Lane Recognition System Noise Suppression via Synthesized Bird's-Eye Image
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing lane recognition systems face misrecognition issues due to noise generated by raindrops or foreign matters on the image input device, leading to inaccurate lane detection, especially when transforming original images to bird's-eye images or detecting lanes directly from original images.
Innovation Solution
A lane recognition system that creates a synthesized bird's-eye image by connecting multiple bird's-eye images from different times and uses both original and synthesized images to detect lane line candidates, thereby offsetting noise and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a bird's-eye image is created from an original image by integrating luminance, then lane detection can be performed, but noise is generated that cannot be discriminated from the lane, leading to misrecognition
Solution Approach 1:
The patent segments the image processing into multiple independent modules: original image processing module for detecting lane candidates, bird's-eye image processing module for verifying lane candidates, and synthesized image processing module for final confirmation. This segmentation allows each module to handle specific aspects of noise filtering and lane detection independently, improving overall accuracy while managing noise effectively.
Solution Approach 2:
The patent merges the advantages of both original image processing and bird's-eye image processing by creating a synthesized image that combines information from both sources. The synthesized image processing module uses both the original image and bird's-eye image to detect and verify lane lines, thereby combining the strengths of both approaches while mitigating their respective weaknesses and noise issues.
2Device complexity
If lane detection is performed directly from original images, then processing is simpler, but misrecognition occurs due to noise from raindrops or foreign matters
Solution Approach 1:
The patent introduces a synthesized image as an intermediary that combines information from both original images and bird's-eye images. This intermediary synthesized image serves as a mediator that filters out noise while preserving lane information, allowing the system to achieve high detection accuracy without the complexity of processing multiple independent image types separately.
Solution Approach 2:
The patent creates a composite processed image by integrating information from both original images and bird's-eye images. This composite image combines the clarity of original images with the noise-reducing properties of bird's-eye images, achieving superior lane detection accuracy while maintaining manageable processing complexity through integrated processing.
3Measurement precision
If multiple images from different times are connected to create synthesized bird's-eye image, then noise is suppressed and accuracy is improved, but processing time and complexity increase
Solution Approach 1:
The patent performs preliminary processing by creating bird's-eye images from original images in advance, and then uses these pre-processed images to create the synthesized image. This preliminary action allows the system to reduce the computational burden during final lane detection, as the heavy lifting of image transformation has already been completed, thereby reducing overall processing time while maintaining high accuracy.
Data Source
AI summary
To provide a lane recognition system which can improve the lane recognition accuracy by suppressing noises that are likely to be generated respectively in an original image and a bird's-eye image. The lane recognition system recognizes a lane based on an image. The system includes: a synthesized bird's-eye image creation module which creates a synthesized bird's-eye image by connecting a plurality of bird's-eye images that are obtained by transforming respective partial regions of original images picked up at a plurality of different times into bird's-eye images; a lane line candidate extraction module which detects a lane line candidate by using information of the original images or the bird's-eye images created from the original images, and the synthesized bird's-eye image; and a lane line position estimation module which estimates a lane line position based on information of the lane line candidate.