Lidar full-waveform decomposition based on empirical mode decomposition and local-Levenberg-Marquard fitting

Area of Science:

• Geospatial technology
• Signal processing
• Remote sensing

Background:

• Full-waveform Light Detection and Ranging (LIDAR) data provides rich information but requires sophisticated processing for accurate interpretation.
• Traditional methods like the zero-crossing (ZC) technique can be sensitive to noise, limiting precise echo decomposition.
• Decomposing complex LIDAR echoes into Gaussian components is crucial for extracting detailed surface information.

Purpose of the Study:

• To propose and validate a new method for decomposing full-waveform LIDAR echoes.
• To enhance the accuracy and robustness of component detection and parameter estimation in LIDAR data.
• To improve the anti-noise performance compared to existing echo decomposition techniques.

Main Methods:

• Utilizing Empirical Mode Decomposition (EMD) to decompose the full-waveform echo into intrinsic mode functions (IMFs) and a residual.
• Calculating average period and energy densities (EDs) of IMFs to select a suitable IMF based on noise spread lines.
• Employing local and global Levenberg-Marquardt (LM) fitting for precise initial parameter estimation of Gaussian components.

Main Results:

• The proposed EMD-LM method successfully decomposes full-waveform echoes into Gaussian-like components.
• Accurate detection of echo components and precise estimation of their initial parameters were achieved.
• The method demonstrated superior anti-noise performance compared to the traditional zero-crossing (ZC) method.

Conclusions:

• The developed EMD-LM method offers a robust and accurate approach for full-waveform LIDAR echo decomposition.
• This technique significantly improves the reliability of LIDAR data analysis, especially in noisy environments.
• Validation with synthetic, recorded LIDAR, and Land, Vegetation, and Ice Sensor data confirms the method's effectiveness.