verification system

I am developing a facelogin application. I could able to identify who the person with 80% accuracy. I need to differentiate between

1. A person standing infront of camera and
2. A poster of the same person kept infront of the camera.开发者_开发技巧

Is this possible?

I did the login using opencv

Help is welcomed from any programming language.

Thanks

---

Take multiple snapshots in short succession. The facial expression on a photo does not change, the face on a photo does not blink with the eyelids and extrapolating the relative depth of the facial elements by means of tiny changes in angle between the snapshots yields a distinct result for photos.

If you can use 2 cameras, calculate the transformation between shots taken by both cameras at the same time. This transformation should be different for plane surfaces and faces (even though you might not be able to extrapolate the actual threedimensional structure).

---

Any such method can be fooled by several videorecords, showing the same person in different positions. For example, you may ask a person to turn around, and this can be recorded and submitted to your program. In the same way, you may decide to, for example, highlight person's face with external light sources, but this also can be recorded and reproduced.

My advice is - do several auth movement series (e.g. nod, head-bobbing, closing eyes, etc) - more is better, then ask user to execute one or more of them in some random order pretty fast. You also may log and count insuccessful auth attempts. Then you will be notified, when someone missordered records for a few times.