Geometric Distortion Correction for Real-Time Object Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object recognition methods struggle to apply real-time processing to high-resolution images from broad-range image sensing devices due to increased processing volume required for distortion correction, making real-time object recognition at high frame rates unfeasible.
Innovation Solution
An image processing apparatus that acquires both high-resolution and low-resolution images from an omni-directional camera, detects objects in the low-resolution images, and applies geometric distortion correction only to the object recognition regions in the high-resolution images, allowing for real-time object recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If distortion correction is applied to the entire high-resolution image, then geometric accuracy is improved, but processing time increases and real-time recognition becomes difficult
Solution Approach 1:
The patent divides the image processing into two stages: first detecting objects in a low-resolution image, then applying distortion correction only to the identified object regions in the high-resolution image. This segmentation of processing scope resolves the contradiction by maintaining geometric accuracy for relevant regions while minimizing overall processing time.
Solution Approach 2:
The patent applies distortion correction selectively only to regions containing detected objects rather than uniformly processing the entire high-resolution image. This local quality approach ensures geometric accuracy is improved where needed (in object regions) while avoiding unnecessary processing in background areas, thus reducing overall processing time.
2Manufacturing precision
If distortion correction is applied to the entire high-resolution image, then geometric accuracy is improved, but processing volume increases
Solution Approach 1:
The patent segments the processing workload by first identifying object regions in a low-resolution image, then restricting distortion correction operations to only those specific regions in the high-resolution image. This segmentation dramatically reduces the total processing volume compared to correcting the entire image while preserving geometric accuracy for the object regions.
Solution Approach 2:
The patent implements local quality processing by applying distortion correction exclusively to object-containing regions rather than the entire image. This approach reduces processing volume by eliminating redundant corrections in background areas while maintaining the necessary geometric accuracy for recognition tasks.
3Measurement precision
If object recognition is applied to high-resolution images from broad-range sensing devices, then recognition accuracy is improved, but processing complexity increases
Solution Approach 1:
The patent performs preliminary object detection in a low-resolution image before applying distortion correction to the high-resolution image. This preliminary action simplifies the overall processing complexity by identifying target regions early, allowing subsequent high-precision processing to be focused only on relevant areas rather than the entire high-resolution image.
Solution Approach 2:
The patent applies distortion correction with local quality control, processing only the regions containing detected objects in the high-resolution image. This approach maintains recognition accuracy by ensuring geometric correctness in object regions while reducing processing complexity by avoiding unnecessary corrections in the entire image.
Data Source
AI summary
A high-resolution image obtained by an image sensing operation by an image sensing unit, and a low-resolution image having a resolution lower than the high-resolution image are acquired. An object which satisfies a predetermined condition is detected from the low-resolution image, and an object recognition processing for a region corresponding to the object in the high-resolution image is performed, thus correcting geometric distortions of the region.