Autonomous Vehicle Vision Control Under Mixed Lighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in reliably detecting obstacles under varying light conditions, particularly when using image processing systems, as existing methods often result in artifacts due to combining images captured with different illumination settings, leading to poor performance in mixed lighting environments.
Innovation Solution
The system captures two images with different settings in close time relationship, processing each separately with intensity thresholds to exclude unreliable pixels, avoiding the combination of images and thus minimizing artifacts, and using a stereo camera or bracketing technique to ensure accurate obstacle detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Illumination intensity
If images are captured with different camera settings to illuminate dark areas and avoid overexposure in bright areas, then the coverage of visible areas is improved, but artifacts occur at boundaries between differently illuminated areas
Solution Approach 1:
The image data is segmented into multiple depth layers based on intensity thresholds. Pixels are divided into first image data (below first threshold) and second image data (above second threshold), allowing separate processing of different illumination zones without contamination from artifacts at boundaries
Solution Approach 2:
Different intensity thresholds are applied to different regions of the image based on local illumination conditions. The system adapts the processing parameters locally - using first threshold for dark areas and second threshold for bright areas - rather than applying a single global threshold, thereby optimizing obstacle detection for each local condition
2Adaptability or versatility
If a single HDR image is generated by combining images with different camera settings, then both dark and bright areas are visible, but artifacts occur particularly at boundaries between dark and bright areas
Solution Approach 1:
The harmful artifacts are extracted and removed by applying intensity thresholds that exclude problematic pixels. The system extracts only the reliable pixel data from each image (below first threshold for dark areas, above second threshold for bright areas) and processes these extracted sets separately, eliminating the need to handle artifact-contaminated combined HDR images
Solution Approach 2:
Instead of combining images first and then trying to remove artifacts, the invention inverts the approach by filtering and processing images separately based on intensity thresholds before any combination occurs. This reverse sequence - filter then combine rather than combine then filter - prevents artifact generation in the first place
3Illumination intensity
If exposure time is increased to capture sufficient light in evening conditions, then image intensity is improved, but obstacles in bright areas become overexposed and harder to detect
Solution Approach 1:
The system dynamically adjusts processing parameters based on local image conditions. By applying different intensity thresholds to different regions (first threshold for dark areas, second threshold for bright areas), the system adapts the detection criteria to match local illumination conditions, maintaining precision across varying light environments
Solution Approach 2:
The invention changes the intensity threshold parameter based on local illumination conditions. Rather than using a fixed threshold, the system employs multiple thresholds (first and second intensity thresholds) that are selected based on whether the local area is dark or bright, thereby optimizing detection precision for each condition
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a control system for an autonomous vehicle, a method and an autonomous vehicle. The system comprises an image capturing means capable of capturing at least a first image of the environment of the vehicle and a second image of the environment, wherein the images are captured in a close time relationship but with different image capturing parameters. A processing means configured to obtain and process the images captured with different image capturing parameters separately and taking into consideration a first intensity threshold when processing the first image and a second, different intensity threshold when processing the second image. A control means for generating and outputting a control signal on the basis of a result of the at least one of the processed images.