With the deepening of intelligence, automobiles have increasingly become a typical technology-driven industry. Especially under the joint promotion of Moore’s Law and Kumei’s Law, automotive chips continue to innovate, computing power continues to break through, and cost and power consumption continue to decline. So thanks to this, what kind of iteration will the entire auto industry have? Aiming at the future trend, at the recent China Automotive semiconductor Conference, Dr. Chen Feng, deputy general manager of ArcSoft’s Vision Vehicle Business Group, demonstrated the newly upgraded ArcSoft one-stop vehicle vision solution VisDrive 6.0 on the spot, from the perspective of computer vision technology From the perspective of empowering cars, it details how visual AI cooperates with on-board chips to create a safe and intelligent driving experience in the future.
From mobile phones to cars, “algorithm + chip” is still the key variable to ignite intelligence
Regarding the development direction of automobile intelligence, Dr. Chen Feng believes that it will follow the same development path of “interaction change-architecture upgrade-ecological evolution” like smartphones.
As an epoch-making intelligent product, smartphones have completed a magnificent turn from communication devices to universal “scene tools”. According to Dr. Chen Feng, in the process of this change, the application of the camera enables mobile phones to carry out innovative designs from the perspective of visual interaction, thereby subverting traditional functions. He introduced: “Starting from the Sharp J-SH04, the world’s first mobile phone equipped with a rear camera, and then from single camera, dual camera to intelligent multi-camera, more and more lenses need to be applied on mobile phones. Following that, we can Video phone, beauty selfie, background blur, face unlock, AR ranging, etc., and its intelligent capabilities are becoming more and more powerful.”
The mass production and application of so many complex visual AI algorithms to smartphones is behind the evolution of mobile phone chips from the early single-core architecture to the current multi-processor core architecture such as DSP, CPU, GPU, and NPU. The rapidly increasing computing power and memory capacity make smartphones a powerful terminal computing platform.
Following the same path, we can find that similar changes are taking place by analogy with smart cars. Dr. Chen Feng pointed out that the traditional role of automobiles as a means of travel is being gradually broken and transformed into a “mobile third space” integrating entertainment, office and consumption. At the same time, the on-board camera is already one of the necessary sensors for the mainstream configuration of smart cars. Front view, surround view, rear view, side view and built-in five types of cameras can provide drivers with all-round safe and intelligent service capabilities, such as lane assist warning. , panoramic parking, blind spot detection, parking assistance, driver status detection, pedestrian collision warning and many other intelligent driving functions.
Dr. Chen Feng, Deputy General Manager of ArcSoft Vision Vehicle Business Group, delivered a keynote speech
Compared with the single window of the mobile phone, the application of the car is more complicated, and most of them are multi-function and multi-task concurrent mode, requiring higher security. With the increasing electronicization of vehicles, especially the addition of functions such as automatic driving and active safety, the number of single-vehicle ECUs is also increasing rapidly, with an average of 50-70 or more. So many ECUs not only bring complicated wiring harness design, but also lead to mixed logic control. This forces vehicle manufacturers to convert the automotive E/E architecture from distributed to centralized, thereby integrating a large number of ECUs and handing them over to the domain controller for unified management and scheduling. The centralized E/E architecture will separate sensing and processing, which is conducive to the standardization and generalization of underlying information resources, and further reduces the coupling between software and hardware, laying a solid foundation for automobiles to enter the “era of common definition of software and hardware”. base.
But in this way, higher requirements are put forward for the high performance, low power consumption and stability of the hardware, so the on-board chip with powerful processing capability emerges as the times require. It is precisely because of the corresponding hardware changes that Dr. Chen believes that the requirements for various in-vehicle software have also been greatly improved. Because the resources given by the hardware to a single application are extremely limited, in the process of processing various sensory information, the algorithm must not only achieve fast and accurate processing under the condition of low resources, but also maximize the use of the hardware under the condition of ensuring low power consumption. Ability. For a long time, the honing and accumulation of ArcSoft algorithms and various general-purpose chips on smartphones are in line with these requirements.
ArcSoft VisDrive 6.0 is newly upgraded for all scenarios of smart cockpit
As a core algorithm provider for the global smartphone imaging industry, ArcSoft has made remarkable achievements in the iterative process of smartphones. Among the top five mobile phone brands by global shipments, except for Apple, which completely adopts self-developed algorithms, the main mid-to-high-end models of other mobile phone brands are equipped with ArcSoft’s computational photography and computer vision artificial intelligence technology.
Dr. Chen Feng said that it is precisely because of the deep business accumulation in smartphones that ArcSoft has maintained long-term cooperation with top companies in the industry chain such as Qualcomm, MediaTek, and arm. Moreover, ArcSoft has a lot of practical implementation experience for in-depth optimization of the algorithm according to the characteristics of different hardware platforms.
So extending from smartphones to smart cars, what kind of imagination will ArcSoft bring to the industry? At the conference site, Dr. Chen Feng brought the newly upgraded ArcSoft one-stop vehicle vision solution VisDrive 6.0. It runs through the entire scene of the smart cockpit, covering DMS (driver monitoring system), ADAS (advanced driver assistance system), BSD (blind spot detection system), OMS (passenger monitoring system), Interact (visual interaction system), Authenticate (biological certification), AVM (3D surround view monitoring system), AR HUD (AR head-up Display) and other related software solutions based on core algorithms.
Among them, the driver monitoring system of VisDrive 6.0, in addition to commonly used driver behavior recognition, state recognition, emotion recognition, and seat belt detection, also has functions such as fatigue behavior reminders, steering wheel hands-off behavior reminders, and mobile phone behavior reminders. In the visual interaction system, VisDrive 6.0 has a wealth of functions such as gesture recognition, gaze tracking, lip motion wake-up, and gaze wake-up.
In the passenger monitoring system, it has personalized functions such as passenger number detection/expression recognition, leftover living body reminder, child crying detection, cabin photo beautification/video selfie, 3D virtual driving assistant and other personalized functions; in terms of biometric authentication, in addition to the basic face Identification and fingerprint recognition, VisDrive 6.0 has smart services such as keyless start of the car, safe and anti-counterfeiting and convenient payment of parking fees; in the 3D surround view monitoring system, VisDrive 6.0 is equipped with technologies such as surround view splicing, reversing assistance, 3D viewing angle, and transparent body .
At present, VisDrive 6.0 has been deeply optimized and adapted to various chip system platforms such as Qualcomm, Texas Instruments TI, MediaTek MTK, NXP NXP, Renesas, Ambarella, and Huawei. In fact, in addition to establishing long-term and stable cooperative relations with chip manufacturers, ArcSoft has also established close cooperation with major sensor and camera module factories such as Sony Sensors, Samsung Semiconductor, Gekewei, Sunny Optical, and ON Semiconductor. relation.
Right now, the process of cars becoming intelligent has just started. Whether it is various sensing hardware such as chips, camera modules, lidar, or visual algorithm software, every link is crucial. Only the deep integration and cooperation of software and hardware can continuously accelerate the development of the industry. ArcSoft believes that with the cooperation of many excellent manufacturers, it will bring users a new mode of “future driving” that is safer, more comfortable and smarter.
The Links: LQ104V1LG61 PD10M44OH7B06 LCD-SOURCE