 2020, and will begin deploying its third-generation satellites in 2021. By the end of 2019, BlackSky will have the world’s only taskable mid-latitude constellation. By focusing on mid-latitude (unlike the polar orbits used by mapping missions), BlackSky enhances its monitoring capabilities by increasing revisit rate; enhancing imaging time diversity, which can image from sun-up to sun-down; and lowering delivery latency. BlackSky is currently operating three satellites, and a fourth satellite will launch later in the summer.
What are the different resolution images that BlackSky offers?
BlackSky focuses on maximizing the temporal resolution of its imagery in order to allow users to access a previously un-addressable and unaffordable dimension of ground activity. Typical temporal resolution for leading providers varies from one to three images per day and is heavily restricted by satellite maneuvering ability and cost. BlackSky will exceed this average temporal resolution and become the highest temporal resolution system available by the end of 2019. By the end of 2020, BlackSky’s added satellites will enable unprecedented temporal resolution for real-time monitoring of situations around the globe.
Can you name some of the major organizations that use BlackSky feeds and data?
A wide range of commercial and government organizations including US and foreign defense and intelligence, transportation, financial, energy sectors rely on BlackSky’s timely Geospatial information.
Do you think in the era of automation, Earth Observation and data analysis industry too would witness a disruption, and what would be its impact on the industry players?
Automation is fundamental to the new wave of space enterprise. BlackSky’s satellite and open-source ground detection systems are entirely automated allowing us to reduce latency and cost for customers. In a broader view, automation of image processing via traditional algorithms and machine learning algorithms has rapidly augmented the relationships of image producers, value-added resellers, and customers.
At each level, the proportion of work performed toward image processing is fluctuating as business models and uses for visible, IR, RF, and SAR data change. Each segment showcases the right balance of risk and reward. This can be seen when major satellite operators purchase analytics firms, satellite-independent algorithm specialists commit to new capital, and governments wade into outsourcing imaging and processing through numerous and diverse commercial contracts.
How do you ensure that rising customer expectations, which often outpace technological advancements, are fulfilled?
Our focus on providing affordable rapid revisit imaging with low latency is opening opportunities to solve problems customers never expected to answer with a space system. In that sense, we’ve been able to jump ahead of traditional customer expectations.
Traditionally, organizations have been unable to commit to the cost and volume required by existing providers to task satellites to take current imagery. Now, they are pleased to have a low barrier means to access remote sensing for business cases, humanitarian work, and other applications.