AI is often presented as a cure-all solution to video surveillance problems, delivering near 100% accuracy, unlimited functionality, and cost savings too.
But the truth is that AI does have its faults and isn’t always necessary. There are many systems out there that offer perfectly good solutions using video data analytics alone, with far fewer false alarms than early versions. And even the most advanced AI system can’t achieve 100% accuracy. Companies must be wary of wasting money waiting for a perfect AI solution rather than picking a good-enough HD plug-and-play system that would pay for itself within a year.
However, there are two areas where AI does offer benefits traditional analytics can’t match: video search and real-time alerts. AI surveillance can capture metadata like clothing colour in a way general data analytics can’t, meaning an eyewitness report of a suspicious loiterer in a green coat carrying a black backpack can be fed into the video search facility and save security personnel from searching hundreds of hours of footage. And it can make real-time alerts far more accurate, reducing the need for on-site surveillance staff.
While analytics is still in widespread use for basic object detection, like counting the number of people in an area, AI is becoming the go-to for real-time anomaly detection, such as spotting a person going somewhere they shouldn’t late at night or a vehicle driving down a road the wrong way.
However, AI doesn’t provide all the answers and is not infallible, no matter how large a dataset it’s been trained on or for how long. Many organisations would do better to get the basics right, such as making sure there are cameras installed in all the right places and all the cameras are working correctly, before investing in the latest AI technology.