From traffic to transit, deep learning-based surveillance is ready to takeover

From traffic to transit, deep learning-based surveillance is ready to takeover
Automated video surveillance is not just going to be a game changer in the security industry, but will also have a tremendous impact on processes like traffic management undertaken in the future. In fact, with the arrival of machine learning-based analysis, the opportunities for this solution are endless, including monitoring traffic flows and congestion, traffic surveys and incident detection as well as mass transit, military, critical infrastructure and commercial uses in casinos and malls.

Key benefits of this development include increased efficiency, better compliance, reduced costs, ease of auditing and boost of public image, according to Calipsa, a UK-based startup specializing in this field.

The company’s product is a machine learning-based software platform designed to automate analysis, enumeration and reporting of video surveillance footage. The core engine is able to support real-time video surveillance monitoring for traffic enforcement, incidents, public disorder, mass transit, commercial and other applications. Video material is meta-tagged for classification and analysis allowing users to output an audit report and annotated video to support statistical reporting.

“Calipsa is next generation video surveillance technology built on state-of-the-art deep learning models,” said Anthony Fulgoni, Chief Revenue Officer of Calipsa. “The company has developed algorithms that can process and analyze hours of video feeds to provide alerts and detailed reports for applications where real-time video monitoring is key, from traffic enforcement and public spaces to road accidents and public disorder.”

The core engine is able to support real-time video surveillance monitoring for traffic enforcement, incidents, public disorder, mass transit, commercial and other applications.

He added that there are around 250 million video surveillance cameras in operation worldwide today, capturing 1.6 trillion hours of video annually. Yet the majority of video surveillance is still done by humans. This doesn’t make sense, as it’s expensive and inefficient. Manually viewing huge quantities of video data for hours and hours leads to fatigue, loss of attention, and errors at a time when video surveillance has never been more critical.

"The Calipsa engine uses a feedback loop to continuously evolve and improve over time," Fulgoni said. "Human operators can ‘teach’ the artificial intelligence using a simple point and click interface, which automates repetitive parts of their jobs. Deployed via the cloud or on premise, Calipsa is adaptable to all weather and lighting conditions, with 95 percent accuracy. Just as important, Calipsa is camera vendor agnostic, meaning it can operate with pretty much any video surveillance equipment that is already in use.”

Of course, at the center of this product is artificial intelligence (AI). Calipsa’s algorithms improve over time with supervision from human operators. The algorithms are rewarded for correct notifications and penalized for false alarms, potentially improving the solution every day. 
Share to:
Comments ( 0 )