Abnormal Image Detection Device for Vehicle Stereo Cameras
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing abnormality diagnosis devices for stereo cameras in vehicles struggle to accurately detect abnormalities, particularly when extraneous materials are attached to one camera, leading to incorrect determination of parallax image data as normal, especially under severe brightness/darkness conditions.
Innovation Solution
An abnormal image detection device that determines pixel reliability based on edge intensity and parallax value, counting pixels with low reliability to identify abnormal parallax images, even when images are taken from different scenes by each camera, using a reliability determination part, computation part, and abnormality determination part to set thresholds and determine abnormality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing abnormality diagnosis devices use parallax value distribution and amount of change determination methods, then they can detect some abnormalities, but they fail to accurately detect abnormalities when extraneous materials are attached to one camera
Solution Approach 1:
The patent divides the image into multiple regions and performs edge intensity determination for each region separately. By segmenting the image processing into region-based edge detection and pixel-based reliability determination, the system can identify local abnormalities caused by extraneous materials more effectively than global parallax distribution analysis.
Solution Approach 2:
The patent performs edge intensity determination and reliability assessment on original images before parallax computation. By preliminarily identifying pixels with low edge intensity (likely obscured by extraneous materials) and marking them as low reliability, the system prevents these pixels from corrupting the parallax image data in the first place.
2Measurement precision
If the system processes all pixels in parallax images without reliability filtering, then computation is simpler, but detection accuracy decreases due to inclusion of low-reliability pixels
Solution Approach 1:
The system performs reliability determination on pixels before parallax computation by analyzing edge intensity in original images. Pixels with low edge intensity are preliminarily marked as low reliability, so that when parallax images are later computed, these pixels can be excluded or weighted appropriately, improving obstacle detection accuracy without requiring complex post-processing.
Solution Approach 2:
The patent applies different processing quality levels to different pixels based on their reliability. High-reliability pixels (with strong edge intensity) undergo full parallax processing, while low-reliability pixels (with weak edge intensity likely obscured by extraneous materials) are excluded or handled differently. This local differentiation improves overall detection accuracy without uniformly increasing processing complexity for all pixels.
Data Source
AI summary
An abnormal image detection device includes a reliability determination part that determines whether or not each pixel of a parallax image has a low reliability. The reliability determination part has an edge intensity determination section. For each respective pixel of original images, the edge intensity determination section determines whether or not the respective pixel is smaller than or equal to a threshold value, and if the respective pixel is smaller than or equal to the threshold value, the reliability determination part determines that a corresponding one of the pixels in the parallax image, which corresponds to the respective pixel, is a pixel having the low reliability. A computation part computes a total number of pixels of the parallax image that each has the low reliability. An abnormality determination part that when the total number of pixels exceeds a predetermined value, determines that the parallax image is abnormal.