Map Generation Using Satellite and Vehicle Image Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing map generation systems from satellite images struggle with occlusions caused by obstacles like buildings and trees, which obscure road features and hinder accurate identification of lane lines and other road features, leading to incomplete or inaccurate map generation.
Innovation Solution
An apparatus and method that utilize a processor to identify lane lines and other road features from both satellite images and vehicle-captured images, align and deform the images to match road features, and combine them to generate a high-precision map, even in the presence of occlusions, using affine transformation and spline deformation techniques.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If satellite images are used for map generation, then coverage area is large, but occlusions by obstacles obscure road features
Solution Approach 1:
The patent combines satellite images (wide coverage) with vehicle-mounted camera images (clear road features) by aligning and superimposing them. The processor identifies corresponding points between the two image types and merges them to create a comprehensive map that overcomes the occlusion problem while maintaining wide coverage area.
Solution Approach 2:
The patent uses a third reference system (map data with known coordinates) as an intermediary to bridge satellite images and vehicle images. By transforming both image types into the map coordinate system and aligning them through common reference points, the system resolves the occlusion issue while preserving the advantages of both image sources.
2Loss of information
If vehicle camera images are used for map generation, then road features are clearly visible, but coverage area is limited
Solution Approach 1:
The patent merges vehicle camera images (with clear road features) with satellite images (with wide coverage) by identifying corresponding points and superimposing the images in a unified coordinate system, thereby achieving both accurate road feature identification and extensive coverage area.
Solution Approach 2:
The patent transitions from two-dimensional image processing to three-dimensional spatial alignment by introducing map coordinate data as a third dimension. This allows accurate registration of road features from vehicle images with the broader satellite image context through coordinate transformation and spatial alignment.
3Manufacturing precision
If occluded regions are present in satellite images, then map generation becomes inaccurate, but avoiding occluded regions reduces coverage
Solution Approach 1:
The patent extracts and utilizes the clear road feature information from vehicle camera images to fill in and correct occluded regions in satellite images. By identifying corresponding points between image types, the system extracts accurate feature data from the non-occluded vehicle images and applies it to reconstruct the occluded areas, maintaining both accuracy and coverage.
Solution Approach 2:
The patent uses map coordinate data as an intermediary to resolve the contradiction between accuracy and coverage. By transforming both image types into the map coordinate system and using known accurate positions as reference points, the system can accurately reconstruct occluded regions while maintaining the original wide coverage area of the satellite images.
Data Source
AI summary
An apparatus for generating a map identifies a lane line and a feature other than a lane line on a road from a first image of a predetermined location of the road taken downward from the sky; identifies the lane line and the feature from a second image representing the predetermined location of the road and made based on images taken by a camera provided for a vehicle; aligning the first images with the second images, based on the predetermined location; deforms the second image so that the feature in the second image best fits the feature in the first image; further deforms the second image in a direction perpendicular to the front-back direction of the road so that the position of the feature remains unchanged and that the positions of the lane lines in the first and second images match; and combines the first and deformed second images.