LIDAR Point Cloud Wavelet Terrain Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing LIDAR systems face challenges in detecting obscured targets due to tree foliage and vegetation, which obstruct the view of real targets in 4D point cloud data, making it difficult to accurately reconstruct terrain profiles and distinguish between natural terrain and man-made objects.
Innovation Solution
The method involves receiving 3D point cloud data from a LIDAR system, reformating it into a 1D signal, applying a wavelet transform to decompose the signal, and then reconstructing a low-pass filtered profile to classify terrain features, while setting detail coefficients to zero to remove noise and distinguish between terrain and objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If LIDAR systems capture 4D point cloud data including vegetation and terrain, then comprehensive scene information is obtained, but obscured targets become difficult to detect due to foliage obstruction
Solution Approach 1:
The patent segments the point cloud data into different categories (terrain points, vegetation points, and object points) by analyzing height profiles and applying clustering algorithms. This segmentation allows the system to separate obscured targets from foliage obstructions, resolving the contradiction between capturing comprehensive scene information and detecting obscured targets.
Solution Approach 2:
The patent extracts vegetation points from the mixed point cloud data by identifying and removing points that belong to foliage based on height profile analysis and distance thresholds. This extraction process isolates the obscured targets from the vegetation obstructions, enabling clear target detection while preserving the complete scene information.
2Measurement precision
If wavelet transform decomposes the height signal to separate terrain from objects, then terrain profile accuracy is improved, but processing complexity increases
Solution Approach 1:
The patent replaces complex iterative terrain classification algorithms with a wavelet transform-based signal processing approach. By transforming the height signal into the frequency domain and applying thresholding in the wavelet coefficient space, the system achieves accurate terrain profile extraction with reduced computational complexity compared to traditional mechanical classification methods.
3Difficulty of detecting and measuring
If detail coefficients are set to zero to remove vegetation noise, then target visibility is enhanced, but high-frequency terrain features may be lost
Solution Approach 1:
The patent applies local quality by selectively processing different regions of the point cloud data. Instead of uniformly setting all detail coefficients to zero, the system identifies vegetation-dominated regions and applies denoising selectively in those areas while preserving high-frequency features in terrain regions. This localized approach enhances target visibility without losing important terrain feature information.
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
This approach effectively removes vegetation and other obstructions, allowing for accurate detection and tracking of obscured targets by preserving low-frequency content associated with terrain features and decimating non-ground features, thereby enhancing the visibility of bare earth and edge structures in real-time.
Implementation Method 1
One example of a 3D type sensing system is a Light Detection and Ranging (LIDAR) system. The LIDAR type 3D sensing systems generate data by recording multiple range echoes from a single pulse of light
Data Source
AI summary
A method for detecting terrain, through foliage, includes the steps of: receiving point cloud data in a three-dimensional (3D) space from an airborne platform, in which the point cloud data includes foliage that obscures the object; reformatting the point cloud data from the 3D space into a one-dimensional (1D) space to form a 1D signal; and decomposing the 1D signal using a wavelet transform (WT) to form a decomposed WT signal. The decomposed WT signal is reconstructed to form a low-pass filtered profile. The method classifies the low-pass filtered profile as terrain. The terrain includes a natural terrain, or a ground profile.