1 Introduction
As the importance of automobile safety has become more prominent, Driving Assistance System (DAS) has become more and more widely used in the automobile field. In such applications, the graph vision system is an indispensable component. It must quickly and accurately perceive various road conditions, and then advance warning or control vehicle behavior in advance. This requires the front end of the system-the camera can adapt to different lighting conditions to the greatest extent and provide detailed information on the road scene.
Figure 1 shows the main applications of automotive vision systems, including front view, rear view, side view, and interior view. Formerly regarded as an example, applications such as Night Vision (NV), Lane Departure Warning (LDW), Collision Warning (CW), Traffic Sign Recognition (TSR), etc. Under various weather and road conditions, it can clearly identify lane lines, vehicles, obstacles, traffic signs, etc. All these functions require that the dynamic range (DR) of the image sensor is high enough to provide rich detailed information for the post-level image processing system.
This article mainly discusses the application of image sensors in automotive vision systems. The second part introduces the requirements of the automotive vision system on the performance of the image sensor, and the third part discusses the high dynamic range image sensor technology. The fourth part summarizes this article.
Main applications of automotive vision systems
2. Image sensors in automotive vision systems
An image sensor is a device that converts an optical image into an electronic signal and is an important part of a camera, as shown in Figure 2. In a typical automotive application, the vision system has to deal with some complex road conditions, such as driving towards the sun, entering and exiting the tunnel, and headlights of the opposite vehicle when meeting at night. Such road conditions require the camera and the image sensor in it to automatically adapt to such complex scenes, otherwise the loss of detailed information in the image is difficult to recover through post-level signal processing methods.
Image sensor in camera
In a picture, the ratio of the brightness of the brightest part to the darkest part is called the dynamic range. For example, the brightness of a bright object under direct sunlight is about 100kcd / m2, and the brightness of a dark object under shadow is about 100mcd / m2. If a scene includes both of these objects, the dynamic range of the image is 120dB. Only if the dynamic range of the image sensor is higher than 120dB, the car vision system can effectively capture images of the scene.
Considering driving at night, some roads have no street lights or the street lights are very dark, and the car vision system also needs to collect road condition information in low-light scenes. For automotive applications, image sensors need to meet a series of automotive-grade requirements, such as wide temperature range, high reliability, high integration, and low cost. Table 1 summarizes the performance that image sensors in automotive vision systems should have:
Automotive vision system requirements for image sensors
3. Melexis high dynamic range image sensor technology
Due to the high integration, high stability, and low cost of CMOS technology, image sensors based on CMOS technology have become more and more widely used, and have become the first choice for most vision systems. The pixels of the CMOS image sensor contain a photodiode, which generates a photocurrent that is positively related to the light intensity under the excitation of light. By integrating the photocurrent within a certain exposure time, the optical image signal can be replaced with the electronic signal output of the pixel, and its response characteristic is shown in FIG. 3 (a). Because of the limitation of the supply voltage amplitude and sensor noise, the dynamic range and imaging details of the traditional CMOS image sensor are a contradiction: increasing the exposure time can improve the details but at the same time reduce the dynamic range, otherwise reducing the exposure time can increase the dynamic range but not conducive to the details . This major contradiction in traditional CMOS image sensors makes it impossible to fully meet the requirements of automotive vision systems.
In response to the characteristics of CMOS technology, Melexis has developed a CMOS image sensor MLX75412 based on Autobrite technology, which solves the aforementioned contradiction with the non-linear response characteristics shown in FIG. All standards listed. Through Autobrite technology to carry out segment exposure / reset control on pixels, the dynamic range of the image sensor is extended to 154dB. The integrated feedback control algorithm analyzes the collected images, adjusts the sensor's exposure time and dynamic range in real time, and significantly improves the quality of image acquisition under complex road conditions.
MLX75412's integrated image enhancement function, Autoview, can adjust the image brightness and sharpness in real time, reducing the amount of calculation in the post-level signal processing system. At the same time, because of the 12-bit image processing pipeline integrated in the MLX75412, its image processing accuracy is higher than the commonly used 8-bit image processing. MLX75412 provides standard raw video signal output and I2C control signal input, compatible with most DSP or FPGA. Considering automotive applications, the MLX75412 provides 1024 × 512 resolution widescreen images to capture a wider range of images, making it easy to identify pedestrians or traffic signs on both sides of the road.
Response characteristics of CMOS image sensor: (a) traditional linear curve; (b) dynamic range expansion curve based on Autobrite® technology
Figure 4 compares the imaging effects of a common CMOS image sensor and an image sensor based on Autobrite technology. In Figure 4 (a), because the dynamic range of the sensor cannot effectively adapt to the dynamic range of the scene, the details of the dark and bright areas of the image are poorly reflected. In contrast to this, by using Autobrite technology, the image shown in Figure 4 (b) has very good details for both bright and dark areas. For applications such as lane departure warning and obstacle recognition, the details in high dynamic range images can greatly improve the accuracy of driving assistance systems.
Autobrite® technology imaging effect
4 Conclusion
This paper analyzes the demand for image sensor performance in automotive vision systems. The work of the automotive vision system depends on its front end, the ability of the camera to collect image information, and this is largely determined by the image sensor in it. In order to collect scene details under complex road conditions, the sensor must have high dynamic range and high light sensitivity. Considering automotive applications, it must also have high reliability, high integration, and high flexibility.
Melexis MLX75412 is a high dynamic range CMOS image sensor for automotive vision systems based on Autobrite technology. Autobrite technology can collect high-quality images under complex road conditions with extremely different brightness, and provide accurate and reliable scene details for driving assistance systems, thereby improving driving safety.
Slimming Machine,Ems Sculpting Machine,Ems Sculpting
Xi'an Double H Health Technology Co., Ltd , https://www.xadoubleh.com