AIoT, short for AI and IoT, has become a popular term. In this regard, Taiwan manufacturers have much to offer. In this article, we look at some of the eye-catching MIT AIoT solutions that were on display at Secutech 2023.
AIoT, short for AI and IoT, has become a popular term. In this regard, Taiwan manufacturers have much to offer, helping users overcome various challenges in security and non-security applications. In this article, we look at some of the eye-catching MIT
AIoT solutions that were on display at Secutech 2023.
Smart search
Needless to say,
smart search can work wonders in post-event investigation. Combining AI with smart search can help related authorities identify individuals in question in matter of minutes. In this regard, VIVOTEK demonstrated its third-generation central management system, called VSS, which comes with a Deep Search function that allows users to do post-event search. They just need to enter the search criteria; no pre-configuration is needed.
“Our cameras have AI edge processing capabilities and extract metadata values – for example humans, vehicles, color of clothes, so on and so forth. Should something unfortunate happen, for example a bomb explosion, and the eyewitnesses cite a suspicious individual wearing a yellow shirt and black hat, the police can just enter the criteria and all video with the individual in question will pop up in a matter of seconds. If a face is identified, it can further be used to match against all faces in the video pile,” said P. J. Tsai, Project Manager with VIVOTEK.
“In fact,” he added, “our slogan is ‘turn hours into seconds.’ In the past officers need to check video clips one by one, an endeavor that takes hours, not to mention the manpower involved. With our solution, this can be shortened into minutes.”
In a similar vein, AIRA exhibited their facial recognition solution that can also help users in security and non-security applications. For the latter, retrieving missing children can be an example. “Say you have a child who has gone missing in a park. We can retrieve the face of that kid from the camera at the entrance and match it against all faces in the video, and see what might have happened,” said F. D. Wei, Manager with AIRA.
What makes their solution unique is that their facial recognition is so optimized, it can be run on systems without GPUs. “We can run on an Intel NUC with a Core i7 CPU supporting four IP cameras, processing 10,000 faces per second. Of course you can also run it on a system with GPU. It depends on your use case and application,” Wei said.
IoT in cold chain
Needless to say, IoT is a key element in AIoT. Among the myriads of IoT applications in the everyday life,
cold chain is a sector that can benefit from IoT particularly well. In this regard, Kiwi Technology demonstrated their end-to-end solution consisted of sensors, routers and cloud processing, all of them made or owned by Kiwi. The solution has achieved success in various markets, including Japan, where retail giant AEON is using their solution.
“We’ve deployed some 150 to 200 temperature sensors in refrigerators across AEON supermarkets. They transfer data to the router, which relays the data to our cloud for further analysis. In the event of malfunctioning, for example a sudden surge in temperature, the system will alert the managers immediately. This is important because if frozen or refrigerated food is not well preserved, food poisoning may occur, and no one wants to see that happen,” said Maverick Lee, VP of Kiwi.
Kiwi currently focuses on two major applications, retail and medical logistics. Both can benefit from cold chain especially in the post-pandemic era where vaccines are in strong demand. “Medical ingredients are very sensitive, and special attention must be paid to it. There was an incident in Taiwan where a batch of vaccines (not related to COVID) had gone bad when delivered to the hospital. At which point did things go wrong, nobody knew. But with cold chain, each step in logistics is monitored and recorded, thus preventing this from happening,” Lee said.
AI and analog cameras
Popular notion dictates that AI works better with IP cameras, which offer higher resolution and image quality. Advanced Video System demystified that notion as they showcased their analog cameras with AI processing capabilities on the edge – that is, on the camera.
“The core of the camera is an AI chip that we developed, capitalizing on our 15 years of experience in AI chips and algorithms, as well as systems and modules for imaging applications,” said Steven Shen, CEO. “Our AI can detect humans, cars, as well as other objects that the AI can do training on. We target all applications as long as there’s DVR that works with camera.”
When asked about the prospects of analog cameras, Shen expressed optimism. “Analog cameras have advantages that IP cameras don’t have. IP cameras are limited by distance, and resolution and frame rate may drop when bandwidth is not sufficient,” he said. “For analog, fps is 30 for NTSC and 25 for PAL, and these will stay constant. High-resolution analog cameras have also become common, with 2MP to 5MP models available.”