Vision System Road Marking Detection Edge Classifier
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vision systems for motor vehicles face challenges in accurately and efficiently detecting road surface markings, especially under adverse conditions such as faded lines, light-colored road surfaces, and varying weather, which leads to computationally demanding and unreliable edge detection.
Innovation Solution
A vision system that combines edge detection with a road classifier to calculate the probability of image elements belonging to road markings or drivable surfaces, using a convolutional neural network for classification and a filter to remove edges with low probability, thereby improving detection efficiency and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If edge detection is performed on all edges in captured images, then road surface markings can be detected, but computational resources are excessively consumed and false positives increase
Solution Approach 1:
The patent segments the image processing task into two independent parallel streams: edge detection and road surface classification. The edge detector identifies potential edges while the road surface classifier independently classifies picture elements as drivable or non-drivable. This segmentation allows each module to focus on its specific function, reducing overall computational complexity while maintaining detection reliability.
Solution Approach 2:
The patent introduces a road surface classifier as an intermediary component that provides probability information about drivable surfaces. This intermediary module filters the search space for edge detection by identifying regions that are likely to contain road markings, thereby reducing the number of edges that need to be processed and evaluated for potential road surface markings.
2Productivity
If traditional edge detection methods are used without classification, then processing is simpler, but detection accuracy deteriorates under adverse conditions
Solution Approach 1:
The patent performs preliminary road surface classification before final road marking detection. The road surface classifier pre-identifies drivable and non-drivable regions, creating a probability map that guides subsequent edge evaluation. This preliminary action filters out regions unlikely to contain road markings, allowing the system to maintain high processing speed while improving detection accuracy in adverse conditions.
Solution Approach 2:
The patent changes the parameter space by introducing probability values for each picture element being drivable or non-drivable. Instead of binary edge detection, the system evaluates edges based on the probability context provided by the road surface classifier. This parameter transformation allows the system to adapt detection sensitivity based on local road surface characteristics, improving accuracy without sacrificing processing efficiency.
Data Source
Figure 1
Figure 2
AI summary
A vision system and method for autonomous driving and/or driver assistance in a motor vehicle are presented. The vision system (10) comprises an imaging apparatus (11) adapted to capture images (6) from a surrounding of the motor vehicle, and a data processing apparatus (14) adapted to perform image processing on the images (6) captured by the imaging apparatus (11). The data processing apparatus (14) is adapted to execute an edge detector (90) adapted to detect edges (100) in the captured images (6). The data processing apparatus is adapted to execute a road classifier (110) adapted to calculate for each picture element of an input image (50) the probability that the picture element belongs to any one of the classes road marking and drivable surface, yielding a road probability image (120). The data processing apparatus (14) is adapted to combine the results from said edge detector (90) and from said road classifier (110) in order to detect road markings in images (6) captured by said imaging apparatus (11).