Parking Map Generation Using Camera and LIDAR Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current vehicle parking systems in large cities face challenges in scalability and dynamic updates due to limitations in detecting available parking spaces, as existing systems rely on static sensors or crowd-sourcing methods that require manual updates and accurate mapping.
Innovation Solution
A method utilizing a camera and LIDAR device to capture images and point clouds, which are analyzed together to generate a parking map by detecting parked vehicles and generating bounding boxes for parking spaces, allowing for real-time updates and improved accuracy in identifying parking space availability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If static sensors (magnetometer, infrared, or radar) are used to detect parking space occupancy, then detection capability is provided, but scalability to other vehicles and dynamic updates are limited
Solution Approach 1:
The patent uses visual copying through images and point clouds to represent physical parking spaces and vehicles. Instead of requiring physical sensors at each parking location, the system creates digital copies (images from cameras, point clouds from LIDAR) that can be processed to detect parking space occupancy. This allows any vehicle equipped with these sensors to contribute to and access the parking map, enabling scalability and dynamic updates without physical infrastructure at each location.
2Quantity of substance
If crowd-sourcing methods are used to collect parking data, then data collection capability is provided, but manual updates and accurate mapping become complex
Solution Approach 1:
The system enables self-service through automated processing of images and point clouds. Vehicles automatically capture data, the system automatically processes the data through object detection and clustering algorithms, and automatically updates the parking map without manual intervention. This eliminates the complexity of manual updates while maintaining accurate mapping through automated computer vision and point cloud processing.
3Loss of information
If manually provided parking maps are used to indicate parking space locations, then parking space identification is provided, but ease of provision and updates to include additional parking spaces deteriorates
Solution Approach 1:
The patent transforms static, manually updated parking maps into dynamic maps that automatically update in real-time. The system continuously collects new images and point clouds from vehicles, processes them to detect new parking spaces and changes in occupancy, and automatically updates the parking map. This dynamic approach eliminates the difficulty of manual updates while maintaining complete and current parking space location information.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enhances the detection of parking spaces and their availability by combining visual recognition of vehicles with precise location and orientation data from LIDAR, providing a scalable and dynamically updated parking map system.
Implementation Method 1
obtaining, via a camera, an image acquired at a location
Implementation Method 2
obtaining, via a light detector, a point cloud acquired at a second location
Data Source
AI summary
A parking map generated based on determining a plurality of object clusters by associating pixels from an image with points from a point cloud. At least a portion of the plurality of object clusters can be classified into one of a plurality of object classifications including at least a vehicle object classification. A bounding box for one or more of the plurality of object clusters classified as the vehicle object classification can be generated. The bounding box can be included as a parking space on a parking map based on a location associated with the image and/or point cloud.