Shulou(Shulou.com)06/01 Report--

This article will explain in detail what are the disadvantages of face recognition on the Internet. The editor thinks it is very practical, so I share it for you as a reference. I hope you can get something after reading this article.

The disadvantages of face recognition: 1, there will be errors, which will affect the judgment results; 2, the reliability and stability of information are weak; 3, the amount of information contained in human face is relatively small, and the complexity of its change is not enough. Identification is not very high; 4, the internal changes of people and the changes of the external environment will affect the stability of face information when collecting.

The operating environment of this tutorial: windows7 system, Dell G3 computer.

From the perspective of technical level, human face is the only biometric information that can be collected without the active cooperation of users. The acquisition process of other biometric features, such as fingerprint, palmprint, iris, vein and retina, all need to be based on the active cooperation of users, that is, if users refuse to collect, they can not obtain high-quality feature information. From a social and psychological point of view, through face recognition identity, in line with human visual recognition experience, easy to be accepted by users. For example, when people collect fingerprints and iris, they worry about privacy leakage, but they do not feel violated when they are photographed by hundreds of surveillance cameras on the street every day, because faces are naturally exposed and are considered to be a natural feature of identification. So let's talk about the disadvantages of face recognition technology.

Technical drawbacks of face recognition

There will also be errors in face recognition technology, which will affect the result of human judgment.

One of the disadvantages of face recognition is that the reliability and stability of information are weak.

The amount of information contained in human face is less than that of fingerprint, iris and other biological features, and the complexity of its change is not enough. For example, if two people's fingerprints or iris are basically the same, it may take dozens or even hundreds of bits to be completely coincident. But if it is a human face, a dozen bits can be coincident. Many similar faces can be found all over the world. Therefore, the face recognition is not very high, it is not so unique.

In addition, the internal changes of people and the changes of the external environment will affect the stability of face information when collecting. Compared with the previous face recognition technology, the current face recognition technology has been improved, but the specific application is still not perfect. It is conservatively estimated that the accuracy of face recognition technology can reach 99%.

Technical difficulties of face recognition

1. Lighting problem

Illumination change is the most key factor affecting the performance of face recognition, and the degree of solution to this problem is related to the success or failure of the practical process of face recognition. Because of the 3D structure of human face, the shadow cast by light will strengthen or weaken the original facial features. Especially at night, the facial shadow caused by lack of light will lead to a sharp decline in the recognition rate, which makes the system difficult to meet the practical requirements. At the same time, the theory and experiment also prove that the difference of the same individual caused by different light is greater than that of different individuals under the same light. Lighting is an old problem in machine vision, especially in face recognition. The solutions to the lighting problem include 3D image face recognition and thermal imaging face recognition. However, these two technologies are far from mature, and the recognition effect is not satisfactory.

2. Attitude problem

Face recognition is mainly based on human facial features, and how to recognize facial changes caused by posture has become one of the difficulties of this technology. The pose problem involves the facial changes caused by the rotation of the head around three axes in the three-dimensional vertical coordinate system, in which the depth rotation perpendicular to the two directions of the image plane will result in the partial loss of facial information. It makes the pose problem become a technical problem of face recognition. The research on posture is relatively less. At present, most face recognition algorithms mainly focus on frontal and quasi-frontal face images. When pitching or left and right sides are more severe, the recognition rate of face recognition algorithm will drop sharply.

3. Expression problem

The changes of facial expressions such as crying, laughter and anger also reflect the accuracy of facial recognition. The existing technology can deal with these aspects well, whether it is to open the mouth or make some exaggerated expressions, the computer can correct it through three-dimensional modeling and posture expression correction.

4. Occlusion problem

For non-cooperative face image acquisition, occlusion is a very serious problem. Especially in the monitoring environment, the monitored objects often wear glasses, hats and other accessories, so that the collected face images may be incomplete, thus affecting the subsequent feature extraction and recognition, and even lead to the failure of face detection algorithms.

5. Age change

With the change of age, a person changes from youth to youth to old age, his appearance may change greatly, which leads to the decline of recognition rate. For different age groups, the recognition rate of face recognition algorithm is also different. The most direct example of this problem is the identification of ID card photos. In our country, ID cards are generally valid for 20 years. During these 20 years, everyone's appearance is bound to change considerably, and there are also great problems in identification.

6. Face similarity

There is little difference between different individuals, the structure of all human faces are similar, and even the structure and shape of facial organs are very similar. This feature is advantageous for using human face for location, but it is disadvantageous for using human face to distinguish human individuals. Human factors such as makeup and plastic surgery aimed at imitating a star make this problem more difficult. Especially the question of twins, whether the face recognition system can be recognized correctly or not is also debated in academic circles. Some experts believe that twins should not be distinguished by face recognition technology at all, and it is impossible to distinguish them accurately by face recognition technology.

7. Dynamic identification

In the case of non-matching face recognition, the blurred facial image caused by motion or incorrect focus of the camera will seriously affect the success rate of facial recognition. This difficulty is obvious in the use of security and monitoring identification, such as subway, highway, station, supermarket, border inspection and other security and monitoring identification.

8. Face anti-counterfeiting

The mainstream deception method of forged face images for recognition is to establish a three-dimensional model, or the grafting of some expressions. With the improvement of face anti-counterfeiting technology and the introduction of intelligent computing vision technology such as 3D facial recognition technology and camera, the success rate of forged facial image recognition will be greatly reduced.

9. Lack of samples

The face recognition algorithm based on statistical learning is the mainstream algorithm in the field of face recognition at present, but the statistical learning method needs a lot of training. Because the distribution of face image in high-dimensional space is an irregular manifold distribution, the samples that can be obtained are only a very small part of the face image space. How to solve the problem of statistical learning under small samples needs further research. In addition, the face image databases participating in the training are basically foreigners' images, and there are few face image databases about Chinese and Asians, which makes it more difficult to train face recognition models.

10. Image quality problem

The sources of face images may be varied, and the quality of face images obtained is also different due to different acquisition equipment. especially for those face images with low resolution, high noise and poor quality (such as face pictures captured by mobile phone cameras, pictures taken by remote monitoring, etc.), how to effectively recognize faces is a problem that needs to be paid attention to. Similarly, the impact of high-resolution images on face recognition algorithms also needs further research. Now, in face recognition, we generally use face pictures of the same size and similar definition, so the problem of image quality can basically be solved, but in the face of more complex problems in reality, we still need to continue to optimize processing.

Security risks of face recognition

In recent years, face recognition technology has made increasingly innovative breakthroughs, and the application projects that have landed among various industries are obvious to all, but as far as the current technology is concerned, it is still unable to keep up with the rapidly changing social changes and market demand. For example, under the surprise attack of novel coronavirus this year, a large number of face recognition products in our country are unable to scan and recognize with masks. After that, the major manufacturers immediately updated the algorithm, but it also reminded us at this time. In the face of future uncertainty, technology cannot remain unchanged and requires constant innovation and breakthroughs.

In addition, how to better recognize the face under different lights and angles? How to clearly and accurately determine the identity and other problems are still the technical pain points to be solved urgently.

A study conducted in 2012 showed that facial algorithms provided by supplier Cognitec performed 5% to 10% lower than Caucasian recognition in identifying Caucasians; in 2011, researchers also found it difficult to distinguish Caucasian from East Asian facial recognition models developed in China, Japan and South Korea. In February, researchers at the MIT Media Lab pointed out that face recognition technologies from Microsoft, IBM and Chinese manufacturer Megvii had an error rate of 7% in identifying lighter-skinned women, 12% in darker-skinned men and 35% in darker-skinned women.

Examples of algorithmic errors are much more than that. The results of a recent survey show that the system deployed by the Metropolitan Police produces up to 49 mismatches in each actual application. At a hearing last year by the House Oversight Committee on face recognition technology, the FBI admitted that its algorithm for identifying suspects had a misjudgment rate of up to 15 per cent. In addition, an ongoing study by researchers at the University of Virginia found that two famous sets of research images, ImSitu and COCO (COCO, built by Facebook, Microsoft and startup MightyAI), show significant gender bias in the description of sports, cooking and a variety of other activities (for example, shopping images are generally associated with women, while coach images are often associated with men).

How to better recognize the face under different lights and angles? How to clearly and accurately determine the identity and other problems are still the technical pain points to be solved urgently.

However, even if the problem of bias is solved, that is, the face recognition system can operate in a fair and fair way to everyone, there is still a potential risk of failure. Like many other artificial intelligence technologies, even if biased factors are completely excluded, face recognition schemes usually have a certain degree of error. All tools can be used for good or evil, and the more powerful the tool itself, the more obvious the benefits or damage it may bring.

This is the end of the article on "what are the disadvantages of face recognition on the Internet". I hope the above content can be of some help to you, so that you can learn more knowledge. If you think the article is good, please share it for more people to see.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.