Multi-sensor fusion method for parking space detection

Abstract

The invention discloses a multi-sensor fusion method applied to parking space detection. Combined with an environment perception module, various parking space detection scenarios are pre-divided and the working modes of various sensors are determined. According to the working modes of the sensors, the various sensors are combined into There are various working strategies for parking space detection. In the actual detection stage, the environment perception module can be used to determine the target type of the parking space detection scene corresponding to the current scene, and then according to the target type, a multi-sensor fusion work strategy suitable for the current scene is selected. , to accurately identify and output the parking spaces in the current scene. The present invention can cover multiple types of parking scenarios, so that the automatic parking system has fewer conditional restrictions, and can improve the practicability, especially compared with the full sensor fusion with extremely high redundancy, the present invention divides different scenarios and selects The maneuvering strategy of sensor fusion can significantly save computing power, thus ensuring the real-time performance of the automatic parking system.

Description

technical field [0001] The invention relates to the technical field of automatic parking, in particular to a multi-sensor fusion method applied to parking space detection. Background technique [0002] In recent years, with the continuous development of autonomous driving technology, the cost of on-board sensors has gradually decreased, and the hardware configuration of autonomous vehicles is first redundant and then integrated. On the basis of the number of sensors and the performance upgrade, gradually realizing multi-sensor fusion is an important strategy to improve the safety and redundancy of autonomous driving. [0003] Parking space detection is a very important part of the automatic parking system. At present, the relatively mature mainstream solution for parking space detection is a pure vision solution based on the vehicle surround view camera. The parking space is identified by detecting the parking space corners and parking space lines on the 360 ​​panoramic imag...

AI Technical Summary

Problems solved by technology

[0006] (1) The parking space detection method based only on the ultrasonic sensor cannot obtain the semantic information of the parking space line, and only provides the space information for parking

Such a parking space detection scheme cannot accurately identify the parking space in a relatively open area, and it is difficult to accurately and automatically park the vehicle in the parking line, often failing to achieve the desired parking effect

At the same time, due to the loss of visual information, the user has no direct experience of the parking space in the car, and the user needs to judge the provided parking space through the external environment

[0007] (2) The parking space detection method based only on the camera is greatly affected by the lighting environment. Low light conditions or the road surface of the indoor parking lot with serious exposure will affect the recognition of the parking space, resulting in the failure to recognize

It is very difficult to identify non-standard parking spaces, blurred parking spaces, and parking spaces without parking spaces, and the requirements for the computing power of the controller are also high, which means that as the complexity of the model algorithm becomes higher, satisfying Real-time requirements will also be difficult

[0008] (3) At present, the cost of lidar is relatively high, and the application of lidar mostly exists in experiments and some special vehicles, and the advantages of distance and accuracy of lidar make it much more valuable in driving scenes than in parking scenes , the automatic parking solution relying only on lidar is still difficult to popularize

[0009] (4) Obtaining a parking space through interaction with the parking space monitoring system in the parking lot requires high external support, while the automatic parking system recognizes the parking space as the basis for automatic parking, and automatic parking can only be realized in special scenarios. Practical application significance, and it is also difficult to popularize

Images