In-Vehicle Camera Lens Dirt Detection Using Luminance Region Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image monitoring devices falsely detect dirt on in-vehicle camera lenses due to regions with smooth luminance values, even when no dirt is present, leading to unnecessary cleaning notifications.
Innovation Solution
An image monitoring device with a hardware processor that analyzes images captured by in-vehicle cameras, distinguishing between sky and ground regions based on luminance values and thresholds, and only notifies dirt presence when specific conditions are met, such as a high ratio of smooth regions and a significant difference in average luminance values between these regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a technique determines dirt adhesion on the lens based on the number of blocks with smooth luminance values, then the detection sensitivity is improved, but false detection occurs when smooth regions are formed without dirt present
Solution Approach 1:
The image is divided into multiple blocks, and the detection is performed by counting blocks with smooth luminance values. This segmentation approach allows the system to distinguish between localized smooth regions (potential dirt) and extensive smooth regions (environmental conditions), thereby reducing false detection while maintaining detection sensitivity.
Solution Approach 2:
The patent applies different detection criteria to different regions of the image. By analyzing the distribution and characteristics of smooth luminance blocks across various image regions, the system can differentiate between dirt-induced smooth regions and environment-induced smooth regions, improving reliability without sacrificing detection sensitivity.
2Measurement precision
If the detection threshold for smooth regions is lowered to increase detection sensitivity, then more dirt cases are detected, but false detection increases in clear conditions
Solution Approach 1:
The system uses feedback from the distribution pattern of smooth blocks across multiple image blocks to distinguish true dirt detection from false alarms. By analyzing whether smooth regions are isolated or widespread, and their spatial distribution characteristics, the system can adjust its interpretation of detection results, reducing false alarms while maintaining sensitivity.
Solution Approach 2:
The patent transitions from analyzing single-block luminance properties to analyzing the spatial distribution and quantity of smooth blocks across the entire image. This dimensional expansion from local to global analysis enables the system to differentiate between dirt (localized smooth regions) and environmental conditions (widespread smooth regions), reducing false detection frequency.
Data Source
AI summary
An image monitoring device includes a hardware processor. An image of the outside of a vehicle is captured by a camera. The hardware processor notifies that dirt adheres to a lens of the camera when a ratio of a smooth region in the image is more than a threshold. In the smooth region, differences in luminance value between pixels are small. The hardware processor does not notify that dirt adheres to the lens of the camera when: a ratio of the smooth region in a sky region in the image is more than a threshold, a ratio of the smooth region in a ground region in the image is more than a threshold, and a difference between an average of luminance values of the smooth region in the sky region and an average of luminance values of the smooth region in the ground region is more than a threshold.