How might ML-powered video surveillance improve security?

How might ML-powered video surveillance improve security? ...

The increased use of surveillance cameras, whether in service of public safety, health monitoring, or commercial use, has raised concerns about privacy. Today, it seems people''s movements will be captured on CCTV cameras regardless of where they go.

According to the Bureau of Labor Statistics, the number of surveillance cameras in the United States increased from 47 million to 85 million from 2015 to 2021, up 80%, putting the total number of cameras in use at around one million in the country. Globally, the number of surveillance cameras in use was expected to increase at an annual rate of 10 percent through 2026.

The increased scope of these systems has raised concerns about privacy breaches, particularly when it comes to facial recognition in China. Besides the limitations of privacy such as the widespread use of facial recognition, MIT and Stanford University, as well as other institutions, have revealed strong biases in facial recognition systems.

San Francisco had banned facial recognition in local agencies surveillance cameras in January, and at least a dozen other states have instituted facial recognition restrictions for one or another. More surveillance does not necessarily require reduced privacy.

Transform 2022

In person on July 19 and almost from July 20-28, join us at a leading conference on applied AI for enterprise business and technology decision makers.

Improvements in machine learning (ML) technology can both enhance the efficiency of retrieving data from surveillance cameras, while also increasing the protection of those who appear in those feeds locally. A smart camera can, for example, perform processing locally, eliminating the need to disclose and store data. Besides, a smart camera can also help prevent both intentional and unintentional data misuse.

How deep learning protects privacy

surveillance cameras have become more popular as a result of their high-resolution cameras, improved local computing capacity, and increased Internet connections. In some systems, the use of machine learning and artificial intelligence (AI) have increased the capability to search hundreds or thousands of hours of footage.

While increasing video surveillance systems to be more powerful and potentially intrusive, ML and AI can also be used to protect privacy. A subset of AI can be trained to focus on what it should be watching and effectively look away from what it should not.

Deep learning, which is adapted to mimic the functions of the human brain by using a three-dimensional layer, can on its own identify and classify objects and patterns. By using tagged data to train the system, a machine can become more proficient as it becomes exposed to additional data over time. Significantly, it can do this with a small footprint that allows for embedded, localized processing to effectively manage data privacy.

Using deep learning software, a CCTV system can classify people approaching a building entrance (like an office, a theater, or other), permit or refuse entry, and then dispose of any captured information. In another example, a camera monitoring a commercial parking lot might also have a visual view onto the window of a property. The software is also capable of rectifying for any possible hazards caused by the positioning of the camera, and thus avoids accidental errors or intentional activity involving recording images not on the property.

ML makes data actionable

While not only is monitoring or retrieving information from video recordings tedious, it''s also beneficial to watch and treat situations like safety hazards and privacy dangers. ML video content analysis software with deep learning can easily extract, classify, and quickly index targeted objects such as humans or vehicles, making video feeds significantly more searchable, actionable and quantifiable.

Intelligent alerts when certain objects, behaviors, or anomalous activity are detected in objects, such as count-based alerts when the number of people in a specific area exceeds a set limit, alerts triggered by object identification or, if applicable, face recognition.

Other video content analysis tools from live or archived feeds are aggregated, thus giving analysts the ability to comprehend trends and develop protocols for improving safety, operations, and security. And by using proper deep learning techniques, it can do it without posing any limitations to privacy.

While preserving data privacy, video surveillance is enhanced.

Despite concerns about privacy and attempts to deactivate facial recognition, the amount of video and other information being collected is unlikely to slow down. Video systems can, for example, assist health officials in locating people who wear masks or are observing safe-distancing practices. Businesses can monitor peoples shopping habits. The security of public spaces is increasingly dependent on effective video surveillance.

Besides those methods, the spread of home systems with surveillance capabilities has fueled fears about compromised privacy. More than 128 million cloud-connected voice assistants such as Google Home, Amazon Echo, and Facebook Portal are currently being used in homes in the United States, with the capability to record and share information. 76% of TV households say they have smart TVs, which have raised concerns about their potential to spy on users.

However, the way video is collected, processed, and searched can achieve goals of improved security, improved operations, or improved privacy. The current approach of using cloud-connected surveillance cameras with cloud-based analytics does not support privacy and bias concerns. However, ML software with deep learning capabilities allows for localized, embedded intelligence and analytics to deliver high performance at low power that can improve safety while managing data privacy. In most CCTV video surveillance systems, intelligent video technology can also be seamlessly integrated with most existing systems.

Deep learning techniques may help develop future capabilities, thus empowering organizations to continue to enhance their systems by using additional AI capabilities.

Sima AI''s vice president is David Gamba.

You may also like: