** detection is an important guarantee of face recognition security, is an important competitiveness of face recognition manufacturers to popularize products in various industries. At present, there are many methods of ** judgment, but it is difficult to achieve the ideal effect based on a method, often need a variety of algorithms of cross judgment, this paper mainly introduces a simple and effective ** judgment method.

In many cases, high definition photos are very close to what a real person looks like on camera under certain lighting conditions, making it difficult to accurately judge ** based on facial features alone. In most scenarios, when holding a photo, the phone or pad will not stand still completely. However, when there is movement, there is a big difference between the real person and the photo: the real face movement has no correlation with the background, while when the photo, phone or pad is moving, the face movement has a strong correlation with the surrounding background movement.

Based on the above principle, we can judge the correlation between face movement trend and background movement trend to distinguish. First of all, after the face is detected, it expands outwards with the face as the center to form a relatively large area. Then feature points are found in this area and tracked in the next frame. The movement trend of feature points on the face of the person and that of feature points in the background can be calculated

There are many choices for the implementation of the above method, you can use the optical flow field for tracking, or other technologies such as KCF for tracking, as long as the face and background movement has a strong correlation, then it is very likely to be non-* *.

Of course, the judgment based on this method will fail in some scenarios, such as when the photo remains basically still, or when the photo is cropped along the contour of the face. However, this method can detect part of the photos without affecting the passage of real people, so it is valuable, and it needs to be combined with other algorithms to achieve better results.

Computer vision technology service provider, its core technical team from Carnegie Mellon University, including experts from leading biometrics research institutes at home and abroad, developed independent property rights of face recognition, ** detection technology. In 2019, the company free open face recognition and ** detection SDK, can help developers and entrepreneurs save a lot of technology development costs, login shenmotech AI open platform https://ai.deepcam.cn/#/home can be downloaded, Developers can achieve product landing based on the low-cost monocular RGB camera, with excellent real-time performance, silent recognition and good user experience.