Curb Detection for Parking Occupancy Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video-based methods for determining parking occupancy rely heavily on vehicle detection, which is challenging due to variations in vehicle colors, shapes, sizes, and weather and illumination conditions, requiring complex algorithms and high computational costs.
Innovation Solution
A method and system for on-street parking occupancy estimation using curb detection, where a computer system is trained to identify curbs in image data, allowing for robust and computationally efficient estimation of parking availability by determining the length of visible curb and correlating it with parking space occupancy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If vehicle detection algorithms are used to determine parking occupancy, then parking occupancy can be identified, but the computational cost and algorithm complexity increase significantly
Solution Approach 1:
The patent extracts the curb as a separate detection target from vehicle detection. Instead of detecting vehicles directly, the system detects the curb and infers parking occupancy from curb visibility. This extraction simplifies the detection task by focusing on a single, consistent geometric feature rather than diverse vehicle characteristics.
Solution Approach 2:
The patent inverts the traditional detection approach by not detecting vehicles to determine occupancy, but rather detecting the absence of curbs (occluded curbs) to infer vehicle presence. This inversion transforms a complex object detection problem into a simpler edge detection and occlusion analysis problem.
2Measurement precision
If vehicle detection algorithms are used to determine parking occupancy, then parking occupancy can be identified, but computational cycles and processing time increase
Solution Approach 1:
The patent extracts curb detection as the primary task, which is computationally less intensive than full vehicle detection. Curb detection involves identifying linear geometric features and continuity, which requires fewer computational resources and processing time compared to detecting diverse vehicle objects with varying characteristics.
Solution Approach 2:
The patent uses simple geometric feature detection (curb edges and continuity) instead of complex vehicle object detection. This approach uses computationally cheaper operations that can be performed quickly, sacrificing the detailed information of full vehicle detection while maintaining sufficient accuracy for occupancy determination.
3Measurement precision
If vehicle detection is used, then parking occupancy can be determined, but robustness to varying weather and illumination conditions decreases
Solution Approach 1:
The patent detects curbs which have relatively homogeneous geometric characteristics (linear edges, consistent structure) compared to the diverse appearance of vehicles. This homogeneity in the detection target makes the detection more robust to varying weather and illumination conditions, as curb geometry remains consistent regardless of environmental factors.
Solution Approach 2:
The patent inverts the detection approach to focus on curb visibility rather than vehicle appearance. By detecting the presence or absence of curb features rather than vehicle characteristics, the system achieves greater robustness to weather and illumination variations that significantly affect vehicle appearance detection.
Data Source
AI summary
A method and system for on-street vehicle parking occupancy estimation via curb detection comprises training a computer system to identify a curb, evaluating image data of the region of interest to determine a region wherein a curb is visible in said region of interest, and estimating a parking occupancy of said region of interest according to said region where said curb is visible.