May 09, 2023
Let dog let dog go, masked theft, criminals are a bit stupid this year
On July 14, at a garage in Pukou, Nanjing, a man stole property from a car and left two dogs on guard. The police clearly identified characteristics of two dogs through observation, and "locate dog on map" arrested suspect.
August 17, residents of Wenzhou City, Zhejiang Province reported to police that their electric tricycle was stolen. After a surveillance call, police found a suspicious man in a plastic bag who pushed electric tricycle away. After a thorough investigation and detention, police arrested suspect next day.
So many stupid thieves this year. But from these incidents, we also need to reflect. If criminals are deliberately disguised, our facial and fingerprint recognition system cannot identify them. How is the public safety system supposed to block suspects?
Is there technology to identify criminals even if they cover their faces? There really is!
Walk detection at a distance of 50 meters can recognize you even if you cover your face
Gait Recognition is a new biometric recognition technology that combines computer vision, pattern recognition, and video/image sequence processing to identify a person by analyzing a person's body shape and walking posture.
Compared to biometric technologies such as face and fingerprint recognition, gait recognition has natural advantages.
Hard to fake
People can use props to drastically change their appearance and even their gender, but they can't change their posture while walking.
Different body shapes, head shapes, muscle strength characteristics, motor nerve sensitivity, walking posture and other characteristics determine that gait has a better discrimination ability. With well-designed algorithms and lots of data, machines can better recognize these fine-grained features.
Remember Benji from Mission: Impossible 5, who was fully armed and dressed up as a villain? Because "gait recognition" technology was caught...
Wide range of applications
Typically, to achieve a good recognition effect, iris recognition requires target to be within 60 cm, face recognition within 3 meters, and gait recognition can recognize people as far as 50 meters away from conventional 2K cameras in real time.
Uncontrolled Identity
It can also be recognized without active cooperation and participation of target. Fingerprint recognition, iris recognition, face recognition, etc. require active cooperation of target. And it's 360-degree recognition of full view, cross dressing, cross posture.
An artifact of public security system for solving crimesIn course of work of law enforcement agencies, it often becomes necessary to find traces of suspects in crimes on mass videos. Although traditional video surveillance can provide real-time monitoring and screen playback, it is limited by quality of video, and it is difficult to recognize faces in video quickly. Some criminals opposed to investigations make it even harder to identify themselves by covering their faces.
Gait recognition technology can just solve this problem. No matter how a person disguises himself, his gait cannot be disguised. Even if a person wears a mask at a distance of tens of meters and walks with his back to conventional surveillance cameras, gait recognition algorithm can still judge his personality. And it could replace manual blocking of targets from massive surveillance footage, saving police time and effort. It also serves as a deterrent to criminals.
However, gait recognition is not perfect. In simple scenes or with small crowds, gait recognition is indeed very useful for police in identifying criminals. However, in case of a large flow of people, serious overlap of human bodies or mutual overlap, the gait recognition technology may encounter relatively large technical problems.
Integrating several biometric methods such as face and gait recognition and complementing each other could be an important way for future applications.
