A few days ago, Xiaopeng Automobile released a demonstration video of the parking visual recognition system equipped in its upcoming Xiaopeng P7 model. It can be seen that the system can use the unique side-forward camera on the vehicle to discover the parking space in front of it in advance and draw a map of the parking lot. The recognition rate is high for both normal and side parking spaces.

Based on the video we can briefly analyze and analyze, in accordance with the computer simulation screen in the figure, you can see that the P7's auto-parking perception ability range is larger, and change the G3 need to drive through the parking space, the vehicle can only sense the parking space. the P7 is optimized to be able to automatically capture the forward-facing parking spaces to facilitate the vehicle to automatically make a decision, and the forward detection of more than three normal standard parking space location.

Trajectory tracking capabilities, from the video of the simulation screen to see, an orange track route automatically marked the vehicle driving track, this track recording capabilities, help P7 record the driving route in this garage. In the future to trigger the automatic valet parking in this environment, according to this trajectory driving. In addition, it can also be triggered by the user to summon the vehicle, in accordance with this track to return to the user to get off the location.

Garage mapping ability, from the computer simulation screen and the real car screen can be seen, the computer simulation map basically restore the actual location of the garage, P7 will be combined with high-precision maps and big data technology, the parking lot of the location of the mapping management, when the owner of the car into the garage, the P7 will be automatically pulled out of the original mapping, to improve the accuracy of the parking.

Limit parking space recognition ability, in the figure you can see the right parking space is a dead-end parking space, the recognition of the parking space in accordance with the previous drive through the ability to recognize the judgment, this dead-end parking space can not be identified. But in the P7 computer simulation, the space is perfectly recognized, and the computer simulation shows green, the system is judged to be able to automatically park.

So in some unlit parking spaces can be recognized, from the video in some of the lighting lighting is not enough parking spaces, P7 still able to complete the identification of all the parking spaces, reflecting the P7 strong perceptual ability, the utility is greatly improved.

Underground garage network signal is not good and how, P7 to take a weak network dependence of the parking work strategy, in the underground garage this network signal coverage is poor, the vehicle basically rely on their own camera and the rest of the sensory equipment for visual fusion perception, to avoid network signal dependence, to increase the reliability of the automatic parking.

In addition, Xiaopeng P7 already has the possibility of autonomous parking, autonomous parking of the engineering argument is to establish a three-dimensional model of the parking lot through the SLAM, drawing three-dimensional map, Xiaopeng P7 can be added through the new side of the forward-facing camera to assist in drawing, which is the basis for the realization of the autonomous parking later, I believe that the subsequent upgrades of the Xiaopeng P7 will be a more pleasing change.

From the technical specifications given by the official point of view, Xiaopeng P7 has the strongest assisted driving hardware architecture in the industry, using NVIDIA's latest self-driving chip Drive?Xavier, the vehicle is equipped with 13?cameras, 12?ultrasonic sensors, 5 Bosch fifth-generation millimeter-wave radar and high-precision maps + GPS positioning combination of the ?L3?level of automatic driving hardware

This article comes from the author of Automotive Home Car, and does not represent the viewpoint position of Automotive Home.