Robots in Japanese technology darkness can also recognize faces

In today's robotics field, vision-based facial facial recognition and detection has developed very rapidly and has been used in many fields.

Face recognition has now been applied to applications including payment, remote service processing, security, and access control. After nearly two months of trial operation, the security face recognition system has been officially put into use in the domestic security inspection area of ​​Shenzhen Baoan International Airport. This is the first security information system in the country to embed the face recognition system into the airport. Many people say that we are welcoming the era of brushing the face. At that time, take the bus to check the ticket, get banned, and pay for the shopping... In many similar scenes, you can get it by simply brushing your face.

However, for now, robots can successfully perform facial recognition because they are already under normal lighting conditions. In order to further develop this technology, we still need to overcome an obstacle: how to successfully face in the case of unclear light? Therefore, before reaching the era of brushing, we need to develop an algorithm that allows the robot to perform facial recognition under abnormal lighting conditions.

Now, researchers at the School of Computer Science and Engineering at Toyohashi University of Science and Technology in Japan have developed a cool technology that uses a reflection model to adjust the effects of light on human faces. This model has a variable illumination ratio, controlled by FIS (Fuzzy Inference System), which can use GA (Genetic Algorithm) to enable robots to perform face recognition in different light.

The research team leader BimaSenaBayuDewantara said:

In order to reduce the impact of light on face recognition, we should adjust the image contrast appropriately. For example, in order to produce a stable facial image, we need to brighten the light from the cheeks and dim the light from the eye. The reflection model we use can achieve this, and combining FIS and GA is also very important for operating such a model.

Another person in charge, Professor JunMiura, said:

By adjusting the image contrast, we can improve the accuracy and effect of face recognition and detection, and also apply it to robot face recognition and human-computer interaction in real time.

The human face can not only distinguish itself from other people, but also provide other information: if you are tired, whether you are happy or sad, peace is still tense. Capturing the information contained in the face is crucial for human-computer interaction. The researchers said that by adjusting the image contrast, the robot can perform facial recognition and detection under different conditions, especially in the case of insufficient illumination. On July 15, the team published the results of the study on Machine Vision and its Applications.

In fact, Toyohashi University of Technology has made some achievements in human-computer interaction. Last year, the University's Interaction and Communication Design Lab developed a robot called "TalkingAlly" that keeps an eye on the user's behavior and responds accordingly.

Its shape looks like the jumping light in Pixar's well-known animated short film "Naughty Jumping Light". It has an oval head and the body is spring-loaded. It has good elasticity and the robot can flex the body flexibly. It has one eye and a camera in the eye, the camera will always focus on the user, TalkingAlly will recognize the face and expression. TalkingAlly was designed primarily to keep the user focused. By observing where the user's gaze falls, the robot will determine where and where his attention is.

It seems that with the rapid development of human-computer interaction, the era of brushing the face is not out of reach. But to get there, you have to overcome the difficulties.