Although Face Recognition has been developed for 3 to 40 years, several difficulties that it has always existed have not been completely solved yet.
Illumination problem is an old problem of machine vision, especially in the Face Recognition System. Due to the 3D structure of the face, the shadow cast by the light will strengthen or weaken the original face features.
Similar to the illumination problem, the pose problem is also a technical difficulty that needs to be solved in the current face recognition research. The pose problem involves the face changes caused by the rotation of the head around three axes in a three-dimensional vertical coordinate system, where the depth rotation in two directions perpendicular to the image plane will cause partial loss of facial information. There are relatively few researches on gestures. At present, most face recognition algorithms mainly focus on frontal, quasi and face images. When pitching or left and right sides are severe, the recognition rate of face recognition algorithms will also be lower. Will drop sharply. Large facial expression changes such as crying, laughing, and anger also reflect the accuracy of facial recognition.
For face image acquisition in non-cooperative situations, the occlusion problem is a very serious problem. Especially in the monitoring environment, the monitored object often wears glasses, hats and other accessories, making the collected face images may be incomplete, which affects the subsequent feature extraction and recognition, and even leads to face detection algorithms Of failure.
With the change of age, the appearance of the face also changes, especially for teenagers, this change is more obvious. For different age groups, the recognition rate of face recognition algorithms is also different. When a person changes from a young person to a young person to an old person, his appearance may undergo a relatively large change, resulting in a decrease in the recognition rate. For different age groups, the recognition rate of face recognition algorithms is also different.
Face Recognition
There is not much difference between different individuals. The structures of all human faces are similar, and even the structure and appearance of facial organs are very similar. Such a feature is advantageous for using human faces for positioning, but it is disadvantageous for using human faces to distinguish human individuals.
The sources of face images may vary. Due to different collection devices, the quality of the face images obtained is also different, especially for those face images with low resolution, high noise, and poor quality (such as those taken by mobile phone cameras). Face pictures, pictures taken by remote monitoring, etc.) How to effectively recognize faces is a problem that needs attention. Similarly, the impact of high-resolution images on face recognition algorithms needs further research.
The Face Recognition Algorithm based on statistical learning is currently the mainstream algorithm in the field of face recognition, but the statistical learning method requires a lot of training. Since the distribution of face images in the high-dimensional space is an irregular manifold distribution, the samples that can be obtained are only a sampling of a very small part of the face image space. How to solve the problem of statistical learning under small samples needs to be further Research.
Traditional face recognition methods such as PCA and LDA can be easily trained and learned in small-scale data. But for massive amounts of data, these methods are difficult to train and may even crash.
As the scale of the face database grows, the performance of the face algorithm will decline.
