Stereo Image Vertical Shift Correction for Solid Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In stereo cameras mounted on vehicles, the heavy load of processing feature detection across multiple frames leads to inadequate correction of camera parameter deviations, particularly in vertical shifts, resulting in erroneous object detection.
Innovation Solution
An image processing device that includes input units for stereo images, a geometric correction unit, a parallax calculating unit, a solid object estimating unit, a vertical shift detecting unit, and a determining unit to accurately detect and correct vertical shifts between image regions, ensuring proper object identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If feature detection processing is performed across multiple frames to correct camera parameter deviation, then correction accuracy is improved, but processing load increases and real-time performance deteriorates
Solution Approach 1:
The patent performs geometric correction using preset correction information before parallax calculation and solid object estimation. This preliminary correction of vertical shifts and geometric distortions ensures that subsequent processing operates on corrected images, improving accuracy without requiring extensive multi-frame feature detection
Solution Approach 2:
The patent divides the image processing into distinct stages: geometric correction, parallax calculation, and solid object estimation. By segmenting the processing pipeline and applying corrections at the geometric stage, the system achieves accurate camera parameter correction while maintaining efficient real-time processing in subsequent stages
2Productivity
If feature detection processing is divided across multiple frames to reduce load, then processing speed is improved, but vertical shift correction accuracy deteriorates
Solution Approach 1:
The system performs geometric correction as a preliminary step before solid object estimation, using preset correction information to correct vertical shifts and geometric distortions. This ensures that even when processing is divided across frames, the correction accuracy is maintained because the geometric relationships are established beforehand
Solution Approach 2:
The patent introduces geometric correction information as an intermediary that mediates between the raw stereo images and the parallax calculation process. This intermediary correction layer ensures accurate vertical shift compensation without requiring all processing to occur within single frames
3Productivity
If geometric correction is performed using preset information, then processing efficiency is improved, but adaptability to unexpected shifts deteriorates
Solution Approach 1:
The patent calculates parallax based on geometrically corrected images and uses this parallax information to determine solid objects. The system effectively uses the parallax calculation results as feedback to validate the geometric correction, ensuring that the preset correction information remains appropriate and adapting to actual scene conditions
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In the present invention, even when there is a vertical shift in a stereo image, a solid object can be accurately detected from the stereo image. A solid object estimating unit 121 performs, on the basis of a parallax image calculated from a parallax calculating unit 120, the estimation of a solid object for determining whether an object included in a criterion image and a reference image is a solid object or not. When the object is estimated as a solid object by the solid object estimating unit 121, an object region vertical shift detecting unit 122 detects, by using a criterion image and a reference image in the same frame for which the estimation of a solid object has been performed, a vertical shift between an object region including an object in the criterion image and an object region including the object in the reference image. An object region solid object determining unit 123 calculates the parallax of the object on the basis of the vertical shift, detected by the object region vertical shift detection unit 122, between the object region in the criterion image and the object region in the reference image, and determines whether the object is a solid object.