The following is a good representation of corn rows in LiDAR data. Corn rows appear for a number of reasons and can be seen in a single scan or in overlapped areas of two adjoining scans. The corn rows in a single scan can be caused by the back and forth scanning of the system. The manufactures say that any given point needs to be within a certain distance of the actual location on the ground. Well this means that that point measurement can be below the actual elevation on the ground or above the actual elevation of the ground. This makes the distance between the points greater creating the appearance of corn rows. Since the error of a LiDAR sensor is greatest at the edge of scan these corn rows are most apparent at the edges of scans. The presents of corns rows is compounded when you have overlapping flight lines because of the inherent error of the scanner or measurement of any given point, the back and forth scan pattern and difference between the elevation measurements between the two scans. This differences will usually be a result of bad calibration. A lot of processing techniques have been developed to remove corn rows but almost always you can see some existences of corn rows in raw LiDAR ( unclassified strips ). The manufactures of LiDAR systems have improved on this issue and continue to make their systems better. The only time I have seen no existence of corn rows is from a system that generates scans in a single direction but I am confident that at some point the technology will improve to the point that we don’t have to talk about corn rows as a result of system function. If you have any insight on the latest systems or systems that address this issue effectively please let me know or leave a comment.
Thanks… Great information and explanation. Corn rows will continue to be found due to the standards of 80% accuracy where if all the 3rd party used a simple las datasets tool and first (dsm) and last return (dem) the problems would been noted and adjusted. The users of published data are the ones that have to find high end software products that cost over one thousand dollars to correct them. One solution was to go back to the DSM and Las dataset to id the problem tiles and then manually correct the example high vegation to noise. Another option was to create a slop analysis to ID areas of high slop variance which noise was the largest section. A vendor offered the mean x 3 to off set outliers. Perhaps our physics people will kindly refer to their text boookfind related programs or references to fix the gaps between ground vegetation open space and height of the sensor / platform. In short to QC have your distributor provide raw dem / dsm to qc their work before end of contract.