Bird's Eye View Image Distortion Correction via 3D Grid Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current driving assistance systems using multiple image capturing modules to create a Bird's eye view image face issues with image distortion, making it difficult for drivers to accurately assess distances to objects on the road, leading to potential accidents.
Innovation Solution
An image processing method and system that captures multiple images, projects a three-dimensional curved surface, and transforms two-dimensional planar images into three-dimensional images by dynamically adjusting grid drawing algorithms based on angles between normal vectors, effectively reducing distortion and improving visual accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If traditional bird's eye view transformation is used to combine multiple 2D images, then a full aerial view is provided for driving reference, but image distortion occurs making it difficult to accurately evaluate distances to objects with actual heights
Solution Approach 1:
The patent transforms 2D planar images into a 3D curved surface coordinate system, adding a vertical dimension to the transformation process. This dimensional change allows objects with actual heights to be represented in three-dimensional space, preserving their true spatial relationships and enabling accurate distance evaluation while maintaining the bird's eye view capability for driving reference
Solution Approach 2:
The patent introduces normal vector angles as a new parameter in the transformation process. By calculating the angle between the normal vector of the 2D image plane and the normal vector of each grid on the 3D curved surface, the system dynamically adjusts the projection parameters to minimize distortion and preserve accurate spatial measurements for distance evaluation
2Measurement precision
If dynamic grid redrawing based on normal vector angles is implemented, then transformation accuracy is improved, but computing complexity increases
Solution Approach 1:
The patent divides the transformation process into discrete grid units, where each grid on the 3D curved surface is independently processed. By segmenting the continuous transformation into discrete grid operations with corresponding normal vector calculations, the system achieves high transformation accuracy through localized precision while managing computing complexity through systematic decomposition
Solution Approach 2:
The patent pre-establishes the correspondence relationships between 2D image grids and 3D curved surface grids, including their normal vectors and angle calculations. This preliminary setup of transformation parameters and grid mappings enables accurate real-time transformation without repeatedly computing complex geometric relationships during actual image processing
Data Source
AI summary
An image processing method and system using the same, wherein the image processing method includes capturing a plurality of images corresponding to the surroundings of an object using a plurality of image capturing modules to generate a two-dimension planar image; providing a three-dimension projected curved surface; defining a plurality of first grids on the three-dimension projected curved surface and a plurality of second grids on the two-dimension planar image, wherein each of the first grids correspond to each of the second grids; transforming the first grids on the three-dimension projected curved surface and the second grids on the two-dimension planar image into a plurality of first redrawn grids and second redrawn grids respectively based on the angles formed between the normal vector of the two-dimension planar image and the normal vector of each first grid, wherein each first redrawn grid corresponds to each second redrawn grid.