2022-05-19

Pls click or email to

mh_elec@126.com or jacky@mh-elec.com

In-vehicle infrared cameras, and most likely to have a place in autonomous driving perception systems. The infrared thermal imaging technology used was first applied to individual combat systems (infrared sights, night vision glasses), and after being converted to civilian use, it has made great achievements in industrial temperature measurement, temperature measurement in public places, forest fire prevention and other fields.

At present, some manufacturers are developing high-reliability and stable automotive-grade infrared camera products, in order to play their strengths in the field of autonomous driving perception sensors, improve the high-level autonomous driving perception system, and contribute to the early sunset of high-level autonomous driving. Power. Didi is the first company I have seen to plan the application of infrared camera products in high-level autonomous driving systems. It is currently unknown whether its products are vehicle-grade.

Working principle

Vehicle-mounted infrared cameras use infrared thermal imaging technology, which has been widely, fire protection, medical, industrial production, and customs. During the epidemic, the loyal guards who are responsible for temperature detection in public places will not let go of any beautiful blushing girl, nor will any young man try to hide the fart.

Infrared thermal imaging technology is a method of receiving infrared radiation from an object through a detector. Any object whose temperature is higher than absolute zero (-273°C) will emit infrared radiation. The infrared radiation of the object is focused on the detector through the lens. Photoelectric processing is performed on the received infrared radiation, and finally an infrared image visible to the human eye is generated.

The core of infrared thermal imaging technology is infrared thermal imaging detectors. At present, the countries that master the core technology of uncooled infrared thermal imaging detectors mainly include the United States, France, Israel and China. The U.S. is the dominant player in the world, and has an absolute leading position. China has developed rapidly in the past 10 years. Domestic companies have developed the world's first 8μm uncooled infrared thermal imaging detector chip (1920×1080).

Main players

The leading companies in infrared thermal imaging technology include SCD in Israel, ULIS in France, FILR in the United States, Raytheon, and DRS in the United States. The follow-up abroad can be introduced in a column, and this article will not list it for the time being. Here we focus on a Chinese company, MH Tech, which has developed automotive-grade infrared thermal imaging detectors.

Founded in July 2015, MH Tech is a wholly-owned subsidiary of MH Group. Mainly responsible for product development and sales of automotive infrared thermal imaging technology. At present, products with a car-grade 25μm pixel size and a resolution of 768x576 have been mass-produced, and products with a 17μm pixel size and a resolution of 768x576 are expected to be mass-produced by the end of 2021.

Some people may question why the resolution of the infrared camera is so low. In fact, it is not the case. Due to the different imaging principles, the infrared camera with a resolution of 640×512 has a very clear and delicate infrared image, and objects the size of a hare can still be distinguished at 500 meters. ; Infrared camera with a resolution of 384×288, 700 meters can still accurately identify the walking human body.

The size of the pixel mainly affects the size, resolution and identification distance of the imaging module. The smaller pixel size will further meet the industry demand for miniaturization and integration of thermal imaging module equipment. At the same time, under the same focal length optical system, the smaller the pixel size, the higher the spatial resolution and the farther the recognition distance.

Vehicle-mounted infrared cameras mainly have the following imaging characteristics:

passive imaging

All weather, unaffected by rain, snow and fog

All day, not affected by glare, backlight, no light

long detection distance

Therefore, it is preliminarily judged that the vehicle-mounted infrared camera will definitely solve the pain point of identifying long-distance obstacles in bad weather and bad lighting conditions.

Evaluation results

Occasionally, we got an infrared camera product with a resolution of 640x512 produced by MH, and conducted an evaluation of the recognition effect of a typical application scenario, and compared it with a visible light camera with a resolution of 2 million and a resolution of 1920x1080. , which is used to evaluate its ability to improve autonomous driving perception in complex environments, and now select several typical scenarios for analysis.

1. Strong backlight environment

For visible light cameras, there will be a short blinding time when entering and exiting the tunnel. During this short blinding time, the recognition of lane lines and traffic signs will fail. We simulate entering the tunnel by testing the interior of the warehouse under strong light.

Judging from the test results, the visible light camera can no longer see the internal structure of the warehouse at this time, while the infrared camera can clearly see the internal structure.

2. Alternating light and dark environment

There are generally street lights on highways or urban main roads. A small amount of lights has very high requirements for the high dynamic performance of visible light cameras. From the point of view of the vehicle-mounted visible light cameras that have been in contact, the effects of identifying lane lines and traffic signs in this scene are excellent enough. Now Let's take a look at the effect of the infrared camera in identifying lane lines in this scenario.

The visible light camera recognizes the lane line very clearly, and the infrared camera also recognizes the lane line, but also recognizes the abnormal lane. The upper red line surrounds the sewer, which is identified as an image that approximates the lane line. In the lower right corner is the old lane line that has been completely erased, which is basically invisible to the human eye, but it was discovered by the infrared camera. The erasing and redrawing of lane lines in the city is not uncommon. How the infrared camera eliminates interference and prepares to identify the lane lines is a very test of the algorithm's ability.

3. In total darkness

In the dark environment, the human eye and the visible light camera are invalid. We tested the recognition effect of the infrared camera in the dark environment.

It may not be very obvious in the picture, but from the actual verification, the effect is very amazing. I only saw the person riding the battery car at about 10 meters, but at a distance of 100 meters, it can be seen from the infrared camera. to this cyclist who is slowly driving in.

At last

Vehicle-mounted infrared cameras will surely become an important member of the all-weather, all-day fully autonomous driving perception system, and become a powerful weapon to solve the long-tail problem of autonomous driving, but there is still a long way to be explored before it truly becomes famous.

AI algorithms adapted to the imaging characteristics of infrared cameras, including infrared and visible light fusion, true color conversion, and monocular ranging algorithms;

Automotive-grade infrared camera products with high resolution and small micro-element size;

It is in line with the cost of front-loading mass production.