|
In LiDAR system, atmospheric backscatter is one kind of important background radiation noise for target detection. When the intensity of atmospheric backscatter signal received by LiDAR system exceeds the detection threshold, the system will make a false alarm. To reduce the atmospheric backscatter interference efficiently, it is necessary to evaluate the importance of influencing factors of atmospheric backscatter. The intensity of atmospheric backscatter noise is mainly related to the parameters of transmitter, receiver, and atmospheric transport properties. We choose nine feature parameters in this study: transmitter features (including Wavelength (λ), Pulse Energy (E), and Divergence (D)), receiver features (including Detection Threshold (ITh), Field of View (FOV), and Responsiveness (Ri)), and atmospheric features (including Visibility (V), Asymmetric Factor (g), and Extinction-to-Backscatter Ratio (EBR)). Based on Mie scattering theory, we establish a LiDAR system atmospheric backscatter impact model with Monte Carlo method and set the false alarm rate as the indicator to evaluate the impact of atmospheric backscatter noise to LiDAR system, and next we assess the importance of these nine selected features by three feature selection methods (F-test, Neighborhood Component Analysis (NCA), and Bagging). The evaluation results prove that these three feature selection methods can successfully access the importance of thesee nine features. Although the importance values of the nine features evaluated by these three methods are not exactly all the same, the features belonged to the first-level, second-level, and third-level are consistent. The most important three features are ITh, g and V, which means atmospheric features are relatively important compared to the features of transmitter and receiver. The feature importance evaluation results can qualitatively provide guidance for LiDAR System design to avoid atmospheric backscatter effect and help improve the performance of LiDAR System.
|
