Vehicle Light Classification Under Occlusion for AV Path Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous vehicles face challenges in accurately and efficiently classifying vehicle lights due to varying configurations, occlusions, and environmental factors, which affect decision-making and safety in dynamic driving environments.
Innovation Solution
A system and method for vehicle light classification using machine learning models, leveraging multiple sensors and data processing techniques to identify and classify vehicle lights, including deep learning models like convolutional neural networks, to enhance perception and decision-making in autonomous vehicles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional sensor-based detection methods are used to identify vehicle lights, then the system can detect basic light presence, but the classification accuracy deteriorates due to varying light configurations, occlusions, and environmental factors
Solution Approach 1:
The system performs preliminary actions by capturing multiple images of the same scene from different angles and positions before classification. This allows the machine learning model to have access to multiple views of the vehicle lights, reducing the impact of occlusions and environmental factors on classification accuracy
Solution Approach 2:
The system creates multiple copies of the light detection task by using multiple sensors and capturing multiple images. Each sensor and image serves as a copy that can independently detect and classify vehicle lights, with the final classification being aggregated from all copies to improve accuracy and robustness against occlusions
2Reliability
If multiple sensors and complex data processing techniques are deployed to improve light classification, then classification reliability improves, but system complexity increases
Solution Approach 1:
The system segments the light classification task into multiple independent detection processes, each handled by individual sensors or image processing streams. This segmentation allows the system to improve reliability through redundancy while managing complexity by keeping each segment relatively simple and modular
Solution Approach 2:
The machine learning model serves multiple functions by simultaneously classifying different types of vehicle lights (headlights, taillights, turn signals, brake lights) and handling various environmental conditions. This multi-functionality improves reliability across diverse scenarios while avoiding the need for separate specialized systems for each light type
3Measurement precision
If machine learning models are used to classify vehicle lights, then classification accuracy improves, but processing time and computational resources increase
Solution Approach 1:
The system performs preliminary action by capturing multiple images in advance before the actual classification decision is needed. This allows the machine learning model to process pre-captured data rather than requiring real-time processing, reducing the perceived processing time while maintaining high accuracy
Solution Approach 2:
The system skips unnecessary processing steps by using efficient machine learning models that can quickly classify vehicle lights from the captured images. The pre-processing and feature extraction are optimized to rush through the computational steps rapidly, minimizing processing time while maintaining classification accuracy
Data Source
AI summary
The described aspects and implementations enable vehicle light classification in autonomous vehicle (AV) applications. In one implementation, disclosed is a method and a system to perform the method that includes, obtaining, by a processing device, first image data characterizing a driving environment of an autonomous vehicle (AV). The processing device may identify, based on the image data, a vehicle within the driving environment. The processing device may process the image data using one or more trained machine-learning models (MLMs) to determine a state of one or more lights of the vehicle and cause an update to a driving path of the AV based on the determined state of the lights.