In-Vehicle Camera ROI Setting for Accurate Posture Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to appropriately set a processing target region for posture estimation of an in-vehicle camera due to variations in vehicle type and attachment conditions, which affect the appearance of the camera image, making it difficult to accurately determine a suitable region of interest (ROI) for optical flow-based posture estimation.
Innovation Solution
The method involves setting a processing target region based on a horizontal line position derived from the intersection of parallel lines in the real space, considering the attachment height and horizontal line deviation of the camera, and adjusting the region to exclude vehicle body portions from the ROI to ensure accurate posture estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If a fixed ROI is set in the camera image for posture estimation, then the posture estimation can be performed with a simple method, but the estimation accuracy deteriorates due to variations in vehicle type and camera attachment conditions
Solution Approach 1:
The patent applies dynamics by making the ROI position adaptive rather than fixed. The lower end position of the ROI is dynamically determined based on the detected horizon line position and predetermined offset distance, allowing the processing target region to automatically adjust to different camera attachment conditions and vehicle types, thereby maintaining high posture estimation accuracy across various scenarios
Solution Approach 2:
The patent changes the parameter of ROI position from a fixed value to a variable determined by horizon line detection. By using the horizon line position as a reference and adding a predetermined offset distance, the system dynamically adjusts the ROI's lower end position to account for variations in camera attachment height and vehicle type, resolving the contradiction between simplicity and accuracy
2Area of stationary object
If the ROI includes the vehicle body portion in the camera image, then the processing area is maximized, but the posture estimation accuracy deteriorates due to interference from vehicle body features
Solution Approach 1:
The patent applies the extraction principle by removing the vehicle body portion from the processing target region. By setting the ROI's lower end position above the horizon line using a predetermined offset distance, the system automatically excludes the vehicle body (which appears below the horizon line in the image) from the processing area, thereby preventing interference with posture estimation while still maximizing the usable processing area
Data Source
AI summary
An information processing device includes: a controller configured to execute a posture estimation processing by estimating a posture of an in-vehicle camera based on an optical flow of a feature point in a processing target region set in a camera image captured by the in-vehicle camera. The controller is configured to set a position of the processing target region in a vertical direction based on a horizontal line position which is a position of a horizontal line derived from a point where two parallel lines intersect with each other in a real space observed in the camera image, an attachment height of the in-vehicle camera, and a horizontal line deviation amount which is a deviation amount between the horizontal line position and an optical axis center of the in-vehicle camera.