With strong R&D and manufacturing capabilities, SunWin, a major camera module manufacturer, has successfully entered the field of vehicle panoramic parking. This is the exact news released by our sales director, Mr. Li at the "4th Smart Car Developer Summit" in Shenzhen on June 23.

"The company has had an in-vehicle operation team since 2017. In August last year, it officially established Sanyingxing in-vehicle intelligent technology. It is currently delivering related products to domestic first-class OEM customers." Mr. Li Zheng said that in addition to in-vehicle camera products In addition, Sanyingxing also provides services for algorithms and domain controller solutions.

Hubei Sanyingxing Optoelectronics Technology Co., Ltd. was established in 2011 and is headquartered in Tongcheng Economic Development Zone, Xianning City, Hubei Province. The company is mainly engaged in intelligent car imaging modules, smartphone camera modules, biometric modules, 5G artificial intelligence imaging R&D, manufacturing and sales of four major categories of imaging digital solution products including modules, related products are widely used in smart phones, smart cars, biometrics, artificial intelligence, notebook computers, medical imaging, Internet of Things, smart transportation, multimedia teaching and other fields.

At the end of April this year, the China Securities Regulatory Commission announced that the initial listing application of Hubei Sanyingxing Optoelectronics Technology Co., Ltd. was reviewed and approved by the Issuance Examination Committee of the China Securities Regulatory Commission.

A number of vehicle industry experts gave Sanyingxing a high degree of recognition on the spot. In view of the very exciting content of Mr. Li Zheng's speech, the following content of Chaodian Smart Car has been edited slightly to try to present it in its original flavor.

01 The outbreak of smart cars, the APA market size will exceed 44 billion

From 2016 to 2022, the growth rate of the entire self-parking industry is 32.1%. Its market size is expected to reach 44.21 billion yuan in 2025, with a five-year growth rate of more than 50%. In the field of autonomous parking and intelligent driving, China has always been at the forefront of the world because of its largest market and largest consumer group.

APA, that is, the demand characteristics of automatic parking are very clear. The first is to detect the angle and distance information of the vehicle body and the surrounding environment through perception systems such as cameras and ultrasonic radars. Second, the detection and recognition of the entire parking space, on the one hand, is to identify the parking space through information detection, and may also recognize the rotation angle of the steering wheel and the sensing orientation of the tire; in addition, including the ultrasonic parking space and visual parking space brought by ultrasonic radar, the second step It is detection and identification. The third step is through the central processing unit, domain controller or platform, and the planning of the parking path, while integrating and adjusting to the steering wheel, brakes and accelerator.

At present, NIO, Ideal, Xiaopeng, GAC, Dongfeng Blueprint, Geely and many others have realized automatic parking. The next step is the tracking of the parking trajectory, which is commonly referred to as memory parking or trajectory parking. It is a vague variable, but it must be a development trend. In the end, the parking status of the whole vehicle will be displayed in the central control of the body or in the rearview mirror and streaming media, and sent to the background or the owner's mobile phone through the T-BOX and the body information unit. .

In the future, from APA automatic parking, remote control parking RPAL2+, to memory parking HPPL3, it will evolve to the ideal autonomous parking AVPL4. The car owner does not need to be on the scene or online, and can directly park or summon the vehicle through the accumulation of scenes and data without user interference. This is the path of technological development.

02 From panoramic to automatic, five links will smooth the future of parking

Mr. Li Zheng believes that moving from panoramic parking to automatic parking is a breakthrough in the quality of intelligent driving, but five key steps need to be completed.

The first step is to have a very high pixel or fisheye vision camera with a recognition rate. The previous 1-megapixel and 2-megapixel camera needs will be eliminated from 2020, and now the mainstream is 3-megapixel to 8-megapixel car high-definition cameras.

In Li Zheng's view, this is the most difficult step. Because it means that all material supply chains and production and research equipment need to be changed, all technical data and original experimental data must be accumulated, and all materials, even a lens module and a glass lens, must be re-engineered. 's verification.

The second step is an algorithm based on fisheye vision multifunctional target. In terms of cameras, there are generally 4 fish-eye around, 2 side-view, 2 ADAS, plus the 2-megapixel rear view at the rear of the car, a car needs nearly 9 cameras to form the entire vehicle visual sensor layout.

The third step is the perception fusion of camera and ultrasonic radar. This requires two different transmission and image recognition methods, all data to be perceptually fused.

The fourth step, after the multi-sensing system is integrated, is the identification and algorithm of vehicle parking spaces, which requires a large amount of data and scene identification.

The fifth step is the algorithm function module of the entire parking control. Regardless of whether it is intelligent driving or autonomous driving, 100+ valid parking scene data accumulation is required, and more than 1 million on-site data training is required.

At present, the automobile industry has many whole machine factories with different system standards, and also has a very large automobile supply chain. The integration and coexistence of multi-sensing systems is the biggest challenge facing the development of automatic parking.

03 Do a good job in industrial "middleware" and create original multi-platform design

In addition to cameras, Sanyingxing also provides services for algorithms and pre-controller solutions. Conceptually, it is called multi-platform design. Currently, it is compatible with chips such as Black Sesame, Ambarella, and Horizon. The next step will be compatible with Huawei HiSilicon, Qualcomm, etc. Multichip platform.

Why do embedded system design? Li Zheng explained that Sanyingxing has always positioned itself as a provider of in-vehicle hardware and solutions, creating a middleware and algorithm container platform architecture that complies with the AutoSAR functional safety standard domain controller, which can be integrated with various excellent chips. In addition, from the bottom layer to the driver, and finally to the fusion, it will be compatible with all software and hardware, including body information.

Since 2017, the entire team of Sanyingxing Vehicle has been operating. Previously called Smart Imaging SID, Sanyingxing Vehicle Intelligent Technology was officially established in August last year, and it has begun to deliver products to domestic first-tier complete machine factories.

At present, Sanyingxing is moving from technology-driven to product-driven, strengthening the closed-loop construction of data. Because the entire intelligent driving and autonomous driving are driven by data, this step is an inescapable path for all OEMs and algorithm platforms.

Li Zheng said that in the future, Sanyingxing will achieve a closed loop in the development of perception fusion, map, regulation, algorithm container, and tool chain, and the final step is to realize a cross-domain and hybrid domain controller platform.

04 Mass production is king, and the monthly production capacity of in-vehicle cameras has reached 3KK

On the intelligent driving track, it's not about how dazzling your technology is, or how big your team and company are. The primary hard indicator is whether you can pass the OEM's audit of the supplier's delivery capabilities.

The second is the review of product production stability and consistency, then the production line equipment, and the last step is the technical review. Software delivery will actually be more difficult, because its verification is too difficult and there are too many bugs. It needs to be modified and modified at the same time. Therefore, we define software delivery at the same level.

For Sanyingxing, it currently hopes to cooperate with more domestic first-level central control panel and navigation hardware manufacturers, as well as Tier1 suppliers who have obtained software algorithms from domestic OEMs. Sanyingxing has the strength to be responsible for hardware template, project management and the entire production delivery.

SunWin Electronic has three production bases in Xianning, Tongcheng and India, with 38,000 square meters of constant temperature and humidity class 100 dust-free workshops, world-leading COB automatic semiconductor packaging production lines, automatic SMT production lines and industry-leading optical imaging experiments laboratory, material analysis laboratory, reliability laboratory. Among them, the third-phase project of Tongcheng is for the construction of vehicle supply. There are many automatic AA production lines for vehicle cameras, with a monthly production capacity of 300K.

In addition, SunWin Electronic has the best in-vehicle cameras and AA equipment both domestically and internationally. At present, the company adopts the AA automatic vehicle production line, which is at the same level as the camera production of Mercedes-Benz, BMW and Volkswagen. The equipment is basically compatible with the entire production and testing process of 1 million to 8 million pixels. In terms of hardware production capacity, technical testing, and product stability, the best domestic ADAS hardware can be mass-produced.

The ultimate dream of SunWin Electronic is to provide a complete solution for safe and reliable image acquisition for global artificial intelligence equipment.

Data source: Chaodian Vision and Perception​WeChat Public Account