In December 2019, Central Security Service announced a security robot with a unique design. In the future, we believe that the use of security robots will be essential to address the labor shortage of security guards and improve security quality. rice field.
Security robots are required to be able to (1) prevent damage, and (2) prevent the spread of damage, which is the essence of security. The company says that in recent years, due to the development of technology such as image analysis technology, there are great expectations for robots to complement security guards.
And NVIDIA's blog reported that "Jetson AGX Xavier" was adopted for its "eyes" (vision).
■Main functions
1.Face recognition function (whitelist detection, blacklist detection)
2.Object recognition (white cane detection, knife detection, etc.)
3.High-temperature object detection
>4.Image recording function with omnidirectional camera
5.Remote threat function (siren threat, flash light threat)
■Assumed use cases
1.Patrol in public spaces such as stations and airports Security
2.Security patrols in office buildings
3.Patrol security in commercial facilities
4.Other facilities that require patrols and patrols
Examples of use cases( Railway security)
Robots detect suspicious persons with knives on the premises
Alert notifications of abnormal situations and locations to smartphones of station staff and security guards
From the history While giving instructions, rush to the scene
This autonomous mobile security robot is equipped with eight high-performance cameras. One of these cameras and the ultra-compact AI computer board "Jetson AGX Xavier" perform "object recognition" to detect dangerous objects such as suspicious persons and knives. This object recognition adopts a CNN (Convolutional Neural Network) based deep neural network. By combining AI inference implemented as a trained model with Jetson AGX Xavier's image analysis, it is said that suspicious people and dangerous objects can be detected. When detected, an alert is sent to terminals such as smartphones and tablets owned by the disaster prevention center and security guards assigned to the neighborhood, and if necessary, sirens and flashlights are turned on to threaten suspicious individuals. In addition to detecting suspicious persons and dangerous objects, it can also detect white canes carried by visually impaired people, quickly find people who need support at security sites, and notify nearby personnel can.
In a real environment where many people come and go, high-performance computing power is required to recognize objects such as suspicious persons and various dangerous objects. Another problem is that high-performance computing consumes a large amount of battery power. It will place severe limits on the power consumption and form factor of computers that can be implemented in battery-powered autonomous robots. After evaluating and examining many computing platforms, including CPUs and FPGAs, the Central Security Security development team selected Jetson AGX Xavier as the most suitable platform for object recognition for this autonomous mobile security robot. NVIDIA says, "Jetson AGX Xavier, an embedded SoM (System On Module) that boasts computing power equivalent to that of an AI supercomputer, can be said to be a unique computing platform for realizing the 'eyes' of an autonomous mobile security robot. I can,” he explains.
Central Security has its own learning data set for autonomous mobile security robots based on knowledge accumulated over many years of experience and achievements. Created in In order to enrich the data set, they also shot in a studio that reconstructed the on-site environment as necessary, and created a data set to create a more accurate model. The CNN model is also trained on the NVIDIA platform using the collected data, and the trained model is optimally deployed on Jetson AGX Xavier.
In addition, when the operation of the autonomous mobile security robot in the real environment starts from now on, in order to respond to new situations that we encounter on a daily basis, we will periodically perform additional learning using new data, We have to update to a versatile and highly accurate model. For that reason, Central Security will work on continuous model improvement using the supercomputer "NVIDIA DGX Station". The ability to realize such a smooth AI model development workflow is also an advantage of the end-to-end solution realized by NVIDIA's unified GPU architecture based on CUDA," says NVIDIA.
NVIDIA said in a blog post, ``In recent years, the introduction of various autonomous robots, such as automatic cleaning robots and mobility support robots, has progressed. , Central Security Service uses AI-powered robotics in the field of security, which they have specialized in for many years, and aims to have autonomous mobile security robots perform security better than humans in the future. NVIDIA's end-to-end platform, including Jetson AGX Xavier, selected by Central Security, is the best solution to achieve this objective quickly and efficiently."
(Quoted from NVIDIA's blog)
There are no related articles yet