Semantics-Based Landing Area Detection for UAVs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous landing area detection systems for unmanned aerial vehicles rely solely on 3D terrain data, which is insufficient for distinguishing between different terrain types such as water, grass, and sand, making it difficult for UAVs to select a safe landing site, especially in complex scenarios.
Innovation Solution
A semantics-based safe landing area detection algorithm that combines 3D perception systems with camera image information, co-registers and segments the data, classifies regions into semantic classes, and prioritizes contextual information to determine a suitable landing area.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If only 3D terrain data is used for landing area detection, then the system complexity is reduced, but the ability to distinguish between different terrain types (water, grass, sand) deteriorates
Solution Approach 1:
The patent combines 3D terrain data from LIDAR with 2D image data from cameras to create a multi-modal perception system. This merging of different data types enables the system to distinguish terrain types (water, grass, sand) that cannot be differentiated using only 3D information, thereby resolving the contradiction between system simplicity and discrimination capability.
Solution Approach 2:
The patent transitions from purely 3D terrain analysis to a 2.5D representation by co-registering 2D camera images with 3D LIDAR data. This dimensional integration allows the system to leverage both the depth information from LIDAR and the textural/semantic information from camera images, improving terrain type classification without excessive complexity increase.
2Measurement precision
If multiple data modalities (3D LIDAR and 2D camera images) are integrated, then the accuracy of landing zone identification is improved, but the device complexity increases
Solution Approach 1:
The patent segments the complex multi-modal data processing into distinct modules: LIDAR data processing, camera image processing, co-registration, segmentation, and classification. This modular segmentation allows each component to be optimized independently and simplifies the overall system architecture, reducing the complexity burden of multi-modal integration.
Solution Approach 2:
The patent introduces a co-registration module as an intermediary that aligns 3D LIDAR point clouds with 2D camera images in a common coordinate system. This intermediary component enables seamless integration of different data modalities by establishing spatial correspondence, thereby improving landing zone identification accuracy while maintaining manageable system complexity through structured data flow.
3Speed
If 3D terrain information alone is used, then the data processing speed is maintained, but the reliability of autonomous landing in complex scenarios deteriorates
Solution Approach 1:
The patent performs preliminary processing of LIDAR and camera data separately before integration, including point cloud filtering, feature extraction, and image preprocessing. This preliminary action prepares the data in advance for efficient co-registration and joint analysis, maintaining processing speed while enabling more reliable terrain classification through multi-modal information fusion.
Data Source
AI summary
A method for determining a suitable landing area for an aircraft includes receiving signals indicative of Light Detection And Ranging (LIDAR) information for a terrain via a LIDAR perception system; receiving signals indicative of image information for the terrain via a camera perception system; evaluating, with the processor, the LIDAR information and generating information indicative of a LIDAR landing zone candidate region; co-registering in a coordinate system, with the processor, the LIDAR landing zone candidate region and the image information; segmenting, with the processor, the co-registered image and the LIDAR landing zone candidate region to generate segmented regions; classifying, with the processor, the segmented regions into semantic classes; determining, with the processor, contextual information in the semantic classes; and ranking and prioritizing the contextual information.