Car stuff (public number: chedongxi)
Author | Little Cui
Edit | Juice
Chedong news on February 27, according to foreign media Telsarati, recently, Tesla obtained two technology patents, which focus on pure vision systems, or will improve the level of full automatic driving.
The two patents, which focus on virtualized displays and machine learning, appeared in the U.S. Patent Office's database last week, and it's unclear whether the two patents have been applied in upcoming software versions or have been installed in cars with recent software updates.
In addition, Tesla recently launched a very significant improvement in full self-driving technology, namely FSD Beta V11.3, and the new patented technology also plays an important role in the continued improvement of FSD Beta V11.3.
1. Add adjustable virtual camera The car page is more accurate
The first patent concerns a virtualized imaging system that removes Tesla's original non-visual sensors, and the biggest difference from Tesla's existing imaging system is the addition of an adjustable virtual camera that provides a more accurate page of the vehicle.
▲Tesla car interface
The patent is a vision-based machine learning model system and method that can be used for autonomous driving with adjustable virtual cameras.
During driving, the system can acquire images from multiple image sensors around the car, determine the features associated with the images, and then output the features directly through the machine learning model. Finally, these features are projected into the space associated with a virtual camera of a specific height, and these features are aggregated with some other projected features before them, and finally displayed on the screen.
Through camera systems or improved virtualized interaction systems, drivers can better understand the view outside the car.
Second, aggregate vehicle static object monitoring Machine learning models are more perfect
The second patent is to improve the machine learning model by monitoring the static environment of the road, and its application will be more extensive.
▲Machine learning models
Similar to the first patent, the system acquires images from image sensors around the vehicle, determines the features associated with the image, and outputs the features directly through the machine learning model.
The difference is that the application of this new technology can make the system's monitoring of static objects more accurate, and these features are finally projected into the space associated with the bird's-eye view.
It can be found that Tesla's two new patents are focused on the vision system of the car.
Conclusion: Committed to artificial intelligence, Tesla's lead is more obvious
Tesla's increasing focus on vehicle intelligence is reflected in their various machine learning models and methods for implementing improvements.
While many ridicule Tesla's commitment to artificial intelligence, not just cars, Tesla's lead will become more pronounced as more automakers enter the race for self-driving.