Vehicle mounted image sensor chip

According to installation location and functional differences, in vehicle CMOS image sensors can be divided into two major application areas: exterior and interior, covering comprehensive needs from active safety to passenger management. With the improvement of intelligence, the number and complexity of sensors have significantly increased, forming a three-dimensional perception network.

Extravehicular application: Building a vehicle's "external visual network" forward-looking perception CIS. Behind the front windshield or on both sides of the roof, some models adopt a split binocular layout. Implement core ADAS functions such as Forward Collision Warning (FCW), Lane Departure Reminder (LDW), Traffic Sign Recognition (TSR), and Pedestrian Detection (PD) to provide basic data for L2+autonomous driving. From monocular to binocular and trinocular systems, different combinations of field of view (FOV) are used to achieve both near and far balance. For example, the main field of view camera (FOV of about 50 °) covers a distance of 100 meters, while the narrow field of view camera (FOV of about 25 °) extends to 200 meters, improving reaction time in high-speed scenes.

Around view imaging CIS, near the front and rear logos, under the left and right exterior mirrors, some models add four side fender cameras to form a 360 ° panoramic view system (AVM), and some high-end models expand to six cameras, supporting the "transparent chassis" function. Support automatic parking (APA), blind spot monitoring (BSD), narrow road traffic assistance, and generate bird's-eye view images through image stitching to eliminate visual blind spots.

Rear view imaging CIS, located above the rear license plate frame or integrated into the rear spoiler. Reverse Vehicle Assist Display (RVC), supporting dynamic guide lines, marking obstacle distances and trajectories, reducing parking risks, resolution: upgraded from VGA (640 × 480) to 2MP (1920 × 1080), some models adopt Wide Dynamic (WDR) technology;

Side view perception CIS, located below the exterior rearview mirror or at the door handle, some models are integrated into the B-pillar. Blind spot monitoring (BSM), lane change assistance (LCA), door opening warning (DOW), real-time monitoring of vehicles and pedestrians on the side and rear. Resolution: up to 3MP, supports wide field of view (FOV ≥ 120 °);

Streaming media/electronic rearview mirror CIS, captured by rear and side cameras, compressed and transmitted to the high-definition display screen inside the car. Anti strong light, weak light compensation, rain and fog penetration, no blind spots in the field of view, especially significantly improving safety in rainy or nighttime weather.

Driving recorder type CIS, located on the inner side of the front windshield, often shares modules with front ADAS or streaming rearview mirrors. Record the driving process for accident tracing, driving behavior analysis (such as rapid acceleration and braking statistics), and insurance liability determination. Resolution: mainly 1MP, gradually transitioning to 2MP;

Driver Status Monitoring (DMS), Occupant Monitoring System (OMS), Rear Passenger Detection (child legacy reminder), Seat Belt Status Recognition, Passenger Posture Analysis (such as whether the seat belt is fastened, abnormal behavior warning), Emotional Interaction (adjusting the interior atmosphere based on facial expression recognition).

High Dynamic Range (HDR) has become a standard feature, and the vehicle's operating environment has undergone drastic changes in lighting conditions (such as tunnel entry and exit, night high beam). HDR ≥ 120dB has become the mainstream requirement, and some high-end products have reached 140dB. Technologies such as multi frame synthesis and dual exposure further expand the dynamic range, eliminating "overexposure" or "dead black" phenomena.

The ability to suppress LED flicker (LFM) is crucial in dealing with high-frequency flashing light sources such as traffic signals and electronic displays (50Hz~120Hz) to avoid image distortion and recognition errors. Special pixel design is adopted in hardware, and precise suppression is achieved through frequency synchronization and phase compensation algorithms in software.

Functional safety and reliability, all critical applications must meet the ISO 26262 functional safety standard, especially in ADAS related scenarios where chips need to have fault detection and redundancy mechanisms. For example, adopting a dual core architecture, built-in ECC verification, temperature monitoring, etc., to ensure that single point failure does not affect system operation.

Continuous optimization of low light and night vision performance, improved photosensitive efficiency through back illuminated (BSI) and stacked processes, combined with on-chip noise reduction algorithms such as 3DNR and TDNR, to achieve starlight level imaging (≤ 0.1lux). Part of the high-end chips integrate near-infrared enhancement function, which achieves ambient sensing without light through 940nm infrared fill light.

The trend towards integration and intelligence,

The chip gradually integrates ISP and AI preprocessing units (such as convolution accelerators) to reduce external computing power dependence, lower system latency and costs. For example, the built-in object detection module can output ROI regions in real-time, improving the efficiency of downstream algorithms.

Anti interference and stability

There is strong electromagnetic interference (EMI) in the car environment, and the chip needs to pass strict EMC testing, using metal shielding packaging and anti noise circuit design. At the same time, it is necessary to have high seismic resistance (such as meeting the ISO 16750 vibration standard) to ensure stability under extreme road conditions.

Continuous improvement in resolution, expansion of perception capability boundaries, multimodal fusion perception, and construction of redundant secure network CIS will be deeply integrated with millimeter wave radar, LiDAR, and ultrasonic sensors. Through spatiotemporal synchronization and data fusion algorithms, the shortcomings of a single sensor (such as CIS being affected by weather and LiDAR being costly) will be compensated for. For example, in rainy and snowy weather, supplementing visual perception blind spots with radar data.

The trend of AI on sensors is deepening, and edge intelligence is rising. Chips are embedded with lightweight AI engines (such as convolutional neural network accelerators) to achieve edge intelligence. For example, target detection, classification, and tracking are completed on the chip side, and only key information is transmitted to the central computing platform, reducing communication latency and bandwidth requirements. Meanwhile, privacy data can be processed locally to comply with regulatory requirements.

With the acceleration of domestic substitution and the coordinated upgrading of the industrial chain, under policy promotion (such as the "Intelligent Vehicle Innovation and Development Strategy") and industrial chain coordination, domestic CIS will continue to catch up in performance, reliability, ecological adaptation, and other aspects. Upstream equipment and material (such as lithography machines and target materials) enterprises are accelerating breakthroughs, midstream manufacturing processes are promoting the construction of automotive grade production lines, and downstream host factories are opening up cooperation to form a healthy ecosystem.

The integration of intelligent cockpit and vehicle networking expands application scenarios, and the in cabin CIS will not only be limited to safety monitoring, but also become an intelligent interactive entrance. For example, through passenger expression and behavior analysis, dynamically adjust the air conditioning temperature, music style, and seat posture; Combining the Internet of Vehicles to achieve remote identity authentication and personalized service push.

Cost and power optimization, promoting large-scale applications, integrating advanced processes (such as 22nm/14nm) with SoC to reduce chip costs and power consumption, gradually extending high-end functions to mid to low end models. For example, integrating ISP, AI units with CIS on a single chip can reduce the demand for peripheral devices.