Vehicle Camera Image Distortion Correction via Parallel Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Modern autonomous or semi-autonomous driving systems face challenges in accurately recognizing driving lanes and other vehicles due to distorted images captured by cameras with wide-angle or fisheye lenses, which require correction to provide reliable information for secure driving.
Innovation Solution
A method and device that transform camera images into display images projected in parallel onto a projection plane, preserving the size ratio of objects, allowing for accurate depiction and velocity determination of vehicles in the surrounding environment by averaging, filtering, or interpolating camera pixels to correct distortion effects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of moving object
If wide-angle or fisheye lenses are used to capture the vehicle's surrounding, then the field of view is increased, but image distortion occurs making it difficult to recognize objects accurately
Solution Approach 1:
The patent transforms the captured image from a distorted 2D representation to a corrected 2D representation by introducing a virtual 3D scene model. By defining virtual objects with known dimensions in a 3D space and projecting them through the camera model, the system creates a reference framework that enables accurate dimension measurement in the distorted image, effectively using 3D geometry to correct 2D distortion.
Solution Approach 2:
The patent changes the parameters of the image transformation process by introducing calibration parameters that relate virtual object dimensions to their projected image dimensions. By adjusting these parameters and using the known dimensions of virtual objects as reference, the system can compensate for lens distortion and recover accurate object dimensions from the distorted image.
2Measurement precision
If image distortion correction is applied to recognize objects accurately, then measurement precision is improved, but the processing complexity increases
Solution Approach 1:
The patent performs preliminary action by pre-defining virtual objects with known dimensions in the scene before actual object recognition takes place. These virtual objects serve as reference markers that are already positioned and dimensioned in the 3D space, allowing the system to use them as calibration references without requiring complex real-time distortion modeling during the recognition phase.
Solution Approach 2:
The patent creates a copy of the physical scene in virtual space by defining virtual objects that correspond to physical objects. These virtual copies have known dimensions and positions, and by comparing the projected virtual objects with actual detected objects, the system can derive distortion correction parameters without needing to directly measure distorted physical objects.
3Measurement precision
If parallel projection is used to maintain object size ratios in display images, then measurement precision is improved, but the transformation processing becomes more complex
Solution Approach 1:
The patent introduces a virtual scene model as an intermediary between the distorted camera image and the corrected display image. By projecting virtual objects through the camera model and comparing with detected objects, the system derives transformation parameters that map distorted coordinates to corrected coordinates, using the virtual scene as a mediating reference framework.
Solution Approach 2:
The patent segments the image transformation process into distinct steps: defining virtual objects with known dimensions, projecting them through the camera model, detecting corresponding physical objects, and calculating transformation parameters. This segmentation allows each step to be handled independently and simplifies the overall complex transformation process.
Data Source
AI summary
The invention relates to method for displaying a front-view of a vehicle's (42) surrounding, the method comprising the steps of: capturing (S1) camera images (20) of the vehicle's (42) surrounding through a lens (53) of a vehicle camera (44) on an image sensor (52) of said vehicle camera (44); transforming (S2) said captured camera images (20) by a processing unit (101) into display images (30), said display images (30) being substantially projected in parallel onto a projection plane (90) being parallel to an image sensor plane (54) of said image sensor (52); and displaying (S3) said display images (30) on a display device (102) for viewing by a driver of said vehicle (42).