Curb Detection via Image Recognition for Autonomous Vehicles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing curb detection methods using lidar and ultrasonic radar suffer from low reliability due to small reflection areas and low detection resolution, making it difficult to accurately determine curb shape and distance.
Innovation Solution
A curb detection method utilizing image recognition to identify curb edges and seams, enabling the extraction of geometry, distance, and height information through image processing without relying on reflected signals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If lidar is used to detect curb based on reflected laser signals, then detection can be performed, but the reflection area of laser signals by the curb is small, resulting in few reflected laser signals received and low reliability of curb detection
Solution Approach 1:
The patent replaces the lidar-based optical detection system with an image recognition system using cameras. Instead of relying on laser signal reflection from the small curb surface, the system uses visual imaging to capture and analyze curb features, thereby avoiding the fundamental limitation of small reflection area and improving both reliability and precision of curb detection
Solution Approach 2:
The patent creates a visual copy of the curb through image capturing and uses this optical copy for analysis. The image recognition system processes the captured images to extract curb information, transforming the physical detection problem into an information processing problem that can be solved through pattern recognition algorithms
2Reliability
If image recognition is used to detect curb, then reliability of curb detection is improved, but comprehensive detection information including geometry, distance, and height may be required
Solution Approach 1:
The patent makes the image recognition system multi-functional by enabling it to simultaneously perform multiple detection tasks: geometry recognition, distance measurement, and height estimation. A single image capturing apparatus and recognition system provides all these functions, avoiding the need for separate detection devices for each parameter and reducing overall system complexity
Solution Approach 2:
The patent merges multiple detection functions into a unified image recognition framework. By combining geometry detection, distance measurement, and height estimation into one integrated system that processes images from a single camera, the patent reduces the complexity that would otherwise arise from coordinating multiple separate detection systems
Data Source
Figure 1~2
Figure 3~5
Figure 6~7
AI summary
This application discloses a curb detection method and apparatus, a system, a computer-readable storage medium, and a vehicle, relating to the field of autonomous driving technologies. The method includes: obtaining at least one to-be-detected image shot by an image capturing apparatus located on a vehicle, recognizing the at least one to-be-detected image to obtain a recognition result, and obtaining detection information of a curb based on the recognition result. The detection information includes at least one of the following: geometry information, distance information, or height information of the curb. In the method, curb detection is performed through image recognition, and reliability of curb detection is relatively high because the curb detection is performed without relying on a reflected signal. In addition, because more comprehensive detection information can be obtained, a parking policy or a driving route that is subsequently selected or planned based on the detection information of the curb has relatively high adaptability to the curb, thereby improving the accuracy of the parking policy or the driving route.