Parking Map Localization via Distribution Shifting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current parking detection systems generate a high number of false positives due to inaccuracies in localization, where detected open areas are incorrectly identified as valid parking spaces, especially in busy streets, with up to 100 false positives for every true positive, and these errors persist even when correlated with parking maps.
Innovation Solution
The method involves shifting detected object distributions based on a predefined parking region distribution to align them with permissible parking areas, using a score function to minimize errors and generate a more accurate parking map by processing data from vehicles traveling through street segments, thereby correcting localization inaccuracies and distinguishing between valid and invalid parking spaces.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If sensors are used to detect open areas for parking, then parking availability information can be provided, but false positives increase significantly (up to 100 false positives per true positive)
Solution Approach 1:
The system uses feedback from multiple vehicle detections to continuously refine and update the parking map. By aggregating detection data from multiple vehicles passing through the same street segment and comparing it with the predefined parking region distribution, the system iteratively improves detection accuracy and reduces false positives through cumulative validation.
Solution Approach 2:
The patent introduces a predefined parking region distribution map as an intermediary reference framework. This map serves as a mediator between raw sensor detections and final parking availability determination, allowing the system to validate detected open areas against known permissible parking regions and filter out false positives.
2Area of stationary object
If more vehicles are used to collect detection data, then coverage improves, but processing complexity and time increase
Solution Approach 1:
The system divides the street into multiple street segments and uses separate detection and validation processes for each segment. By segmenting the overall detection task into smaller unit-specific processes, the system can efficiently handle data from multiple vehicles without overwhelming computational complexity, as each segment is processed independently against its own predefined parking region distribution.
3Measurement precision
If localization accuracy is improved using traditional methods, then object positioning improves, but false positives due to systematic errors persist
Solution Approach 1:
The predefined parking region distribution map acts as an intermediary reference that corrects systematic localization errors. Rather than relying solely on improving sensor precision, the system uses this reference map as a mediator to validate whether detected objects and open areas fall within permissible parking regions, thereby correcting systematic errors in localization.
Solution Approach 2:
The system employs feedback mechanisms where detection results are continuously validated against the predefined parking region distribution. This feedback loop allows the system to identify and correct systematic localization errors by comparing detected positions with expected positions based on the reference map, improving reliability without requiring higher measurement precision.
Data Source
AI summary
A method for object detection and parking spot localization includes receiving data corresponding to detected objects located within a street segment, the data being ascertained by an ascertaining vehicle driving through the street section, determining a detected object distribution, shifting the detected object distribution, and generating a parking map of available parking spots on the street segment based on a parking region distribution and the shifted detected object distribution. The method includes receiving the data, determining the detected object distribution, and shifting of the detected object distribution, for example each time at least one of the at least one ascertaining vehicle drives through the street section.