Object Detection for Distorted Images Using Selective Corner Point Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous vehicle systems face challenges in accurately detecting parking spots using distorted images from fisheye or wide-angle lenses, as conventional methods require distortion correction, which reduces resolution and increases computational resources, and labeling such images is labor-intensive due to curved shapes appearing in distorted views.
Innovation Solution
A method that directly detects corner points of parking spots in distorted images, maps these points to a distortion-corrected domain, interpolates intermediate points to generate lines, and maps them back to the distorted domain for precise object location, allowing for efficient detection and adjustment of vehicle direction without the need for full-image correction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If distortion correction is applied to the entire image, then object detection accuracy is improved, but image resolution is reduced and computational resources are increased
Solution Approach 1:
The patent segments the distortion correction process by applying it only to specific regions of interest (corner points and their neighborhoods) rather than the entire image. This selective approach maintains object detection accuracy while significantly reducing computational resources and preserving image resolution in non-processed areas.
Solution Approach 2:
The patent applies local quality by performing distortion correction only in local regions around detected corner points rather than globally across the entire image. This localized processing maintains the necessary precision for detection while avoiding the computational overhead and resolution loss associated with full-image correction.
2Measurement precision
If distortion correction is applied to the entire image, then object detection accuracy is improved, but image resolution is reduced
Solution Approach 1:
The patent segments the image processing into two parts: corner points undergo distortion correction while the rest of the image remains uncorrected. This segmentation preserves the high resolution of the original image in most areas while still achieving accurate object detection through corrected corner point coordinates.
Solution Approach 2:
The patent applies local quality by restricting distortion correction to small local regions around corner points. This ensures that the critical detection points have corrected coordinates for accurate object localization, while the broader image maintains its original high resolution and visual quality.
3Productivity
If corner points are detected in distorted images, then detection efficiency is improved, but accurate object localization becomes difficult without full distortion correction
Solution Approach 1:
The patent uses corner points as intermediary elements that bridge the distorted image domain and the desired object localization. By detecting corner points in the distorted image and applying minimal distortion correction only to these points, the system achieves both detection efficiency and accurate object localization without requiring full-image correction.
Data Source
AI summary
Techniques including receiving a distorted image from a camera disposed about a vehicle, detecting, in the distorted image, corner points associated with a target object, mapping the corner points to a distortion corrected domain based on one or more camera parameters, mapping the corner points and lines between the corner points back to a distorted domain based on the camera parameters, interpolating one or more intermediate points to generate lines between the corner points in the distortion corrected domain mapping the corner points and the lines between the corner points back to a distorted domain based on the camera parameters, and adjusting a direction of travel of the vehicle based on the located target object.