Airborne LiDAR intensity banding: Cause and solution

One of the issues limiting the use of airborne LiDAR intensity data for thematic analysis is the appearance of intensity noise. Although recent development of radiometric correction and normalization techniques can help to reduce the intensity discrepancy occurred within an individual data strip and between overlapping data strips, systematic striping noise found within an individual strip still causes an analytical burden for any thematic application. Nevertheless, there is a lack of studies discussing such intensity banding issue, which is mainly caused by the backscattered laser pulses of a specific scanning direction partially falling outside the receiver's field of view (FOV). In this work, we propose a LiDAR scan line correction (LSLC) algorithm to remove these stripe artifacts caused by such intensity banding effect. The proposed LSLC first splits the LiDAR data into two subsets of each representing an individual scanning direction. A polynomial model is built by first pairing up the closest points between the two subsets, where both the intensity and scan angle serve as the model parameters. With the fitted polynomial function, the LiDAR subset with relatively lower intensity value can be normalized to the other subset. We utilized three LiDAR datasets acquired online for experimental testing, and all of them demonstrated a successful removal of the striping noise, resulting in a reduction of coefficient of variation by up to 13%, 28% and 80% in the scenario with light, mild and extreme level of noise, respectively. Regardless of the settings being used, LSLC still outperformed the histogram equalization in terms of noise reduction and removal in all the experimental trials.