What's the rational behind laser_filters/ScanShadowsFilter?

edit

Hi bros,

I am facing with 'ghost obstacles' that similar to above problem, however, I could not solve the problem even I used ScanShadowsFilter, can you please give me some advices, thanks in advances, this is my params:

scan_filter_chain:

name: box_filter type: laser_filters/LaserScanBoxFilter

params:

box_frame: R_007/base_link

max_x: 0.438

max_y: 0.332

max_z: 0.231

min_x: -0.438

min_y: -0.332

min_z: -0.01

name: shadows

type: ScanShadowsFilter

params:

min_angle: 5

max_angle: 175

neighbors: 5

window: 1

type: LaserArrayFilter

name: laser_median_filter

params:

range_filter_chain:

name: median_3

type: filters/MultiChannelMedianFilterFloat

params:

number_of_observations: 3

unused: 6

intensity_filter_chain:

name: median_3

type: filters/MultiChannelMedianFilterFloat

params:

number_of_observations: 3

unused: 6

Besides, I also used LaserScanIntensityFilter:

scan_filter_chain:

name: intensity

type: LaserScanIntensityFilter

params:

lower_threshold: 34000

upper_threshold: 37000

disp_histogram: 0

As pitpet noticed, ScanShadowsFilter works by computing the angle differences between successive readings and discarding those that don't fit within a user-specified range. I suspect the incorrectly erased readings fell below the minimum angle difference min_angle – you can try reducing it to 0 and checking what happens.

Maybe you want to use LaserScanAngularBoundsFilter instead of ScanShadowFilter. ScanShadowFilter removes laser data points which satisfies the following conditions : for all p1, if exists point p2 in (p1 -window, p1 + window), so that (∠O P1 P2) < minAngular, or (∠O P1 P2) > maxAngular, then, all datapoint in (p1 - neigbor, p1 + neigbor) will be removed. Where O means the origin of the laser.