This site uses cookies. By continuing to browse the site you are agreeing to our use of cookies. Read our privacy policy>Search

If you need help, please click here:

How We See Software-Defined Camera

2019-09-05 656
Emotional male voice
  • Emotional male voice
  • Emotional female voice
  • Standard male voice
  • Standard female voice
Speed 1.0X
  • Speed 0.5X
  • Speed 1.0X
  • Speed 1.5X

Software-Defined Camera (SDC) will inevitably bring major changes to the security industry, where hardware still prevails and intelligence has only recently emerged.

Having studied the industry and major trends, we believe we can anticipate how the industry will be affected when cloud and AI technologies converge. We believe SDC will provide many industries with more opportunities, and that intelligence will greatly improve production efficiency.

What is software-defined technology?

Based on hardware and vitalized by software, software-defined camera will reshape the whole industry.

As a product of hardware development, an SDC is a prime example of improved product value. Take cellphones, for example. Over the past few decades, 2G phones have been replaced by 4G and now 5G models. They are playing increasingly important roles in our lives. We can use them to choose the best routes using mobile maps, keep in touch with friends worldwide via chat apps, and watch movies through video player apps. How do cellphones improve our lives? By supporting games that keep us entertained? By making it easier and more convenient to do business? Or by making our daily lives much more convenient? All of these functions, based on our wants and needs, are supported by smartphones that enable us to download and upgrade apps on demand. Similarly, cameras have also moved from the analog era, through the digital era, and now to the AI era. With this transformation, different industries have varied requirements of cameras. For example, campus surveillance —such as traffic surveillance — requires different camera applications for people and vehicles. With this in mind, we want our software-defined concept to allow cameras to flexibly adapt to various application scenarios in numerous industries.

SDC is a product of the AI era.

Software-defined camera actualizes AI’s potential.

A product’s development reflects the trends of an era. Created in the AI era, SDC reflects both the current industry landscape and its sustainable development in the future.

  • Diverse requirements: Transforming a single data source to multi-dimensional data match, to meet various demands.

From face detection to face and body detection, the dimensions of data analysis are expanding. In the future, algorithms and data in more dimensions may also play a part in this course — such as those for recognizing postures, facial expressions, and 3D modeling. This is where the software-defined camera comes in — to meet the requirements in increasingly extensive application scenarios, with continuous evolution of camera algorithms and performance.

  • Ever-changing technologies: Market changes lead algorithm evolution and technology development.

Because algorithms are continuously being trained, software updates are becoming more frequent than ever. New versions go live every month, even every week in some cases. Because of this, the decoupling of camera software and hardware can ensure regular updates without interrupting basic services.

  • New market landscape: Driven by AI, the number of algorithm developing companies is growing quickly.

In a robust market, infrastructure providers are most effective when they are open and inclusive enough to keep up with trends and cooperate closely with leading technology providers. With this in mind, we have built an open-ended ecosystem to ensure beneficial collaboration among enterprises, serving users with quick response and comprehensive services.

  • Large inventory: In scenarios in which there are large existing inventories, it is difficult to upgrade to intelligent cameras.

During the transition to the AI era, it would be wrong to hastily replace old products with new ones. Instead, it is better make full use of inventory products to maximize your return on investment (ROI). In this case, a highly inclusive product will boost industry development and maximize ROI throughout the lifecycle of inventory products. Huawei SDC can share their compute power; one SDC can connect to multiple common cameras and add intelligence to these cameras to gradually achieve full intelligence.

  • Complicated management: Complex networking and massive data complicate camera management.

In the future, camera algorithms and application scenarios are becoming increasingly diverse, and countless cameras need to be managed. Therefore, we need a visualized and manageable product that can be remotely upgraded to achieve smart operations and maintenance (O&M). This product can interoperate with the cloud, load and upgrade algorithms online, and manage cameras online, making camera management easier and more efficient.

In the AI era, a competent SDC needs to meet three requirements: high quality AI chips, open-ended SDC OS, and a future-oriented algorithm and application ecosystem.

What kind of cameras are software-defined?

Software-defined camera adopts a flexible and open-ended architecture that promotes agile management.

Architecture is like the framework of a software system. We need to reform the traditional all-in-one architecture and build a flexible, open-ended, and manageable software system.

1. High Quality AI chips: Traditional intelligent surveillance devices usually have a separate Central Processing Unit (CPU) chip — such as DSP, ARM, or FPGA — to execute intelligent analysis algorithms inside the camera. Although such a design meets current requirements, the camera upgrade space is limited. As urban security challenges become increasingly complex, deeper and broader analysis is required for people, vehicles, and objects. Improving the accuracy of a single algorithm and supporting concurrent processing of multiple intelligent algorithms in segmented scenarios have become a priority. With these greater and more complex demands, more computing bottlenecks are inevitable. In this context, Huawei launched an AI chip, the Ascend 310, at the end of 2018. It features ultra-high computing power (16 TOPS) and low power consumption (only 8 W). It is the optimal AI chip with extraordinary performance. With superior computing power, one Huawei SDC can add intelligence to three or four common cameras nearby, so that more targets can be detected with detailed attribute information. In practical applications, critical events can be efficiently handled with pre-warning. Huawei SDC adopts high performance neural processing units (NPUs), such as Ascend chips, with a computing power 25 times that of CPUs, enabling visual analysis and computing of trillions of records.

2. Open-ended SDC OS: The SDC OS, based on multiple algorithms operating simultaneously, enables software to be decoupled from hardware. Unlike traditional cameras, which upgrade algorithms through firmware updates and camera reboots, Huawei SDC supports independent operation of multiple algorithms, and services are not interrupted even during algorithm upgrade and switchover, ensuring data completeness.

3. Future-oriented algorithm and application ecosystem: With the open-ended OS, an end-to-end ecosystem tool chain is now available for developers to connect, train, and roll out algorithms and applications following specific standards. Huawei built an app store-like algorithm store to meet customers’ needs for front-end intelligent applications. This algorithm store provides a one-stop application development platform for algorithm partners to develop, launch, and transact algorithms. With this store, Huawei and its algorithm partners promote the development of intelligent applications and response to customers’ needs, making AI a technology truly beneficial to all. In addition, to simplify algorithm development, Huawei has built a prosperous ecosystem with a full-fledged tool chain featuring SDC Studio and ModelArts. Various algorithms turn common cameras into dedicated cameras within seconds and add intelligence to a variety of industries.

This is how we see the development of software-defined camera. We hope to share our thoughts and ideas with the whole industry. Much like computers and smartphones have changed the way we connect to the world — we hope SDC can be the enabler that changes the landscape of the whole security industry, bringing it into the AI era.


Write your comment here.


    More comments

      You have scored successfully.

      You have submitted successfully.

      Evaluation failed.

      Submission failed.

      Please write your comment first.