MCC-LIDAR is currently designed to replace all the classification codes in the input file (for example, if there are errors from a previous classification effort; see ticket #40).
The user intent in this ticket is different: to refine a previous classification effort. For example, run MCC-LIDAR on a file initially with a larger threshold parameter, and then run the program on the file again with a smaller threshold, but only on those points that the first run classified as non-ground (class = 1, unclassified). This new refinement behavior could be expressed with a new option:
--select:class=1
which tells the program to only select those points in the input file with class 1 to feed into the classification algorithm.
The current replacement behavior of MCC-LIDAR would be expressed with this new option as:
--select:class=*
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I'm updating this ticket because if currently fails to document the full description of the problem. Running MCC on just the 1's would result in additional 2's being classified. However, this is only half of the issue. There should be a switch that allows the user to run MCC on just the 2's, in order to apply another set of parameters (and three new scale domains) to the problem, to result in fewer 2's being classified. The old MCC AML allowed you to run MCC multiple times, to achieve a more conservative classification of ground returns.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
MCC-LIDAR is currently designed to replace all the classification codes in the input file (for example, if there are errors from a previous classification effort; see ticket #40).
The user intent in this ticket is different: to refine a previous classification effort. For example, run MCC-LIDAR on a file initially with a larger threshold parameter, and then run the program on the file again with a smaller threshold, but only on those points that the first run classified as non-ground (class = 1, unclassified). This new refinement behavior could be expressed with a new option:
which tells the program to only select those points in the input file with class 1 to feed into the classification algorithm.
The current replacement behavior of MCC-LIDAR would be expressed with this new option as:
I'm updating this ticket because if currently fails to document the full description of the problem. Running MCC on just the 1's would result in additional 2's being classified. However, this is only half of the issue. There should be a switch that allows the user to run MCC on just the 2's, in order to apply another set of parameters (and three new scale domains) to the problem, to result in fewer 2's being classified. The old MCC AML allowed you to run MCC multiple times, to achieve a more conservative classification of ground returns.