Parking Lot Outline Tracing via Template Similarity Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing parking-lot tracing systems face inaccuracies due to image distortion and misrecognition of parking-lot lines, especially when lines are hidden by vehicles or obstacles, leading to decreased recognition accuracy.
Innovation Solution
An apparatus and method that generate a template for parking-lot lines from images, adjusting thickness and width within a range, and using similarity calculations to correct the template based on the highest similarity match, even when lines are obscured, to determine accurate positions and orientations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a parking-lot tracing system uses Automatic Vehicle Monitoring (AVM) images to recognize parking-lots, then the system can provide parking assistance, but image distortion causes recognition inaccuracy and difficulty in tracing when parking-lot lines are hidden by vehicles or obstacles
Solution Approach 1:
The system changes parameters of the template including thickness of the parking-lot line, width of the parking-lot, and position within a certain range to generate multiple templates. This allows the system to adapt to image distortions and find the best matching template even when lines are hidden or distorted in the AVM image.
Solution Approach 2:
The system generates a template based on previously recognized parking-lot information before tracing the current parking-lot. This preliminary template is then used as a reference for similarity calculation, enabling the system to maintain recognition accuracy even when current image conditions are poor due to distortion or hidden lines.
2Measurement precision
If the system generates a template based on previously recognized parking-lot information, then it can improve tracing accuracy, but the system complexity increases due to template generation and similarity calculation
Solution Approach 1:
The system creates a simplified copy of the previously recognized parking-lot as a template, which captures the essential geometric characteristics (position, orientation, dimensions) without requiring complex image processing. This template copy is then used for efficient similarity comparison with current images, maintaining high tracing accuracy while avoiding excessive system complexity.
3Adaptability or versatility
If the system adjusts template parameters such as thickness and width within a certain range, then it can handle image distortions better, but the calculation time and processing complexity increase
Solution Approach 1:
The system efficiently handles parameter variations by defining a certain range for thickness, width, and position adjustments rather than exhaustively searching all possible values. This approach allows the system to adapt to image distortions while limiting the search space, thereby reducing processing time compared to exhaustive parameter optimization methods.
Data Source
AI summary
An apparatus and a method for tracing a parking-lot is provided that includes a controller configured to recognize at least one parking-lot from a previous image frame which photographed a surrounding of a vehicle and extract a template according to a type of a parking-lot line of the recognized parking-lot. In addition, the controller is configured to generate a template transformed based on a position information of the parking-lot and calculate similarity by comparing a template generated from a previous image frame with a parking-lot line recognized from a current image frame. A position of a parking-lot is determined according to the calculated similarity and the controller is configured to correct the template based on an information of a parking-lot line extracted from the determined position.