Face recognition camera module

Face recognition camera module

Depth camera application

The application of depth camera is fully blooming in the fields of intelligent human-computer interaction, face technology, three-dimensional reconstruction, robotics, AR, etc. The most mature application of commercial depth camera is a variety of interesting applications based on face technology on mobile terminals.

Face technology

Two-dimensional face technology has been developed for decades, although it has been relatively mature, but the detection of key points on the face under various conditions such as large angles, multiple expressions, complex lighting, facial occlusion, etc., wants to achieve real-time robust high-precision detection. raletively hard.The emergence of high-precision depth cameras has greatly promoted the development of face technology, directly jumping from two-dimensional to three-dimensional.It has to be said that the front depth camera of the iPhoneX can be said to have raised the three-dimensional face technology to a new level.It is not restricted by natural light, and it can achieve perfect segmentation even if the foreground and background colors are similar, and its performance under complex head poses is much better than the two-dimensional face technology. The 3D face model reconstructed by the depth camera based on structured light through 30,000 infrared speckles is very fine.

With a fine three-dimensional face model, a series of super practical and fun applications can be developed. Some of the following application scenarios have just been commercialized, and some are still in their infancy.

1. More detailed and natural background blur

Compared with the dual-camera background blur, the depth camera can reconstruct a high-precision three-dimensional face model, the blur effect is stronger, the details are richer, and the portrait is more three-dimensional and natural.

2. Portrait light effect

The portrait light effect function can imitate the lighting effect of professional portrait photography, creating an amazing studio-level effect.

3. Animated emoticons

Animated expressions can transfer facial expressions to dolls in real time. It is a very interesting application. It is thanks to the fine three-dimensional face model obtained by the depth camera that such fine expression capture can be achieved.

4. Three-dimensional beauty

It is not an exaggeration to say that three-dimensional beauty is a dimensionality reduction blow to two-dimensional beauty.The two-dimensional beauty effect is generally exaggerated, and many features of the face itself are lost, resulting in the embarrassment of "not recognizing yourself" often.The three-dimensional beauty emphasizes reality and three-dimensionality. It not only can fully inherit the effect of two-dimensional beauty, but also can customize the "micro plastic surgery" according to the three-dimensional face shape, such as rhinoplasty, lip augmentation, cutting off the high cheekbones, and removing the double Chin, adjust facial features, etc.;In addition, you can add light and shadow effects, such as adding shadows on both sides of the nose and cheeks, making the face look more three-dimensional and realistic.

• Three-dimensional face recognition in vivo

Three-dimensional face recognition is a truly safe face recognition, which can effectively avoid cracking methods such as photos/screen images, face-changing algorithms, masks/3D face models, etc. Very suitable for live detection.

Many times we did not look directly at the camera when we took beautiful photos, and when we looked back, we regretted it.In fact, this can be corrected by algorithms.According to the three-dimensional model of the face, combined with the face orientation of the person, we can adjust the direction of the line of sight to the direction facing the camera, and the visual effect produced is more impactful.Of course, this is just a typical application. In theory, we can arbitrarily manipulate the movement of the eyeball, and based on this, we can develop some fun (strange) applications.

Intelligent human-computer interaction

1. Human skeleton extraction and tracking

The Kinect series of depth cameras launched by Microsoft are specifically designed for somatosensory games.

The key technology behind somatosensory interaction is human skeleton extraction and tracking.Traditional RGB image-based skeleton extraction and tracking technology declines quickly when there are multiple people overlapping, while the depth map generated by the depth camera can easily distinguish human bodies and backgrounds at different distances, which is very conducive to multiple people overlapping. Extraction of different human skeletons.

• Gesture recognition and tracking

Similar to human skeleton extraction and tracking, gesture recognition and tracking also belong to natural body language.Compared with RGB cameras, depth cameras can extract and track finger key points faster and more accurately.

Many practical and interesting applications can be developed based on gesture recognition and tracking.The first wide-ranging application is game entertainment: for example, in shooting games, you only need to hold your hand like a pistol and pull your index finger to "fire the gun." This body language familiar from childhood is used in games. In, make users feel natural and kind.The second is that there are great application requirements in special industries with harsh or more dangerous environments:For example, the non-contact control of clean room machines or equipment working in hazardous areas through gestures can solve many practical problems.