Road Surface Extraction from Distance Histograms Under Sparse Points

Solution Overview

Problem

Existing image processing systems for obtaining movement surface information, such as those using stereo cameras, may produce inaccurate results when the number of movement surface candidate points is low or when the gradient and intercept of the approximate straight line do not fall within predetermined ranges, leading to inappropriate processing in later stages like object detection.

Innovation Solution

An image processing apparatus that generates disparity histogram information and segments it into multiple ranges, using collinear approximation and error condition checks to ensure accurate movement surface information by either using the extracted points or interpolating based on past data when error conditions are met.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If the number of movement surface candidate points is low or gradient and intercept are out of range, then the system enlarges disparity value range or executes correction processing, but this may still produce inaccurate movement surface information when there are many errors in extracted points

Engineering Contradiction:

Improveaccuracy of movement surface informationVSAvoiduse of inaccurate movement surface information in later processing

Core Design Contradiction:

ReliabilityVSLoss of information

Solution Approach 1:

The patent implements feedback by checking whether extracted movement surface candidate points satisfy predetermined conditions (number of points, gradient range, intercept range). When conditions are not met, the system performs correction processing and re-checks, creating a closed-loop feedback mechanism that ensures only accurate movement surface information is used in later processing, preventing propagation of errors

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent dynamically adjusts the disparity value range when initial extraction fails to meet conditions. By enlarging the search range and re-extracting points, the system adapts to difficult extraction conditions while maintaining quality control through subsequent condition checking and correction processing

Inventive Principle:

Principle #15Dynamics

2Ease of manufacture

If the system continuously uses extracted movement surface candidate points even when error conditions are satisfied, then processing can be simplified, but appropriate approximate straight line may be unable to be obtained and inaccurate information is used

Engineering Contradiction:

Improvesimplicity of processingVSAvoidaccuracy of movement surface information

Core Design Contradiction:

Ease of manufactureVSReliability

Solution Approach 1:

The patent implements feedback by checking whether extracted movement surface candidate points satisfy predetermined conditions (number of points, gradient range, intercept range). When conditions are not met, the system performs correction processing and re-checks, creating a closed-loop feedback mechanism that ensures only accurate movement surface information is used in later processing, preventing propagation of errors

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent dynamically adjusts the disparity value range when initial extraction fails to meet conditions. By enlarging the search range and re-extracting points, the system adapts to difficult extraction conditions while maintaining quality control through subsequent condition checking and correction processing

Inventive Principle:

Principle #15Dynamics

Data Source

PatentEP3432260B1Image processing device, moving body equipment control system, image processing program, and image processing method

Publication Date: 2023.10.18 RICOH CO LTD

Figure 1
Figure 2~3
Figure 4A~4B

AI summary

An image processing apparatus, which executes image processing of obtaining, based on distance image information, information on a movement surface in a region in front of a moving body in a movement direction thereof, has: a means that generates distance histogram information representing a frequency distribution of distance values in each of plural row regions obtained by division of a captured image in an up-down direction; and a means that: extracts a group of distance values matching a feature represented by distance values of the movement surface from distance values having frequencies exceeding a predetermined frequency threshold, for each of plural segments obtained by segmentation of the distance histogram information according to the distance values; executes, for a segment where a group of distance values extracted does not satisfy a predetermined error condition, processing of generating information on the movement surface using the group of distance values; and executes, for a segment satisfying the predetermined error condition, predetermined error processing, such as linear interpolation processing, without using the group of distance values.