2D LiDAR-Camera SLAM with Feature Point Retention Beyond Camera View
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
SLAM based on 2D Lidar faces issues with creating incorrect maps due to feature points out of the camera's narrow field of view being detected as non-existent by the wide viewing angle of the Lidar, potentially harming driving safety.
Innovation Solution
A method and system that maintains feature points detected by a camera with a narrow viewing angle in conjunction with a Lidar sensor's wide viewing angle, involving a controller to remove feature points exceeding a set distance and add newly searched points, ensuring robustness and reducing memory usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If SLAM is performed based on 2D Lidar with wide viewing angle, then the viewing coverage is improved, but feature points may be detected as non-existing due to areas not at laser height not being recognized, resulting in incorrect map construction
Solution Approach 1:
The patent combines 2D Lidar and camera data to perform SLAM. The controller integrates feature points from both sensors, using camera feature points to supplement areas where 2D Lidar cannot detect (not at laser height), thereby maintaining reliable map construction while preserving the wide viewing coverage advantage of 2D Lidar.
Solution Approach 2:
The camera acts as an intermediary sensor to verify and supplement feature points detected by the 2D Lidar. When the controller detects feature points from 2D Lidar, it uses the camera's narrow-angle verification to confirm their validity, preventing false detections and ensuring map accuracy.
2Measurement precision
If camera feature points are used for SLAM, then measurement precision is improved, but feature points out of camera field of view are lost when vehicle moves, resulting in incorrect map creation
Solution Approach 1:
The controller continuously compares feature points from the camera with those from the 2D Lidar in real-time. When the vehicle moves and camera feature points go out of field of view, the system receives feedback from the 2D Lidar to identify and retain valid feature points, ensuring continuous and accurate map creation throughout vehicle movement.
Solution Approach 2:
The controller preliminarily identifies feature points using the 2D Lidar's wide viewing angle before the vehicle moves. This preliminary detection ensures that feature points are pre-identified and can be retained even when they move out of the camera's narrow field of view during vehicle movement.
3Reliability
If all detected feature points are maintained in point cloud, then map accuracy is improved, but memory usage and computational load increase
Solution Approach 1:
The controller extracts and retains only valid feature points by comparing data from both 2D Lidar and camera. Invalid or redundant feature points are removed, maintaining map accuracy while reducing the quantity of data stored in the point cloud, thereby optimizing memory usage and computational efficiency.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Secures robustness in SLAM by maintaining relevant feature points, reducing memory usage, and computational load while correcting map inaccuracies.
Implementation Method 1
a light detection and ranging (Lidar) sensor configured to detect a feature point in front of a vehicle
Implementation Method 2
a camera configured to acquire a front image of the vehicle
Data Source
AI summary
A system for simultaneous localization and mapping (SLAM) includes a Lidar sensor configured to detect a feature point in front of a vehicle and a camera configured to acquire a front image of the vehicle. The system also includes a controller configured to receive data on the feature point from the Lidar sensor and the front image from the camera, search for the feature point from the front image, set the feature point received from the Lidar sensor and the feature point searched from the front image as a point cloud (PCL), and perform SLAM based on the feature point in the PCL. The controller is further configured to remove a feature point satisfying a removal condition from the PCL in response to an arrival of an object search cycle, and add a newly searched feature point after a previous object search cycle to the PCL.