Stereo Image Region Switching for Accurate Vehicle Distance Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Stereo cameras face challenges in accurately estimating distances to objects that initially appear in monocular regions, leading to increased collision risks and sudden braking due to errors in distance calculation, particularly when vehicles turn or objects enter from the side, necessitating larger electronic circuits to handle increased calculations.
Innovation Solution
An image processing device that selectively processes regions of captured images from multiple cameras based on the number of images introduced per unit time, using multiple parallax image creation units to create parallax images, thereby reducing electronic circuit scale and enhancing safety.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If the overlapping region (stereo region) of a plurality of cameras is decreased to decrease the angle of view, then the electronic circuit scale is reduced, but the distance estimation accuracy for objects in monocular regions deteriorates
Solution Approach 1:
The patent dynamically adjusts the stereo region range based on detected objects. When an object is detected in the monocular region, the system expands the stereo region to include that object in subsequent frames, enabling accurate distance measurement. This dynamic adjustment resolves the contradiction by allowing the system to maintain small circuit scale while temporarily expanding processing regions when needed for accuracy.
Solution Approach 2:
The system performs preliminary object detection in monocular regions using assumed distance information, then prepares to expand the stereo region in advance when objects are detected. This preliminary action allows the system to anticipate when accurate distance measurement will be needed and prepare the expanded stereo region accordingly, balancing circuit scale and measurement precision.
2Measurement precision
If the stereo region is expanded to improve distance estimation accuracy for objects entering from side areas, then the electronic circuit scale increases, but the collision risk remains high due to initial estimation errors
Solution Approach 1:
The system dynamically expands the stereo region only when and where objects are detected in monocular regions, rather than maintaining a permanently large stereo region. This dynamic approach ensures accurate distance estimation for objects entering from side areas while keeping the overall circuit scale manageable by limiting expansion to necessary regions and timeframes.
Solution Approach 2:
The patent applies different processing qualities to different regions: the majority of the field of view uses the smaller monocular region with assumption-based distance estimation, while specific local areas where objects are detected receive the enhanced stereo region processing. This local quality differentiation resolves the contradiction by providing high measurement precision only where needed rather than uniformly across the entire field of view.
3Speed
If the number of captured images processed per unit time is increased to improve detection responsiveness, then the processing speed increases, but the electronic circuit scale and calculation load increase
Solution Approach 1:
The system processes a selective subset of captured images rather than all images at full resolution. It identifies regions containing objects and applies detailed stereo processing only to those partial regions, while other areas use simpler monocular processing. This partial action approach maintains high processing speed while reducing the overall calculation load and circuit scale requirements.
Solution Approach 2:
The patent segments the image processing into different regions: monocular regions use simplified distance estimation, while stereo regions undergo comprehensive parallax-based distance measurement. This segmentation allows the system to process images efficiently by applying appropriate processing intensity to different segments, balancing speed and circuit complexity.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In an aspect of the present invention, there is provided an image processing device for processing a series of a pair of captured images acquired by a pair of camera modules adapted to capture an image of a surrounding environment of a vehicle, thereby obtaining surrounding-environment recognition information for controlling the vehicle, wherein the image processing device is enabled to select a region of the captured images to be used for creating a parallax image from the pair of captured images, and the image processing device changes over the selection of the region according to the number of the pairs of captured images to be introduced into image processing per unit time.