Natural Scene Recognition Based on HRRP Statistical Modeling

Natural scene classification based on high resolution one-dimensional range profile (HRRP) has significant value in the field of target recognition and environmental monitoring. Statistical modeling of HRRP has been widely used to extract useful information from clutter-like signals. However, it is still difficult to distinguish manually based on the extracted parameters. This paper proposes an one-dimensional convolutional neural network (1D-CNN) to automatically recognize natural scenes based on the statistical parameters. generalized Gamma distribution ( $\mathrm{G}\Gamma\mathrm{D}$ ) are used to model the HRRP data and the distribution parameters are estimated by the Method of estimating Log Cumulant (MoLC). Classification results on four scenes validates the proposed method with a 99% accuracy.