This page was machine-translated and may differ from the original. View original

From analysis to judgment...'Intelligent video security' handles it

Google 우선 소스Published2018.03.19 17:16

Intelligent video security that analyzes video and judges situations with its own algorithm comes to the fore

CCTVs are everywhere, whether inside or outside buildings. They are installed in various locations to monitor and track crimes, incidents, and accidents. But then, does CCTV really have a crime prevention effect?

There is no research on the correlation between proactive crime prevention effectiveness and CCTV. However, 'after an incident occurs,' video information is often provided for investigations of violent crimes, assault cases, and the like. The reason CCTV finds it difficult to prevent crime is that real-time monitoring by humans has limitations.

Due to this, interest in 'intelligent video security services' is growing. CCTV plays a sensor role in image collection, while intelligent video security performs video analysis and situation judgment through its own algorithms.

At World Security Expo (SECON) 2018, many companies showcased intelligent video security services.

Distinguishing repetitive movements and monitoring 100+ channels

SK InfoSec exhibited 'SECUDIUM IoT,' an integrated digital security task platform. SECUDIUM IoT collects and analyzes all digital information generated from information security as well as physical security and OT (ICS/SCADA) security systems on a single platform. It can be used in various industrial sectors such as railways and ports, which are critical national infrastructure, smart factories, and smart buildings.

The physical security areas of SECUDIUM IoT, such as video surveillance and access control, were handled jointly with Innodep. Innodep's smart AI monitoring NVR was incorporated. Existing solutions were managed with constant monitoring. As a result, it was difficult to manage 64 or more channels. The smart AI monitoring solution is a solution that reads vector information without decoding to indicate the location and direction of movement. It recognizes objects and reads only movements that are not repetitive or simple movements.

An Innodep official stated, "We are continuously supplementing exception handling for accurate object motion recognition. For example, we remove parts such as rustling leaves or reflections from a parked vehicle's dashcam that move in place without a vector value. As a result, monitoring up to 100-150 channels is possible on an i7-class computer," he explained.

He added, "The existing exception handling was done by detecting motion built into the camera and removing some parts, but now we are expanding the exception handling patterns. Also, it is possible to collect data by monitoring center."

This system uses a selective monitoring method that complements the limitations of simple visual monitoring, and supports hundreds of types of heterogeneous cameras as an open platform software. Since only screens requiring monitoring, such as people and vehicles, are displayed, efficiency can be maximized.

Real-time recognition of vehicle types, facial recognition, and object masking

Kudo Communication unveiled an AI-based smart monitoring solution with Intellix. This solution provides filtering functions according to event types and purposes. It performs intelligent monitoring by using artificial intelligence as a supplementary means of monitoring in a passive monitoring environment.

Using deep learning technology, the types of vehicles detected in video can be analyzed, and by applying this technology, the clothing worn by people can be distinguished. Basic information such as whether long sleeves, short sleeves, or skirts are worn can be analyzed, and it is also possible to identify the gender of the object. Based on stored data, high-speed search analyzes approximately 10,000 images in 5 seconds.

Real-time facial recognition is also possible. By combining fixed cameras and pan-tilt cameras, analysis determines whether it is a person or not, and facial analysis displays age, gender, and expression. This is expected to enhance the efficiency of access control security or be converged into various industries such as retail stores, smart homes, and smart factories.

There is also a solution for cases where objects must be made impossible to identify under personal information protection laws. It is a masking solution that prevents illegal leakage of video data information. Automatic, semi-automatic, and manual masking functions are provided, and real-time masking by head and body is possible using DNN object detection.

Dong-wook Jung, head of the security division at Kudo Communication, said, "Existing CCTVs were only capable of tracking, not facial recognition, and had many errors. When people overlapped, they were recognized as a single object, but this solution recognizes individuals even when they overlap. Third-generation intelligent solutions reduce false alarms and missed detections, and reliably classify data," he said.

Automatically reports fires directly to fire departments via IoT communication network without human judgment

In addition to incident and accident prevention and tracking, intelligent systems can be implemented for disaster prevention. Inkon introduced an intelligent fire monitoring system (Fire Guard). It is a system to secure 240 seconds, the golden time after a fire occurs. The intelligent fire detector transmits fire signals that occur for 5 seconds or more through IoT communication to the manager system and mobile devices.

Existing systems were standalone, where a fire alarm would sound or a report had to be made. The intelligent fire monitoring system sends signals to a server through an IoT communication network when a fire occurs. Once the signal is received, a report is immediately sent to the fire department.

Hyun-chul Kim, executive director at Inkon, said, "The product is wirelessly connected, so we test it twice a day. We check how much battery remains and whether it operates properly."

Currently, the system is installed in 243 stores at Seoul's Amsa Market and Dunchon Market, and is scheduled to expand nationwide to traditional markets, shared and single-family residences, shopping districts, schools, and cultural properties.
본 기사에 대한 정정·반론·추후보도 청구는 보도 청구 안내를, 그간 게재된 보도문은 정정·반론보도 모아보기를 참고해 주세요.
김지혜 기자