AI Upscaling, Automated Monitoring Emerge as Solutions for Aging CCTV Systems

CCTV-based security and surveillance technology is evolving into intelligent systems capable of providing more accurate imaging and real-time response through integration with artificial intelligence. Notable approaches include improving image quality without replacing aging CCTV cameras and automatically detecting abnormal situations by recognizing human movements to send immediate emergency alerts.

According to the IT industry on January 13, the government and companies are continuing efforts to improve CCTV footage analysis accuracy using AI-based upscaling technology that enhances image quality. The importance of AI-based video analysis technology is being increasingly highlighted in evidence-gathering processes such as tracking suspects' movements and identifying physical descriptions after crimes occur.

The global AI-based video surveillance market is expected to grow rapidly in line with this trend. According to market research firm Global Market Insights, the market size is projected to expand from $5.5 billion (approximately 8 trillion won) in 2023 to $19.5 billion (29 trillion won) by 2032.

However, image quality enhancement models based on generative adversarial networks (GAN), which are widely used for video and image generation, have limitations. There is a high possibility of errors that generate non-existent image and video information. In May last year, a Washington State appeals court ruled that smartphone footage upscaled with generative AI could not be admitted as evidence in criminal trials, with the court acknowledging the possibility of unexpected errors and hallucinations.

For this reason, Korean company Four by Four's (389140.KQ) upscaling technology is attracting increased attention. The company's self-developed image quality enhancement AI solution "PIXELL" is built on convolutional neural network (CNN) technology, prioritizing factual accuracy in its image enhancement processing.

"Image enhancement models using CNN technology are more difficult to develop compared to GAN structures, but they are drawing attention as virtually the only model usable in criminal investigations and medical diagnostics where maintaining factual accuracy is the top priority," a Four by Four official explained.

AI is also playing a role in addressing the shortage of CCTV monitoring personnel. CCTV operators monitor numerous video screens in real time to detect signs of crimes or accidents and quickly relay information to relevant authorities such as police and fire departments.

A representative AI technology improving such operational efficiency is "Dejaview," developed by the Electronics and Telecommunications Research Institute (ETRI). Dejaview is based on the tendency of crimes to repeat in patterns similar to past cases in terms of type, method, location, and time of day. It uses AI to compare and analyze past crime situations with current CCTV footage data to predict the likelihood and risk level of crimes occurring.

"The number of CCTVs installed in Korea, including both public and private, is estimated at more than 20 million, and the cost of replacing aging CCTVs among them would be substantial," an industry official said. "If AI technology can improve image quality when needed and predict the possibility of crimes occurring, it could increase CCTV utilization while reducing personnel and cost burdens."