Road Surface Anomaly Detection Using Adaptive Normal Ranges
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anomaly detection systems for autonomous vehicles struggle to accurately identify obstacles of indefinite color, shape, and size on the road surface, which can prevent safe travel, as they rely on predefined classifiers and reference images.
Innovation Solution
An anomaly detection device that extracts features from images using a trained feature extractor, detects abnormal conditions by comparing them to a normal range, and modifies this range based on the distribution of features from images taken during normal travel to improve detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a classifier for detecting predetermined objects is used, then detection of known objects is improved, but detection of obstacles with indefinite color, shape, and size deteriorates
Solution Approach 1:
The system changes from using fixed predefined object classifiers to dynamically adjusting the normal range parameters based on actual road conditions. By continuously adapting the parameter ranges for color, shape, and size based on collected road surface data, the system can detect both known objects and unexpected obstacles with indefinite characteristics.
Solution Approach 2:
The anomaly detection device transitions from static detection thresholds to dynamic adaptation. The normal range is continuously updated based on the distribution of features from images taken during normal travel, enabling the system to adapt to varying road conditions and detect anomalies that deviate from the learned normal patterns.
2Productivity
If a fixed normal range is used for anomaly detection, then detection speed is improved, but detection accuracy under varying road conditions deteriorates
Solution Approach 1:
The system performs preliminary collection of road surface features during normal travel to establish the normal range before actual anomaly detection begins. This preliminary action enables the system to have pre-computed detection thresholds ready, maintaining fast detection speed while ensuring accuracy is optimized for the specific road conditions.
Solution Approach 2:
The system implements feedback by continuously monitoring the distribution of normal travel features and using this information to modify and refine the normal range. This feedback loop ensures that the detection thresholds remain accurate under varying road conditions while maintaining efficient detection performance.
Data Source
AI summary
An anomaly detection device includes a processor configured to extract a feature indicating the condition of a road surface by inputting an image representing surroundings of a vehicle into a feature extractor that has been trained to extract the feature, detect an abnormal condition in which the vehicle is unable to travel normally, when the feature is outside a normal range that is a tolerable range in which the vehicle is able to travel normally, and modify the normal range, based on the distribution of normal travel features each indicating the condition of a road surface and extracted from the respective images obtained while the vehicle travels normally.