Automated Parking Self-Localization via 3D Virtual Viewpoints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing path planning methods for automatic vehicle parking are limited by their reliance on specific parking scenarios and require the vehicle to be positioned near the learned path, failing to generalize to various real-life situations due to the large-baseline matching problem in self-localization, which restricts the usability of automatic parking systems.
Innovation Solution
The method generates a 3D model of the parking environment and creates virtual viewpoints to synthesize virtual images from different observation angles, allowing for feature matching and self-localization even when the vehicle is not on the initial path, thereby expanding the applicability of automatic parking systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If image-based 2D feature matching is used for self-localization in automatic parking, then the system can achieve automatic parking along the learned path, but the system fails when the vehicle is located away from the learned path due to appearance changes of landmarks from different viewpoints
Solution Approach 1:
The patent transforms 2D image features into 3D geometric features by constructing a 3D map of the parking environment. This dimensional transformation allows the system to represent landmarks in a viewpoint-independent manner, resolving the contradiction between reliable self-localization and adaptability to different parking scenarios. The 3D geometric features remain consistent regardless of the observation viewpoint, enabling accurate self-localization even when the vehicle is away from the learned path.
2Extent of automation
If the automatic parking system relies on matching image features from the learning stage, then it can achieve automatic parking in specified spaces, but it requires the vehicle to be near the learned path which limits its usability
Solution Approach 1:
The patent changes the feature representation parameters from 2D image-based features to 3D geometric features. This parameter transformation makes the features invariant to viewpoint changes, allowing the automatic parking system to work from any starting position rather than requiring the vehicle to be near the learned path. This resolves the contradiction between automation capability and ease of operation.
3Productivity
If a two-staged automatic parking system learns parking paths during a learning mode, then it can assist parking along the learned path, but it cannot handle arbitrary parking paths where the vehicle starts from different locations
Solution Approach 1:
The patent creates a 3D copy or model of the parking environment during the learning stage, which can then be used for self-localization in the automatic parking stage regardless of the starting position. This 3D model serves as a viewpoint-independent reference that enables the system to handle arbitrary parking paths, resolving the contradiction between parking assistance efficiency and adaptability to different scenarios.
Data Source
AI summary
A system controls motion of vehicle according to a first parking path ending at a target state and determines a second parking path from a current state to the target state using a model of the parking space. The system acquires images of the parking space during the motion of the vehicle along the first parking path and constructs the model of the parking space. The model is used to generate a set of virtual images of the environment of the parking space as viewed from virtual viewpoints outside of the first parking path. The current state of the vehicle is determined by comparing a current image of the parking space with at least one virtual image. The second parking path is determined from the current state to the target state.