Road Surface Detection Using Disparity Frequency Distribution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for detecting road surfaces using disparity images often result in erroneous identification of objects other than the road surface, leading to potential false alarms or braking errors in vehicle safety systems.
Innovation Solution
An image processing apparatus that generates a frequency distribution of disparity values for each vertical position in a disparity image, selects candidate points based on buffer ranges with the highest frequency sums, and extracts sample points to accurately detect the road surface.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If road surface detection is performed using disparity image frequency histograms, then road surface detection capability is enabled, but detection accuracy deteriorates due to erroneous identification of non-road surface objects
Solution Approach 1:
The patent applies local quality by analyzing frequency distributions separately for different spatial regions (upper and lower halves) of the disparity image. Instead of treating the entire image uniformly, the method divides the detection space into local regions and processes each region independently to identify road surface-specific disparity patterns, thereby improving detection accuracy while maintaining reliability
Solution Approach 2:
The patent segments the disparity image frequency analysis into multiple buffer ranges with different disparity value intervals. By dividing the frequency histogram analysis into segmented disparity ranges and selecting candidate points from specific segments, the method distinguishes road surface disparities from non-road surface objects more effectively, resolving the contradiction between detection capability and accuracy
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An image processing apparatus includes: a generating unit configured to generate, from a distance image including distance values, each corresponding to a distance to a road surface, obtained from a plurality of captured images respectively captured by a plurality of image capturing units, vertical distribution data for indicating frequency distribution of distance values in the distance image; a selecting unit configured to select candidate points based on a density of frequencies of the distance values in the vertical distribution data; and a detecting unit configured to detect a road surface based on the sample points extracted from the candidate points that have been selected by the selecting unit.