Automotive Image Depth Calculation for Repetitive Pattern Ambiguity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image processing methods for automotive vehicle cameras face ambiguity in detecting repetitive patterns, leading to incorrect distance information and potential safety risks due to multiple possible locations of the same pattern in successive images.
Innovation Solution
A method that detects identical points in a sequence of images and calculates their depth using rotation matrices and translation vectors, identifying points with identical depths as representing the same real point, thereby distinguishing repetitive patterns and providing accurate depth information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional optical flow tracking is used to detect points in image sequences, then the processing method is simple and fast, but multiple identical points from repetitive patterns cause ambiguity and incorrect depth information
Solution Approach 1:
The patent segments the problem of identifying corresponding points by dividing it into multiple stages: first detecting candidate points in the epipolar segment, then calculating depths for each candidate, and finally selecting the correct correspondence based on depth consistency across multiple images. This segmentation allows the system to handle repetitive patterns by evaluating multiple candidates systematically rather than making a single ambiguous match.
Solution Approach 2:
The patent introduces depth as an additional dimension to resolve the 2D ambiguity of repetitive patterns. By calculating the depth (distance from camera) for each candidate point and requiring consistency across multiple images, the system transforms the problem from 2D pattern matching to 3D spatial reasoning, enabling correct identification of corresponding points even when patterns repeat in the image plane.
2Reliability
If multiple candidate points are considered for repetitive patterns, then the correct point can be identified, but the processing time and computational load increase
Solution Approach 1:
The patent performs preliminary actions by first detecting all candidate points in the epipolar segment before final identification. By pre-calculating potential correspondences and their depths in advance, the system prepares multiple hypotheses that can be quickly evaluated later, reducing the computational burden during real-time processing while maintaining reliability in identifying the correct point.
Solution Approach 2:
The system uses feedback from depth calculations to validate candidate correspondences. By computing depths for multiple candidates and checking for consistency across image sequences, the system receives feedback that confirms or rejects each candidate, allowing efficient elimination of incorrect matches while preserving the correct one without requiring exhaustive processing of all possibilities.
Data Source
AI summary
A method of processing images allowing the identification of at least one same point appearing on each image of a sequence of images, the sequence of images including a first image, a second image and a third image generated respectively at a first instant, a second instant, later than the first instant, and a third instant, later than the second instant, by a camera mounted in an automotive vehicle. The method includes detecting a plurality of identical points in each image of the sequence of images, relating to a plurality of repetitive patterns in each image of the sequence of images and of identification, in the second image and in the third image, of the identical points corresponding to the points in space for which the calculated depth is identical, these points in space representing respectively in the second image and the third image one and the same real point.