Noise reduction algorithm for full-waveform lidar signal

Noise is an inevitable problem which affects coherent imaging systems and consequently reduces the capability of lidar systems to resolve variations in the target heights. Common noise-reduction procedures, including moving Mean and Median filters, operate on a global level and do not perform adaptively on different segments of the backscattered signal. As such, spikes in the trailing edge of the measured signal due to multiple scattering are systematically classified as noise. Such filtering approaches do not account for changes in the characteristics of the waveform. It is assumed that contiguous scatterers are similarly affected by noise, and therefore need to be treated separately. This paper proposes a new adaptive local filter based on an averaging function, called adaptive mean filter (ADMF). Using a 2-D photographic ray-tracing model of a stand-alone tree, a signal with random white Gaussian noise, carrying the responses of the tree structure was simulated. The new filter along with Mean and median filters were examined on ten simulated signals. According to the results, the ADMF filter is abale to reduce noise, as shown by the lower root mean square error (RMSE), higher contrast and signal-to-noise ratio more efficient than the other two filters.