Parking Zone Localization Using Mixed-Domain Mapping and Altitude Compensation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current parking assist systems face challenges in accurately mapping and localizing vehicles in parking zones, especially in environments with weak GPS signals or multiple floors, leading to inefficiencies and fuel consumption due to repeated searching for parking spots.
Innovation Solution
A mixed-domain neural network approach using image, bird's-eye-view (BEV), and tri-perspective view (TPV) domains, combined with inertial measurement units (IMUs) for altitude compensation, to create detailed 3D maps and localize vehicles within parking zones.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional GPS-based localization is used in parking zones, then the system is simple to implement, but localization accuracy deteriorates in environments with weak GPS signals or multiple floors
Solution Approach 1:
The system segments the localization problem into multiple domains: image domain for feature detection, bird's-eye-view domain for top-down positioning, and tri-perspective view domain for 3D spatial understanding. Each domain processes specific aspects of the parking environment independently, then integrates results to achieve accurate localization in GPS-denied environments
Solution Approach 2:
The system transitions from 2D GPS coordinates to 3D spatial localization by incorporating vertical dimension through TPV images and IMU altitude data. This dimensional expansion enables accurate positioning in multi-floor parking structures where traditional 2D GPS fails
2Productivity
If vehicles repeatedly search for parking spots without accurate localization, then the system maintains simple operation, but time consumption and fuel consumption increase
Solution Approach 1:
The system implements continuous feedback loops where detected parking landmarks are compared against the map database, localization results are continuously updated, and parking spot availability is monitored in real-time. This feedback mechanism enables vehicles to efficiently navigate to available spots without repeated searching
Solution Approach 2:
The system performs preliminary mapping of parking zones including identification of parking landmarks and available spots before vehicles arrive. This pre-processing enables rapid localization and direct navigation to parking spots, eliminating time-consuming search procedures
3Measurement precision
If multiple sensor data and processing domains are integrated, then localization accuracy improves, but computational complexity and processing time increase
Solution Approach 1:
The processing pipeline is segmented into distinct stages: image-domain feature detection, BEV image generation and processing, TPV image generation and processing, IMU data fusion, and final localization. Each stage handles specific computational tasks independently, managing complexity through modular organization
Solution Approach 2:
BEV images serve as an intermediary representation that simplifies the integration of multi-camera data by transforming complex 3D spatial relationships into a standardized top-down view. This intermediary format facilitates efficient feature matching and localization computations
4Measurement precision
If altitude compensation is not applied, then processing speed is faster, but localization accuracy deteriorates in multi-floor parking environments
Solution Approach 1:
The system dynamically adjusts the altitude parameter by integrating IMU acceleration data to compute vertical displacement. This parameter change compensates for elevation differences between floors, enabling accurate positioning in multi-floor environments where fixed-altitude assumptions fail
Data Source
AI summary
Methods and systems for assisting a vehicle to park using mixed-domain image data. Image-domain data is generated based on raw image data received from a plurality of cameras mounted on a vehicle. A bird's-eye-view (BEV) image is generated based on the raw image data. BEV-domain data associated with the BEV image is generated, which includes data associated with parking landmarks in the parking zone. A tri-perspective view (TPV) and associated data can be generated. A computing system localizes the vehicle within the parking zone based on the BEV-domain data, the image-domain data, and the TPV-domain data to generate localization data. The computing system performs mapping of the parking zone based on all three image-domain data and the localization data. A motion sensor such as an inertial measurement unit (IMU) can generate data that is used to compensate the various domain data.