Autonomous Vehicle Stop Line Recognition via Tracking and Neural Network
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous driving systems struggle to accurately recognize stop lines, especially in harsh environments like rainy or nighttime conditions due to poor visibility and external lighting effects, leading to reduced recognition accuracy and difficulty in maintaining vehicle speed.
Innovation Solution
A method utilizing a deep neural network-based stop line recognition system that combines a tracking algorithm with a trained neural network model to detect and re-detect stop lines in real-time, reducing the impact of external lighting and improving accuracy, even in challenging conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a hand-coded algorithm such as line detection is used to detect stop lines, then the stop line can be recognized with over 90% accuracy under normal conditions, but the recognition accuracy deteriorates significantly in harsh environments including nights and rainy days due to external lighting effects
Solution Approach 1:
The system pre-processes the input image by converting it to a specific color space and applying histogram equalization to enhance contrast and reduce lighting effects before detection. This preliminary processing prepares the data to be more robust against varying environmental conditions, allowing the same detection algorithm to perform well both in normal and harsh environments.
Solution Approach 2:
The system changes the parameter representation of the image by converting from RGB to HSV color space and adjusting histogram distribution. This transformation alters how lighting variations affect the detection, making the stop line features more distinguishable regardless of external lighting conditions, thereby improving both accuracy and environmental adaptability.
2Measurement precision
If real-time stop line detection is implemented using deep neural networks, then recognition accuracy can be significantly improved, but the processing speed may decrease making it difficult to maintain high vehicle speeds
Solution Approach 1:
The system segments the detection task into two parts: a fast hand-coded algorithm for initial detection and a deep neural network for verification. This segmentation allows the system to benefit from both the speed of traditional algorithms and the accuracy of neural networks, maintaining high vehicle speeds while improving recognition accuracy.
Solution Approach 2:
The system applies the computationally intensive deep neural network only when the hand-coded algorithm detects a stop line candidate, rather than processing every frame through the neural network. This partial application of the heavy computation maintains processing speed while still achieving high accuracy when needed.
Data Source
AI summary
A method of recognizing a stop line in an autonomous vehicle is disclosed. The method includes detecting valid stop line data in a current frame of an input image, when the valid stop line data is detected in the current frame, calculating a stop line area in the current frame using a tracking algorithm and tracking the stop line in a next frame, and when the valid stop line data is not detected, inputting the current frame to a trained neural network model and performing a redetection of the stop line data.