Crosswalk Edge Estimation Using Template Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing crosswalk marking estimation devices face accuracy issues when crosswalk markings are unclear due to wear or other factors, leading to decreased estimation precision.
Innovation Solution
A crosswalk marking estimating device that creates a planar-view road surface image using onboard sensors, acquires end candidates for band ends based on a template image, selects appropriate end candidates, and estimates the crosswalk marking edge position using these candidates, even when markings are unclear.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If edge points are extracted from road images to estimate crosswalk marking positions, then the estimation process can be implemented, but when crosswalk markings are unclear due to wear, edge points cannot be appropriately extracted and estimation accuracy decreases
Solution Approach 1:
The patent uses template images that are copies or models of typical crosswalk marking ends stored in advance. These templates are matched against the planar-view road surface image to acquire end candidates, replacing the traditional method of extracting edge points directly from worn or unclear markings. This copying approach allows accurate identification even when the actual crosswalk markings are degraded.
Solution Approach 2:
The patent performs preliminary actions by creating a planar-view road surface image through coordinate transformation before attempting to identify crosswalk markings. This preprocessing step transforms the image to a top-down view, making subsequent template matching more effective. Additionally, templates are prepared in advance and stored for quick comparison, enabling the system to handle unclear markings efficiently.
2Measurement precision
If template matching is used to acquire end candidates from planar-view road surface images, then crosswalk marking positions can be estimated even when markings are unclear, but multiple candidates including erroneous ones may be acquired requiring selection processes
Solution Approach 1:
The patent implements feedback by evaluating the distribution of acquired end candidates and using this information to select the most appropriate candidates. The system analyzes where end candidates are located in the image and selects those that form coherent patterns consistent with actual crosswalk markings. This feedback mechanism filters out erroneous candidates while maintaining high accuracy.
Solution Approach 2:
The patent applies local quality by selecting end candidates based on their specific distribution characteristics in the image. Rather than treating all candidates equally, the system evaluates each candidate's position and relationship to other candidates, selecting those that locally form valid crosswalk patterns. This approach handles complexity by making localized decisions about candidate validity based on spatial relationships.
Data Source
AI summary
A crosswalk marking estimating device includes: circuitry configured to: create a planar-view road surface image of a road around a vehicle; acquire end candidates which are candidates for ends of the bands, included in a crosswalk marking, in the planar-view road surface image based on the planar-view road surface image and a template image which is stored in advance; select a plurality of selected end candidates which is a set of the end candidates corresponding to an edge on the one side, in the road extending direction, of the crosswalk marking from a distribution of the end candidates in the road extending direction in the planar-view road surface image; and estimate a position of the edge of the crosswalk marking relative to the vehicle based on the selected end candidates.