Face recognition has been basically implemented and scenarioalized applications have been realized. At present, face attendance, punch in, access control, personnel tracking, and personnel recognition have also been developed and mature. The face recognition developed by Qingxi Video Members has also been put into use, such as face detection in a scenic spot and smart construction site scenes. In the research and development at the same time, we also found some problems and difficulties, this article to summarize the problems and overcome some difficulties in our research and development.

  • Keep out. The face in the image may be blocked by other faces or blocked by the background, so that only partial face is leaked during detection. In addition, the face can also be blocked by facial appendages, such as glasses, masks, long hair, beard and so on.

  • Lighting. Different spectrum, light source position, illumination intensity and so on will affect the appearance of face image. In a backlit environment, the face may be too dark to see details. Under a single strong light source, the face may appear “Yin and Yang”.
  • Facial expressions. The relative position of the face and the camera lens determines the diversity of the face posture, such as up and down pitch Angle, left and right Angle, vertical Angle of rotation, different angles will have different effects.
  • Image quality. The source of the face image may be varied, due to the different collection devices, the face image quality is different also, especially for those low resolution and large noise, poor quality of the face image face images (such as mobile phone camera, remote monitoring, etc.) of images taken by how effectively face recognition is a problem that need attention. Similarly, the influence of high resolution image on face recognition algorithm also needs further research.
  • The motion is blurred and the camera is not focused correctly. Blurred facial image caused by motion or incorrect camera focus will lead to inaccurate face information received, resulting in inaccurate recognition.
  • Face similarity. The global population is large, in addition to the parent-child relationship looks similar, even many unrelated people also have similar, which is beneficial for the use of face positioning, but for the use of face to distinguish human individuals is unfavorable.

  • Physiological changes. Facial appearance changes with age, especially for teenagers. For different age groups, the recognition rate of face recognition algorithm is also different. As a person changes from a teenager to a young man or an old man, his appearance may change considerably, resulting in a decline in recognition rate. For different age groups, the recognition rate of face recognition algorithm is also different.
  • Lack of big data samples. Face recognition algorithm based on statistical learning is the mainstream algorithm in face recognition field at present, but statistical learning method needs a lot of training. Because the distribution of face image in high dimensional space is an irregular manifold distribution, the sample that can be obtained is only a very small part of the face image space sampling, how to solve the problem of statistical learning under small sample needs further research.
  • Face security. In the specific application of face recognition in the future, the main problem is changed from “identification of human identity” to “judging whether the face in front of the system is a real person”, which is also known as face anti-counterfeiting.

However, with the continuous maturity and progress of face recognition technology, the difficulties in face detection will gradually be solved.