Many video analytic customers struggle with false alarms and inaccurate data, says Oyla.
After a few months of using them, most video analytics customers turn off analytics, claims Oyla, a relatively new entrant in the security industry. This is because analytics solutions throw up too many false alarms and cannot function in various harsh conditions.
The San Carlos, California-based company says it offers a Lidar-integrated 3D-aware surveillance system that can collect better data to make analytics systems more accurate. The idea is simple - while the AI algorithms may be powerful, they would work better with more in-depth data.
This is a relatively fresh approach in the surveillance industry, which is leaning more towards investments in the algorithms themselves. Some professionals even feel that it is cheaper to add video-enhancing software than install a high-resolution camera. And while this may be true, nothing can replace accurate data.
Also read: why video analytics may not need high-res cameras
The problem with analytics
Speaking to asmag.com recently, Srinath Kalluri, CEO of
Oyla, pointed out that AI-based analytics are increasingly applied to surveillance video streams to automatically flag unusual activity in real-time. But, when existing surveillance solutions cannot accurately gather sufficient data, the risks posed can progressively become detrimental to the entire security operation.
This could lead to a high rate of false alarms, wasted time and resources on non-existent threats, alarm fatigue resulting in response latency, difficulty in real-time tracking, poor quality in extreme weather or bad lighting, and overall inconsistencies.
“What we’re trying to do is fundamentally different,” Kalluri said. “By obtaining better data as input to the AI engine in the first place, all video surveillance capabilities improve in accuracy and reliability. To this end, we’ve developed an AI-based video surveillance system that augments video with a 3D sensor and proprietary coding.”
3D information is collected across the video field of view using the remote sensing method LiDAR, which is integrated with the video sensor at a hardware level. This approach elevates the level of accuracy users can expect and ensures that the stakeholder has multiple visual layers available to ensure what is happening.
Why don’t analytics work well in harsh conditions?
Thanks to AI and video analytics,
video surveillance systems are becoming smarter – they can automatically flag unusual activities in real-time. However, while great strides have been taken to improve the reliability and accuracy of AI and video analytics, there are still problems in delivering in the real world.
“As the solutions have become more accurate, user trust has gone up, but they are still wary,” Kalluri said. “They have been promised a lot from AI for many, many years, and it has kind of fallen flat on its face in the past. But over the last year and a half, it has now started to get trusted as a viable and valuable solution. In the past, we saw most challenges in extreme weather environments — analytics would identify several false alarms based on environmental conditions, such as trees swaying in the wind.”
Conclusion
Video analytics is considered one of the major factors that could drive the growth of video surveillance systems in the coming years. To this end, there are several companies making significant advancements in this area. However, such expectations may not become reality if the accuracy levels are low.
Solutions like Oyla’s that enhance data collection would definitely add more value to the industry. However, their market penetration might depend on customer awareness. Many customers still consider security as an afterthought. Only after an incident do they even realize if the camera still functions. Hence more customer education is necessary for mutual benefit.