Multi-Camera Vehicle Positioning for Indoor Parking Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle positioning technologies face challenges in indoor environments due to signal attenuation, reduced precision, and complex terrains, making accurate vehicle positioning difficult in large-scale indoor parking lots.
Innovation Solution
A method and apparatus utilizing multiple cameras on a vehicle to capture image data from different angles, processing these images to extract feature points, and transmitting them to a cloud for matching with a pre-established fused map to determine the vehicle's position, reducing data transmission and computing power requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If GPS technology is used for vehicle positioning, then outdoor positioning precision is improved, but indoor positioning reliability deteriorates due to signal attenuation and occlusion
Solution Approach 1:
The patent introduces visual feature points as an intermediary between the vehicle and the positioning system. Instead of directly using GPS signals which fail indoors, the system captures images, extracts feature points from the environment, and uses these feature points as intermediaries to determine vehicle position through matching with pre-stored map data. This intermediary approach enables reliable positioning in indoor environments where GPS signals are unavailable.
Solution Approach 2:
The patent replaces the electromagnetic signal-based GPS positioning system with a visual-based positioning system. Instead of relying on radio frequency signals that are blocked by structures, the system uses optical images captured by cameras, processes them to extract visual features, and matches these features with stored visual maps. This substitution of the positioning mechanism enables operation in indoor environments.
2Loss of information
If multiple channels of image data are captured by multiple cameras, then visual input information is increased, but data transmission volume increases
Solution Approach 1:
The patent extracts only the essential visual features from the captured images rather than transmitting the complete image data. By using feature point extraction algorithms to identify and isolate key visual elements (corners, edges, distinctive patterns), the system retains the necessary positioning information while reducing data volume dramatically. Only these extracted feature points are transmitted to the server, not the entire multi-channel image data.
Solution Approach 2:
The patent segments the visual information into discrete feature points rather than processing images as continuous data. Each image from multiple cameras is divided into identifiable feature elements that can be independently extracted, matched, and processed. This segmentation approach allows the system to utilize information from multiple camera channels while keeping transmission requirements manageable by sending only the segmented feature data.
3Measurement precision
If feature point matching is performed on the cloud, then positioning accuracy is improved, but computing power consumption increases
Solution Approach 1:
The patent inverts the traditional computing architecture by performing the computationally intensive feature matching operations on the cloud server rather than on the vehicle's onboard computer. Instead of having the moving vehicle execute complex algorithms, the system uploads feature points to a stationary server that performs the matching against pre-stored map data. This inversion transfers the computing burden from the mobile platform to the fixed infrastructure, reducing energy consumption on the vehicle while maintaining high positioning accuracy.
Data Source
AI summary
A method and apparatus for vehicle positioning, including: obtaining multiple channels of image data of surrounding environment of a vehicle, where the multiple channels of image data are taken by multiple cameras at different angles concurrently; processing the multiple channels of image data to extract feature points in an image frame of each channel of image data; transmitting the feature points to a cloud, and receiving from the cloud a coordinate of current position of the vehicle in a vector map obtained by feature point matching.