Road Sign Recognition Using Text Probability Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing road sign recognition methods in digital image data suffer from a high false-positive rate, particularly for rectangular structures, due to the frequent occurrence of non-sign rectangular objects like advertising posters in urban environments.
Innovation Solution
A method that segments potential road signs, identifies and evaluates text within the segmented data using OCR or ICR methods to determine a probability value for the text being on a road sign, and applies a classifier only if the probability exceeds a threshold, thereby reducing false positives.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a two-step approach with geometric shape detection and pictogram classification is used for road sign recognition, then the classification rate for circular traffic signs is good, but the false-positive rate for rectangular structures is markedly higher
Solution Approach 1:
The patent segments the road sign detection process into multiple independent stages: geometric shape detection, text region segmentation, text recognition, and classification. By dividing the rectangular sign detection into these segments, the system can filter out false positives (like advertising posters) at the text recognition stage while maintaining true positive detection, thus reducing the false-positive rate without sacrificing classification accuracy
Solution Approach 2:
The patent introduces text recognition as an intermediary step between geometric shape detection and final classification. This intermediary text analysis stage acts as a filter that verifies whether detected rectangular structures actually contain road sign text, thereby reducing false positives from non-sign rectangular objects while preserving true road sign detections
2Object-generated harmful factors
If text recognition methods are applied to all segmented objects before classification, then the false-positive rate is reduced, but the computational demands and processing time increase
Solution Approach 1:
The patent applies text recognition selectively rather than universally - only to segmented objects that match rectangular geometric patterns and are located in regions where road signs are expected. This partial application of text recognition reduces computational overhead compared to applying it to all detected objects, while still effectively reducing the false-positive rate for rectangular structures
Solution Approach 2:
The patent performs preliminary geometric shape detection and filtering before applying text recognition. By pre-segmenting objects based on shape characteristics and spatial location, the system reduces the number of candidates requiring computationally intensive text recognition, thereby minimizing processing time while maintaining false-positive reduction effectiveness
Data Source
AI summary
A method and a device are provided for recognizing road signs in image data. The method includes, but is not limited to segmenting an object in the image data that is a road sign for a predefined probability. A text mapped in the segmented image data is identified using a text recognition method, where this text comprises numbers and/or words and/or abbreviations and/or combinations thereof. A probability value is determined for the text being depicted on a road sign and, in case the probability value is smaller than or equal to a predefined threshold value, is selected as a potential road sign. In case the probability value is greater than the predefined threshold value, a classifier is applied to the segmented image data for recognizing the object as an actual road sign.