Evidential Illumination Mapping for Low-Visibility 3D Scene Assessment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing self-driving vehicles struggle to accurately assess the visibility and uncertainty of illuminated objects and areas in low visibility conditions, such as nighttime or hazy scenarios, leading to challenges in safe navigation.
Innovation Solution
A visibility grid is generated using an evidential illumination assessment based on camera images, converting color images to greyscale, determining intensity thresholds, and applying Ostu's method to separate well-lit and not-well-lit pixels, projecting uncertainty values onto a three-dimensional scan to enhance navigation safety.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If camera images are used to assess visibility in low visibility conditions, then the system can operate in nighttime or hazy scenarios, but the accuracy of visibility assessment deteriorates
Solution Approach 1:
The patent introduces an evidential illumination assessment as an intermediary layer between the raw camera image and the visibility determination. This assessment generates uncertainty values that mediate the interpretation of pixel intensities, allowing the system to operate in low visibility conditions while maintaining accurate visibility assessment by quantifying the reliability of each pixel's illumination state
Solution Approach 2:
The patent transforms the camera image from color space to greyscale space, changing the parameter representation from RGB values to intensity values. This parameter change enables the application of Ostu's method to separate well-lit and not-well-lit pixels, improving the accuracy of visibility assessment in low visibility conditions
2Device complexity
If traditional illumination assessment methods are used, then the processing is simple, but the uncertainty quantification is insufficient
Solution Approach 1:
The patent segments the image processing into distinct stages: color to greyscale conversion, intensity threshold determination using Ostu's method, pixel classification into well-lit and not-well-lit categories, and uncertainty value generation. This segmentation maintains processing simplicity while enabling comprehensive uncertainty quantification through the evidential illumination assessment
Solution Approach 2:
The patent implements feedback by using the evidential illumination assessment to generate uncertainty values that feed back into the visibility determination process. This feedback loop allows the system to adjust its confidence in visibility assessments based on the quality of illumination evidence, preventing loss of uncertainty information
Data Source
AI summary
Systems and methods are disclosed for identifying visibility of a physical space using visibility values and indications of uncertainty associated with the visibility values. One method can include capturing an image of a physical space, dividing pixels of the image into presumably well-lit and presumably not well-lit categories based on an intensity threshold, generating an evidential illumination map for the image based at least partly on a comparison between a value of each pixel to the intensity threshold, and projecting the evidential illumination map onto data representing a three-dimensional scan of the physical space. Evidential values can enable an autonomous vehicle to more safely navigate spaces using camera data by enabling programmatic determination of uncertainty for the camera data.