The importance of in-vehicle cameras for autonomous driving
release time2017-06-04
ADAS system solutions include camera solutions, radar/lidar solutions, and sensor fusion. In the early stage of market development, as radar technology is mature and not affected by weather conditions, radar/lidar solutions are the mainstream of the market. However, with the development of ASIC (application-specific integrated circuit) and the improvement of image processing algorithms, and because radar technology is more accurate in distinguishing metal obstacles, it is unable to distinguish non-metal obstacles such as pedestrians and cannot accurately identify Vehicles approaching from the side and cannot distinguish lanes, debris or road potholes.
The visual processing technology of the camera can better distinguish information such as signs and pedestrians on the road, and can also calculate the movement trajectory of pedestrians and vehicles through algorithms. Compared with radar technology, the cost is lower, the function is more comprehensive, and the accuracy is higher. Camera-based imaging technology is gradually accepted by mainstream manufacturers. Considering the limitations of camera pixels on image recognition technology and the reduced functionality in extreme conditions such as fog and rain, camera-based sensor fusion will become the mainstream
From bottom to top, the Internet of Vehicles architecture is the perception layer, the network layer and the application layer, which are responsible for information collection, transmission, and processing functions. Video capture and storage (perception layer) is the underlying architecture of the Internet of Vehicles. The main technologies include vehicle-mounted DVR and vehicle-mounted IP Camera. Car DVR, commonly known as car video recorder, is based on the application of digital video compression storage and 3G wireless transmission technology, built-in GPS, car black box, CANbus bus, G-SENSOR and other technologies.
The in-vehicle IP Camera is based on digital signal processing technology (DSP) and network technology. The CMOS image sensor converts the light signal of the scene into electrical signals. These electrical signals are converted into digital signals and transmitted to the DSP memory through the data interface to complete image compression and encoding. At the same time, the data stream is sent to the hard disk or other storage devices for storage. It is different from traditional analog systems and DVRs in terms of distance, expansion capability and cost.
Vehicle-mounted cameras have a wide range of applications. According to the application fields, they can be divided into driving assistance (driving recorder, ADAS and active safety system), parking assistance (whole-vehicle surround view) and in-vehicle personnel monitoring (face recognition technology), which penetrates the vehicle The whole process from driving to parking, so there are higher requirements on the working time and temperature of the camera. According to the installation position, it can be divided into four parts: front view, rear view, side view and interior monitoring.