Recently, in the graduation project, a group of students of FPT University in Ho Chi Minh City “launched” a project to identify criminals using artificial intelligence.

 

With the creative plus, the research topic of the group has received high appreciation from the board and won a good grade for graduation projects.  The authors of the project include: Le Hung Son, Vo Hoang Viet, Nguyen Thanh Nha and Nguyen Kien Huy (studying the same field of Software Engineering).

 

“Today image recognition technology is developing rapidly, becoming one of the most attractive technologies to develop.  Current security camera surveillance is often man-made, so it is easy to make mistakes.  Therefore, our group wants to develop image recognition technology to reduce the efforts of the guards and soon, criminals will no longer be a concern, ”Le Hung Son (team leader) explained the reason for implementing this project.

 

With artificial intelligence (AI) technology, the team's project makes the security management work through the camera more stringent with the application "Criminal Face Detection".  Accordingly, many places frequented can use the system to detect unwanted objects appearing in the area.  And the application will alert security personnel to monitor the above objects, in case of possible bad situations.

 

 

About the project, Son shared: “The system recognizes images on multiple cameras.  When an unwanted object appears, the signal is immediately sent to the manager via software, website, email or app on the phone.  In addition, occurrences or criminal information will be stored in a list, convenient for future reference. ”

 

The photos will go through 2 steps that are identifying the frame containing the face, then use AI technology to identify that face.  This system recognizes a number of frames that share the face attribute.  Then, the device will convert the properties of the face (eyes, nose, and mouth) into vector form and calculate and compare the level of that face with the models which are already in the system data.  “The system allows the organization to manage the undesired object list with the addition to the system 30 to 40 images of each case.  Then, through machine learning technology, the computer calculates and 'learns' the characteristics of the faces that the organization has provided to the system.

When each object appears in the camera area, the device identifies the face and calculates the percentage of the similarity between the face that appears in the camera with the learned data set.  If that rate is over 90%, it will be evaluated as the face of the unwanted object and create a warning for the organization”, Son said. According to Son, crowded public places’ security need to be ensured, especially stations, airports, apartments ... The facial recognition system will help the building managers or security guards not only have one more step of screening, but also can scan across the entire area that need to be observed.

“In cases where the subject’s face appears to be partially obscured, such as when looking their profile or if they are covering a mask, the system can still identify and give similarity.  Because during the experiment, our team also identified unwanted objects while wearing masks”, Son is proud of the group's efforts.

When mentioning to future plans, Son excitedly shared: “I have more ideas for the product developments.  For example, when a child is lost in a public place like a mall, the system will help their relatives to find through the camera system by giving some photos of their child to the security guards... ”

   ACCORDING TO THANH NIEN