Camera Calibration Saddle Point Detection Across Shifted Image Regions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing camera calibration methods face challenges in accurately detecting saddle points, particularly when they are located on the border between regions of interest, leading to reduced recall and increased computational complexity.
Innovation Solution
A method involving image shifting and re-shifting using machine learning, specifically an artificial neural network, to detect and align saddle points across multiple shifted images, leveraging a grid structure and non-maximum suppression to enhance detection accuracy and reduce computational time complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing camera calibration methods are used to detect saddle points, then the detection process can be completed, but the recall is reduced and computational complexity increases when saddle points are located on border regions
Solution Approach 1:
The input image is divided into multiple regions of interest (ROIs), and each ROI is processed independently through the neural network. This segmentation allows the system to handle saddle points on borders more effectively by distributing the detection task across multiple smaller regions, improving recall while managing computational complexity through parallel processing.
Solution Approach 2:
The patent processes the image by creating multiple overlapping regions of interest, effectively adding a spatial dimension to the detection process. By evaluating the same image content from multiple overlapping regions, the system ensures that saddle points near borders are captured by at least one ROI, thereby improving detection recall without requiring a fundamentally more complex computational approach.
2Measurement precision
If multiple shifted images are processed to improve saddle point detection accuracy, then detection recall improves, but processing time increases
Solution Approach 1:
The patent applies preliminary actions by creating multiple regions of interest with different shifts before the actual detection process. These pre-defined shifts are optimized to cover border regions effectively, ensuring that saddle points are captured in at least one ROI without requiring exhaustive processing of all possible shifts, thus balancing accuracy improvement with time efficiency.
Solution Approach 2:
Different regions of interest are created with local quality variations through different shifts and overlapping configurations. This allows the system to focus computational resources on regions where saddle points are more likely to be missed (border regions), improving overall detection accuracy while minimizing unnecessary processing in regions that are already well-covered.
Data Source
Figure 1A
Figure 1B
Figure 2
AI summary
A computer implemented method for determining pre-determined points in an input image comprises the following steps carried out by computer hardware components: applying a plurality of shifts to the input image to obtain a plurality of shifted images; detecting pre-determined points in each of the shifted images; re-shifting the detected pre-determined points; and determining the pre-determined points based on the re-shifted detected pre-determined points.