1. Deep learning helps face recognition technology continue to heat up

　　 Face recognition is not a new term for access control applications. Many companies are using face recognition to check attendance. Traditional face recognition technology uses the method of extracting feature points to model the face as a whole. This method is greatly affected by factors such as angle, light, expression, image quality, etc. In addition, the embedded system has limited computing power and the recognition accuracy is limited, which also affects the end user's application experience.

Different from traditional single-machine face recognition based on embedded platforms, the new generation of face recognition algorithms are mostly based on high-density GPU architecture hardware, using deep learning technology to achieve face modeling, comparison, retrieval, and other functions, and can be compatible with various industries. The combination of platforms realizes real-time facial alarm, identity verification, personal identification comparison, database retrieval, and other applications, and builds artificial intelligence and security big data solutions. With the guarantee of high-reliability big data calculations, the new generation of face recognition algorithms uses local feature classification and modeling methods to refine the differences in human facial features, which not only greatly improves the accuracy of face recognition, but also increases with the number of training samples. The accuracy rate will continue to upgrade, which will fundamentally improve the user experience of face recognition. It not only improves the performance of simple applications such as attendance, but also opens up new applications such as identification comparison, personnel retrieval, and face payment. door.

　　 One of the most eye-catching products at the 2017 China Security Expo is the personnel channel based on deep learning algorithms. Unlike traditional face time attendance, the person ID comparison channel does not need to enter the face information in advance, and directly extracts the photos in the ID card and compares them with on-site photos to ensure that the tester passes through the customs with his own ID. The core value of this product is that the algorithm replaces the manual judgment that the person holds a certificate, which not only improves the accuracy of identification but also reduces the uncertainty caused by factors such as the fatigue of the attendant.

There is a broad promotion space for the identification comparison channel, as well as vertical, desktop, and visitor identification products, in border inspection ports, traffic stations, bank insurance, education examinations, visitor registration, etc., and there is a broad promotion space for current ID inspections. In all scenarios, the person-to-card comparison system can be promoted. However, before the introduction of deep learning algorithms, this technique was very risky. As we all know, the photo pixels in the ID card are very low, and the ID card is valid for up to 20 years or even permanently. The traditional overall modeling method is difficult to avoid the impact of face shape changes on the recognition accuracy within 20 years. The local feature classification modeling used by the person-to-document comparison algorithm can select parts of the face that are not easy to change for sampling calculation, thereby minimizing the impact of overall appearance changes and ensuring the accuracy of the comparison. It can be said that it is the continuous increase in computing performance and the introduction of deep learning algorithms that have laid a good foundation for the explosion of face recognition applications.

　　 2. Network applications put forward higher requirements for information security

In recent years, cloud computing and platform strategies have penetrated into all aspects of the security industry. Cloud video, cloud alarm, cloud intercom, and cloud parking have been familiar to everyone. Of course, this also includes cloud access control, but from the landing case and the maturity of the system From a perspective, cloud access control may be relatively lagging behind in many security clouds.

　　The development of cloud access control will be affected by two factors: the first is the core needs of users. Different from cloud video remote monitoring, the core function of the door access control system is to open the door with authentication. In high-security areas, it can even set up multiple permissions to open the door or limit the time permission to open the door. This complex logic setting is highly personalized and has many changes. Platform management requirements are high. Another reason is that access control is different from other systems. It is directly related to the access control of the core protection area. Network security has special significance for the access control system. If it is not effectively protected, the risk will be greater.

　　 We often encounter collisions between different business models in our projects: the cloud platform aims to acquire the number of users, cross-subsidies through advertising operations and other methods, drastically lowers equipment quotations and even third-party payment system construction costs. As a business model innovation, this attempt deserves encouragement, but we should also pay attention to the possible risks. Generally speaking, the business model based on the operating model has a clear investment payback period, which determines that the overall cost pressure of the system will increase, which can easily lead to the reduction of equipment quality; if the system is based on a cloud platform, it will also be due to cost recovery. Under the pressure of time, we need to pay special attention to how much resources the system builder invests in the "iceberg below the water" such as network security, and what effects can be achieved.

For all manufacturers, it is not difficult to be an APP-connected device, but truly choose reliable equipment, pass a certain level of network security experiment certification, can be deployed nationwide and even globally, and users are really like in the cloud no matter where they go It is not easy to be connected everywhere without worrying about the network being invaded. Especially the investment in "the iceberg below the water surface" requires manufacturers to have a strong sense of responsibility and a sense of mission. Mentality can really make achievements.

　　 3. Integration and innovation bring new opportunities for the development of access control

Face recognition is inseparable from monitoring and taking pictures, and cloud platforms cannot be separated from network communication. This is another major feature presented at the 2016 China Expo: the cross-border and system integration of manufacturers has injected new vitality into the traditional security industry. It also brings new opportunities and challenges. Many companies that focused on algorithms in the past set up independent booths at the 2016 China Security Expo, and combined with traditional security products, they released many security products with the "Internet +" gene-a a very important part of them is access control and speed. Doors and other entrance and exit control equipment.

Taking S4A access control products as an example, the leading monitoring manufacturers in the industry are seeking to transform to system solutions and provide integrated, one-stop industry solutions. Many have released access control, personnel access, video intercom, and alarms. Such product lines can basically cover the various subsystems of conventional weak current security, and use the integrated platform to form their own unique advantages, which not only improves the delivery efficiency but also improves the end-user experience.

　　 Access control products, as an integral part of the security system, are integrated with other systems in three levels:

At the equipment level, we saw a video access control all-in-one integrated card swiping, fingerprint, intercom, and IPC functions at the China Security Expo. Not only can it directly record in real-time on the NVR, but it can also identify the QR code through IPC. It will not even be in the future. Exclude deep learning algorithm functions such as the integration of witness comparisons. Similarly, most of the video intercom systems introduced by traditional monitoring manufacturers can also directly call the monitoring system to achieve functions such as doorway video recording and indoor extensions to call community monitoring pictures. Once this type of "cross-border" equipment was released, it aroused great interest from the audience. Not only could it improve the security level of the system, but the product itself was also a novelty and bright spot in the project.

　　 At the system level, the integration and integration of devices also brings a better experience for users, and the most typical one is mobile apps. With the increasing popularity of smartphones, more and more users want to control security equipment through mobile phones. If a system deploys multiple platforms, it usually means that the end-user needs to install multiple APPs on the mobile phone to control each subsystem separately. In contrast, it is very attractive to connect to the system of a comprehensive security management platform to realize the real system integration scheme and realize the association and control between the various subsystems. This has become more and more attractive. The plan that many users desire in their hearts.

　　 In addition to equipment and systems, service and support are also becoming more and more important. The security industry is a highly comprehensive industry with a wide range of products, rapid knowledge updates, and wide regional coverage, which poses great challenges to the technical support capabilities of practitioners. In contrast, traditional access control products are systems with relatively distinct regions and industries. There are more regional brands, and the requirements for system integration and cross-regional project support are relatively high. If we can combine distinctive products with highly integrated platforms and localized support services, I believe it will play a very good role in promoting product development or project delivery.

　　 Conclusion

　　 Driven by integration and innovation, access control products have rapidly leaped from the original access control card application to new technologies such as biometrics, big data, and deep learning. New technologies are not only rapidly breaking through the original technical barriers and establishing new industry technical standards, but also opening up a broader application space for access control products and even the security industry. In the future, under the general trend of "Internet +", only by actively embracing changes and exploring the integration and innovation points of access control products with other systems can we better grasp market opportunities, achieve better performance, and contribute to the integration of products. Contribute to the development of the industry.

Author: Written by S4A from S4A INDUSTRIAL CO.,LIMITED