Lane Marking Detection Using Key Frame Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing lane marking detection methods face inefficiencies due to the need for feature extraction and convolution calculations on every image frame, leading to low detection efficiency and high computational costs.
Innovation Solution
A method that extracts key image frames at a predetermined interval from a video stream, using these frames to detect lane markings, and employs feature maps from previous key frames for non-key image frames, reducing the need for extensive calculations and improving detection efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If feature extraction and convolution calculations are performed on every image frame, then detection accuracy is maintained, but detection efficiency decreases and computational cost increases
Solution Approach 1:
The patent segments video frames into key frames and non-key frames based on motion detection. Only key frames undergo full feature extraction and convolution calculations, while non-key frames use simplified processing. This segmentation resolves the contradiction by maintaining accuracy on important frames while improving overall efficiency through selective processing.
Solution Approach 2:
The patent applies partial action by performing complete feature extraction only when necessary (on key frames with motion changes), and using partial/simplified processing for routine frames. This reduces unnecessary computational effort while maintaining detection accuracy when it matters most.
2Reliability
If feature extraction is performed on every image frame, then complete lane marking information is captured, but computational cost increases
Solution Approach 1:
The patent performs preliminary motion detection and frame comparison before deciding whether to perform full feature extraction. This preliminary action identifies which frames require intensive processing, reducing overall computational cost while ensuring reliable detection when lane markings change.
Solution Approach 2:
The patent applies different processing qualities to different frames based on their importance. Key frames with motion changes receive full processing for reliable detection, while stable frames receive minimal processing. This local quality approach maintains detection reliability where needed while reducing computational cost overall.
Data Source
AI summary
A lane marking detecting method, an apparatus, an electronic device, a storage medium, a program product, and a vehicle, where a specific implementation includes: obtaining a video stream including a lane marking; extracting a key image frame from the video stream at a predetermined interval of frames, with an image frame between adjacent key image frames being a non-key image frame; detecting, for the key image frame, a lane marking according to a feature map of the key image frame; detecting, for the non-key image frame, a lane marking according to the non-key image frame and a feature map of a previous key image frame of the non-key image frame. In this embodiment, the feature map of the non-key image frame is determined and the lane marking is detected by combining the feature map of the previous key image frame, thus enabling flexibility and diversity in lane marking detection.