Cars can also have the ability to perspective See how Koreans do it.-shiyang

By , April 16, 2018 10:23 pm

Cars can also have the ability to perspective? Look at how the Korean people do [Abstract]KAIST team on the front and rear of the two cars are equipped with a camera, the car can be seen in front of the car screen view of the car in the field of information. Remember the beginning of Samsung launched the car has four screens of the safety of the truck? Through the installation of the camera in front of the screen, the rear of the car can "perspective" truck, so as to understand the road ahead when overtaking, reduce traffic accidents. In this year’s IROS 2016, KAIST Korea Institute of science and technology of robotics and Computer Vision Laboratory also demonstrated a similar technology, but they did not use the screen this rude way, but really let the car have the ability to view. In simple terms, the KAIST team before and after the two cars are installed on the camera, see the car in front of the information through the WiFi network to the car, after the car and placed in a proper position in the field of vision, so that you can see the car in front in the field of information on the screen, the relevant papers have been published. In a yellow marker is actually on the yellow marker gray KIA map is actually KIA gray forward this technology is particularly useful in overtaking, presumably older drivers have experienced before the big truck in the way, but dare not overtake it. Although the process is simple, it is not so easy to achieve. In this technology, the front car is equipped with two cameras, is responsible for the acquisition of 3D image of the scene; after the car has a camera, is responsible for estimating and tracking the relative position of the front car. Before the car has two cameras rear camera, so that the image collected before the car can be projected to the rear of the car through the relative position of the line of sight. As long as there is a camera pose (pose) and the 3D data of the scene, any AR display is just a projection. Nanyang a study on visual aspects of a person told reporters, here is the most difficult time, one is the accuracy of camera pose estimation for the latter, there is currently no real good real-time estimation algorithm. And this has been confirmed by the KAIST team, they said that the realization of SLAM in the urban environment, as well as the real-time positioning of the car is very difficult. One of the authors of the Hyowon Ha told Lei Feng, the main difficulty in the development of algorithms and networks. After the car in the environment to achieve continuous changes in positioning, in order to put the image of the car in front of the vision, which is a great challenge. On the network, and now they are using a car to car (V2V) WiFi network transmission of images, the future will use 5G technology. At present, KAIST has been able to do in real time after the car before the car to see the image display, video frame rate of about 15 frames per second, the delay is less than 100ms. They also do experiments in the KAIST Park, and in Germany to do DEMO. Francois, another author of the paper, said that there is now a German company in cooperation with the team, considering the technology used in the actual scene in Rameau. He also said that the future will enhance the stability of the system to adapt to a variety of weather conditions, lighting conditions and time, in addition to perspective, but also to join the scene.相关的主题文章:

Comments are closed