Road Defect Detection and Location From 360-Degree Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current approaches to road infrastructure management are inefficient, requiring manual surveying and data aggregation, which are not scalable to the number and extent of roads managed by transportation departments, limiting their effectiveness and efficiency.
Innovation Solution
A road defect and asset detection system that uses AI and ML models to automatically detect road defects and assets from 360-degree images, generating images and location data, and providing them to transportation departments for management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual surveying methods are used to detect road defects, then measurement precision can be maintained, but productivity is severely limited and loss of time increases
Solution Approach 1:
The patent replaces manual mechanical surveying with an automated computer vision system that uses machine learning models to detect road defects. The system processes images of road segments automatically, substituting human inspectors with algorithmic analysis to maintain precision while dramatically improving productivity.
Solution Approach 2:
The system creates visual copies of road segments through imaging and uses these copies for defect analysis. By working with image representations rather than physical inspection, the system can analyze multiple road segments simultaneously, enhancing productivity without compromising detection accuracy.
2Area of stationary object
If manual surveying is used to cover extensive road networks, then comprehensive coverage can be achieved, but loss of time and scalability are severely limited
Solution Approach 1:
The automated image processing system replaces manual surveying operations, enabling rapid analysis of extensive road networks. The system can process multiple images in parallel, covering large areas significantly faster than manual inspection while maintaining comprehensive defect detection.
Solution Approach 2:
The system performs preliminary defect detection through automated image analysis before physical repair actions are taken. This preliminary identification of defects across extensive road networks allows for efficient planning and prioritization of maintenance activities, reducing overall time loss.
3Productivity
If automated detection systems are implemented, then productivity increases, but device complexity increases
Solution Approach 1:
The patent employs a universal machine learning model architecture that can detect multiple types of road defects (cracks, potholes, signage issues) using the same system framework. This multi-functionality increases productivity across various defect types while managing device complexity through a unified approach rather than separate specialized systems.
Data Source
AI summary
An example method includes receiving multiple 360-degree images each of which include a portion of a road. Multiple images based on the multiple 360-degree images are generated. One or more road defect detection models are applied to the multiple images. A first defect in a first portion of the road in a first image and a second defect in a second portion of the road in a second image. A determination that the first defect and the second defect are part of a single defect is made. A first location for the first image and a second location for the second image is received. An estimate of one or more locations of the single defect is made based on the first location and the second location. Defect location data that indicates the one or more locations of the single defect is generated and provided.