The Importance of Vehicle-Mounted Cameras for Autonomous Driving
ADAS system solutions include camera-based solutions, radar/LiDAR solutions, and sensor fusion. In the early stages of market development, radar/LiDAR solutions were the mainstream choice due to the maturity of radar technology and its immunity to weather conditions. However, with the advancement of ASICs (Application-Specific Integrated Circuits) and improvements in image processing algorithms, along with the fact that while radar technology has high accuracy in detecting metal obstacles, it is ineffective at identifying non-metal obstacles such as pedestrians, cannot accurately detect vehicles approaching from the side, and is unable to recognize lane markings, debris, or road potholes.
Camera-based visual processing technology can better identify road signs, pedestrians, and other information on the road. It can also calculate the movement trajectories of pedestrians and vehicles through algorithms. Compared to radar technology, it offers lower cost, more comprehensive functionality, and higher accuracy. As camera-based imaging technology becomes increasingly accepted by mainstream manufacturers, and considering the limitations imposed by camera pixel count on image recognition technology as well as reduced performance under extreme conditions such as fog and rain, camera-dominant sensor fusion is expected to become the mainstream approach.
The IoV (Internet of Vehicles) architecture consists of three layers from bottom to top: the perception layer, network layer, and application layer, which are responsible for information collection, transmission, and processing respectively. Video capture and storage (perception layer), as the underlying architecture of the IoV, primarily involves technologies such as vehicle-mounted DVRs and vehicle-mounted IP cameras. A vehicle-mounted DVR, commonly known as a vehicle video recorder, is based on digital video compression storage and 3G wireless transmission technology, incorporating built-in GPS, automotive black box functionality, CAN bus, G-SENSOR, and other technologies.
In contrast, vehicle-mounted IP cameras are based on Digital Signal Processing (DSP) technology and network technology. A CMOS image sensor converts the optical signals of a scene into electrical signals, which are then converted into digital signals and transmitted to the DSP memory through a data interface. This process completes image compression and encoding while simultaneously sending the data stream to a hard drive or other storage devices for preservation. These systems differ from traditional analog systems and DVRs in terms of range, scalability, and cost. Vehicle-mounted cameras have a wide range of applications. Based on their application areas, they can be categorized into driving assistance (dashcams, ADAS, and active safety systems), parking assistance (surround-view systems), and in-cabin occupant monitoring (facial recognition technology). These applications cover the entire process from driving to parking, placing high demands on camera operating time and temperature tolerance. Based on installation position, they can be further divided into four categories: front-view, rear-view, side-view, and in-cabin monitoring cameras.