In the past few years, more technology developers in the field of autonomous driving have discussed sensors such as cameras, ultrasonic/millimeter wave radars, and lidars, and debates about how to select and configure these sensors.
Regarding the infrared night vision sensor, it is rarely mentioned. But after Uber's self-driving vehicle was involved in a fatal accident in Arizona, it seems that the industry is starting to re-examine this issue.
Including FLIR Systems, AdaSky, and Chinese company Xuanyuan Zhijia (a subsidiary of AutoNavi Infrared) and other manufacturers of in-vehicle infrared night vision devices, they have once again become the focus of the industry, although infrared night vision devices have been in stock in the past few years. Low-volume configuration on production cars (BMW, Porsche and other luxury cars).
The leading autonomous driving company represented by Waymo has been working hard on lidar before and even started to develop lidar independently.
On January 8th last year, Waymo CEO Krafcik announced that the cost of lidar, a key component of its autonomous driving system, has fallen by 90% compared to when the project started, from more than $70,000 to 7,500. Dollar.
But obviously it cannot benefit the entire industry unless Waymo announces the sale of its own research and development lidar. At the same time, in the past two years, the price of lidar, which has fallen sharply, is still expensive for most market applications. For example, the cost of Velodyne's HDL-64E laser laser system is still about $80,000.
Although, Velodyne announced at the beginning of this year: The global price of its 16-line lidar product VLP-16 Puck has dropped by half, from the previous price of approximately US$7,999 to US$3,999. But this is far from the "sweet spot" of the pre-installed mass production price of $300.
In addition, one car still needs to be equipped with multiple lidars, and the overall cost is still high.
Of course, since last year, more solid-state lidar startups announced that they have developed cheaper lidar technology, which is smaller, and the price may drop to less than 100 US dollars within a few years. But this is still unknown.
More industry insiders said that according to the current development process of the lidar industry chain, lower prices in the short term inevitably mean performance trade-offs, such as lower resolution and shorter ranging range.
In addition, lidar is not adapted to all-weather scenes like millimeter-wave radar. Weather such as fog, rain and snow will seriously affect accuracy, not to mention that there is no consistent standard for lidar in the world so far.
For infrared night vision device manufacturers, they also have enough evidence to explain the shortcomings of the existing sensor combination.
Taking Tesla’s Autopilot as an example, the perception system uses a collaborative work method between a camera and a millimeter wave radar to determine obstacles in front (for example, the difference between a plastic bag and a puppy).
However, under low illumination, the camera cannot work normally, which is the Achilles heel of the system. The current millimeter-wave radar technology has not yet been able to recognize obstacles. They can only provide the distance and speed of the obstacle.
It is worth noting that in the past cases of traffic accident statistics, the number of fatalities in the accident rose fastest after sunset. In fact, the vast majority (three-quarters) of pedestrian deaths also occur at night.
It has been confirmed that the infrared night vision system uses infrared light waves to detect the difference in heat emitted by objects naturally, and can detect objects that cannot be identified by visible light cameras, radar, and lidar. It is crucial that they perform well in low light and bad weather.
At present, BMW is one of the earliest manufacturers to introduce night vision systems in terms of vehicle mass production configuration. Since March 2006, the personalized versions of BMW 760Li and BMW 760Li sold in the Chinese market have been equipped with the new BMW night vision system.
The camera of the BMW night vision system has a wider detection angle and a longer detection distance (detection distance of about 300 meters) than other systems. When the vehicle speed is less than 80 km/h, the 36° horizontal wide angle of the thermal imaging camera can cover a wide detection range.
At moderate speeds, the image area on the display screen can cover the 24° range of the scene in front of the vehicle, and the image area can also be rotated around 6° as the road turns. At higher speeds, the driver can activate the digital zoom function, which in turn enlarges the image of objects at a greater distance by 1.5 times.
The Audi A7 launched in 2012 also began to be equipped with far-infrared imaging cameras. The night vision assistance system with pedestrian recognition function can capture heat sources within 24 degrees and 300 meters ahead. When the heat source appears within the capture range, it will be processed by computer. The image of will be displayed on the dashboard in black and white contrasting colors.
Last year, Xuanyuan Zhijia, a subsidiary of the listed company AutoNavi Infrared, released a new generation of thermal imaging obstacle avoidance system IR313 "based on AutoNavi Infrared Autonomous Detector". Pedestrian recognition and warning, front vehicle collision warning and warning, and many other functions.
Xuanyuan Zhijia's infrared obstacle avoidance system uses infrared thermal imaging technology, which can capture a front view of 28°*21° and a heat source 400 meters away. This product can detect night vision distances of more than 400m, while most vehicle night vision systems on the market have detection distances of about 200m. At the same time, the system will automatically adjust functions according to different vehicle speeds.
At the same time, the system is fully designed in accordance with vehicle regulations and has passed the certification of relevant agencies. The night vision camera has an integrated intelligent image recognition function, without an external image processing box, which is easy to install. The system integrates the pedestrian recognition algorithm function, and has passed long-term, systematic and strict road test verification.
In addition to the cost reduction of the night vision system itself, how to deeply integrate with the current multi-sensor of ADAS and autonomous driving systems, and deep learning based on night vision, is also an opportunity to accelerate the introduction of large-scale markets.
Just recently, FLIR, a global automotive night vision thermal imaging system supplier, released a free thermal image data set for collision avoidance systems, which is also the world's first.
The system trained with this data set can detect objects 250 meters away, or 4 times the distance from the headlights of a car. The data set contains day and night shots in fog, smoke, haze, sun glare, and other lighting conditions in the environment, including pedestrians, animals, bicycles and other vehicles.
In the past few years, more than 500,000 infrared night vision sensors have been deployed in the driver warning systems of General Motors, BMW, Audi, Mercedes, Benz and Volkswagen. At the same time, a software development kit is also provided to build and train neural networks for use in collision avoidance systems.
High Current Terminal Blocks,Panel Terminal Block,Feed Through Terminal Block,Heavy Power Terminal Block
Sichuan Xinlian electronic science and technology Company , https://www.sztmlchs.com