Stereo Camera Road Surface Estimation Using Previous Frame Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional object detection systems using stereo cameras face accuracy issues in estimating the road surface and detecting objects due to temporary image abnormalities, such as wiper blades or unclear images, which lead to incorrect detection results, especially when image conditions change rapidly.
Innovation Solution
An object detection device with a computer system that includes a distance image generation part, road surface estimation part, object detection part, and abnormality detection part, which generates distance images, estimates the road surface, detects objects, and compares current images with previous ones to identify and mitigate image abnormalities, allowing for accurate road surface estimation even in abnormal conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional stereo camera methods are used to estimate road surface and detect objects, then object detection can be performed, but calculation accuracy deteriorates for distant objects and incorrect detections occur (e.g., detecting traffic marks as objects)
Solution Approach 1:
The patent applies preliminary action by estimating the road surface in advance using a plane equation before performing object detection. This preliminary road surface estimation is then used to correct disparity calculations for distant objects, improving both disparity accuracy and preventing incorrect detections of traffic marks as objects.
Solution Approach 2:
The patent uses a simplified plane equation model to create a mathematical copy/representation of the road surface. This plane model serves as a reference that can be used to correct and verify disparity calculations, improving measurement precision without requiring complex additional hardware.
2Device complexity
If stereo camera is mounted on rear view mirror to capture images, then compact installation is achieved, but image abnormality occurs when wiper blades are contained in the captured images
Solution Approach 1:
The patent applies dynamics by making the road surface estimation adaptive and dynamic. The system continuously updates the plane equation parameters based on current image data, allowing it to adapt to changing conditions including temporary image abnormalities from wiper blades. This dynamic adjustment maintains reliability despite occasional poor quality images.
Solution Approach 2:
The patent uses feedback mechanisms where the system continuously monitors image quality and disparity calculation results, then adjusts the road surface plane parameters accordingly. This feedback loop allows the system to recover from temporary image abnormalities and maintain accurate road surface estimation over time.
Data Source
AI summary
In an object detection device, a distance image generation part generates a distance image based of a pair of brightness images transmitted from a stereo camera. A road surface estimation part estimates a road surface based on the generated distance image. An object detection part groups together positions having a distance value located on/over the estimated road surface and satisfying a predetermined relationship, and detects an object based on the grouped positions. An abnormality detection part detecting occurrence of image abnormality in the brightness image captured during a current detection period. The road surface estimation part estimates the road surface at the current detection period based on the road surface estimated at the previous detection period when the abnormality detection part detects the occurrence of image abnormality in the brightness image captured during the current detection period.