The development of REID technology helps improve facial recognition systems

With the development of society, public safety has become a common topic for the whole society, and video surveillance systems that complement it have also been widely popularized. In this topic, we will discuss the technology of video surveillance - REID.

Firstly, what is REID?

ReID, also known as Re identification, is defined as the technique of using algorithms to find the target to be searched in the image database, so it belongs to a sub problem of image retrieval.

Why did such technology emerge?

As mentioned at the beginning of this article, as public safety is increasingly valued, it is only a matter of time before the corresponding improvement in video surveillance system technology. The video surveillance system can intuitively reproduce the target scene and serve as a powerful auxiliary for public security to solve cases. In the work of law enforcement agencies, identifying and locating targets is a crucial step, but under existing monitoring deployments, this critical step is almost completed by manpower. In this era that values data and efficiency, there is clearly a significant waste of resources and low efficiency in manually observing surveillance footage to search for results.

In addition, due to hardware defects such as camera resolution, it is difficult to obtain clear personal images. Therefore, facial recognition technologies that cover a wide range of areas have certain limitations in non target accuracy. Due to the inconvenience of traditional manual queries and the limitations of facial recognition applications, researchers are considering how to invent a more suitable technology to assist facial recognition and replace human analysis with machines in the monitoring field? Therefore, research on pedestrian re identification (ReID) should also be carried out accordingly.

Simply put, in situations where surveillance cannot capture faces, ReID can replace facial recognition to find the target object in the video sequence. Based on such functions, ReID can be widely used in the market by combining security, personal positioning, and other aspects.

ReID technology promotes smart security construction

The concept of ReID was proposed as early as 2006, and now there have been new developments in the security industry. Among them, Yuncong Technology announced a breakthrough in this field in 2018, stating that its core indicators for measuring ReID technology in Market-1501, CUHK03, and DukeMTMC reID datasets have all broken relevant records. Its Rank-1 Accuracy on Market-1501 reached 96.6%, breaking the world record set by Alibaba iDST in January 2018. In addition, Yitu Technology and Kuangshi Technology have also released excellent research results in the field of ReID. It is not difficult to see that in recent years, AI companies in the industry have been investing heavily in the development and research of ReID technology.

Conclusion: Currently, ReID is still imperfect, but we can look forward to the development of ReID technology in the future. For the security industry, combining the characteristics and advantages of facial recognition and ReID at the application level is also a topic worth exploring. The application of ReID technology is one of the important technologies for building smart and secure cities in the future. With deep learning as a key focus of artificial intelligence, combined with ReID technology, its practicality will also be widely expanded.

Prev: Without sensors, the Internet of Things will move towards “hollowing out”