Road Surface Image Analysis for Curb and Gutter Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Drivers face challenges when navigating narrow roads with curbs and gutters, as there is a risk of wheels contacting the curb or falling into the gutter due to careless driving, particularly when pulling over to pass oncoming vehicles.
Innovation Solution
An image analysis apparatus that captures road surface images, performs area division using Semantic Segmentation, estimates depth using CNN-Depth, and detects irregularities such as curbs and gutters, providing real-time alerts and steering interventions to prevent wheel contact.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a driver pulls over to the side edge of a narrow road to pass an oncoming vehicle, then the vehicle can successfully navigate the narrow road, but there is a risk that the wheels may come into contact with the curb or fall into the gutter due to careless driving
Solution Approach 1:
The system performs preliminary detection of curbs and gutters on the road surface before the vehicle reaches them. By identifying these irregularities in advance and notifying the driver, the system enables proactive steering adjustments to avoid contact, thus resolving the contradiction between navigation efficiency and driving safety
Solution Approach 2:
The system continuously captures images of the road surface, processes them to detect irregularities, and provides real-time feedback to the driver through notifications. This closed-loop feedback mechanism allows the driver to adjust steering based on actual road conditions, maintaining both navigation efficiency and safety
2Measurement precision
If the apparatus detects all edges in the captured image, then comprehensive edge information is obtained, but edges corresponding to paint display drawn on the road are incorrectly identified as three-dimensional obstacles
Solution Approach 1:
The system applies different analysis criteria to different regions of the captured image. By performing area division to identify road surface regions and analyzing depth information specifically for these regions, the system can distinguish between paint displays (which have different depth characteristics) and actual three-dimensional obstacles, thus improving detection accuracy while reducing false positives
Solution Approach 2:
The system introduces depth estimation as an intermediary parameter to differentiate between paint displays and three-dimensional obstacles. By combining image data with depth information, the system can identify that paint displays have minimal depth variation compared to actual obstacles, thereby resolving the contradiction between comprehensive detection and false positive reduction
Data Source
AI summary
An image analysis apparatus according to the present disclosure includes: an area division unit that performs area division on an image obtained by capturing a road surface, and generates a result of the area division in which the image is classified into image areas of a plurality of categories; a depth estimation unit that performs depth estimation on the image and generates depth estimation information; and a detection unit that detects irregularities present on the road surface based on the result of the area division and the depth estimation information. The plurality of categories include a road category, and the detection unit detects irregularities present on the road surface within the image area classified in the road category.