[Met_help] [rt.rap.ucar.edu #77876] History for Merging/Splitting of Simple Objects in MODE

John Halley Gotway via RT met_help at ucar.edu
Tue Sep 13 11:05:18 MDT 2016


----------------------------------------------------------------
  Initial Request
----------------------------------------------------------------

I was wondering if you could give me some more information on how MODE handles the numbering of simple objects when there is  merging or splitting of existing simple objects.


I understand the fuzzy logic engine and how that's applied to the simple and single objects to get cluster and matched objects.  I am more concerned with the merging and splitting of simple objects before they are clustered or merged.


Thanks, Josh


----------------------------------------------------------------
  Complete Ticket History
----------------------------------------------------------------

Subject: Merging/Splitting of Simple Objects in MODE
From: John Halley Gotway
Time: Mon Sep 12 15:57:57 2016

Hello Josh,

I see you have questions about the merging of simple objects in MODE.
To
be clear, MODE does absolutely no "splitting" of simple objects.  It
simply
(1) applies a convolution filter to smooth the field, (2) applies a
threshold, (3) defines simple objects as being contiguous grid squares
which meet the threshold criteria, and (4) applies matching/merging
logic.

If you're wondering about the "numbering" of objects... I believe MODE
just
works through the rows of data from the upper-left grid square to the
lower-right grid square.  And the simple objects are numbered in the
order
that they are encountered.

MODE supports 3 types of merging that are controlled by options set in
the
configuration file:

(1) Merging while matching... MODE compares simple forecast objects to
simple observation objects.  Any pair exceeding the total interest
threshold (default = 0.7) constitute a match.  Any two objects in one
field
matching the SAME object in the other field are merged.  So if
forecast
objects 1 and 3 both match observation object 7, 1 and 3 are merged.

(2) Engine merging... MODE can compare the forecast objects to
themselves
and the observation objects to themselves using the fuzzy logic
approach.
Then apply the logic described in (1).  Obviously each object will
match
itself perfectly but two nearby objects may have a total interest of
0.8.
Objects with interest values exceeding the total interest threshold
are
merged.

(3) Double threshold merging... MODE applies a "higher" threshold to
define
the objects (conv_thresh) and lower threshold to apply this merging
method
(merge_thresh).  If two objects defined by conv_thresh fall within the
same
object defined by merge_thresh, then they are merged together.

Selecting the desired logic is controlled by the merge_flag and
match_flag
in the MODE configuration file.

Does that answer your questions?  If not, just let me know.

Thanks,
John Halley Gotway
met_help at ucar.edu

On Mon, Sep 12, 2016 at 3:00 PM, Markel, Joshua via RT
<met_help at ucar.edu>
wrote:

>
> Mon Sep 12 15:00:12 2016: Request 77876 was acted upon.
> Transaction: Ticket created by joshua.markel at und.edu
>        Queue: met_help
>      Subject: Merging/Splitting of Simple Objects in MODE
>        Owner: Nobody
>   Requestors: joshua.markel at und.edu
>       Status: new
>  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=77876 >
>
>
> I was wondering if you could give me some more information on how
MODE
> handles the numbering of simple objects when there is  merging or
splitting
> of existing simple objects.
>
>
> I understand the fuzzy logic engine and how that's applied to the
simple
> and single objects to get cluster and matched objects.  I am more
concerned
> with the merging and splitting of simple objects before they are
clustered
> or merged.
>
>
> Thanks, Josh
>
>

------------------------------------------------
Subject: Merging/Splitting of Simple Objects in MODE
From: Markel, Joshua
Time: Tue Sep 13 08:46:21 2016

Hi John,


Thanks for your timely feedback, but I should have specified in my
original email that I am working with MODE Time Domain and am
referring to objects and their evolution over time.


For example, if a simple object is one polygon then splits into two
polygons the next forecast time, how does MTD handle the numbering?
Are the objects in the later time step still considered one object or
is an additional object created?


Thanks, Josh

________________________________
From: John Halley Gotway via RT <met_help at ucar.edu>
Sent: Monday, September 12, 2016 4:57:57 PM
To: Markel, Joshua
Subject: Re: [rt.rap.ucar.edu #77876] Merging/Splitting of Simple
Objects in MODE

Hello Josh,

I see you have questions about the merging of simple objects in MODE.
To
be clear, MODE does absolutely no "splitting" of simple objects.  It
simply
(1) applies a convolution filter to smooth the field, (2) applies a
threshold, (3) defines simple objects as being contiguous grid squares
which meet the threshold criteria, and (4) applies matching/merging
logic.

If you're wondering about the "numbering" of objects... I believe MODE
just
works through the rows of data from the upper-left grid square to the
lower-right grid square.  And the simple objects are numbered in the
order
that they are encountered.

MODE supports 3 types of merging that are controlled by options set in
the
configuration file:

(1) Merging while matching... MODE compares simple forecast objects to
simple observation objects.  Any pair exceeding the total interest
threshold (default = 0.7) constitute a match.  Any two objects in one
field
matching the SAME object in the other field are merged.  So if
forecast
objects 1 and 3 both match observation object 7, 1 and 3 are merged.

(2) Engine merging... MODE can compare the forecast objects to
themselves
and the observation objects to themselves using the fuzzy logic
approach.
Then apply the logic described in (1).  Obviously each object will
match
itself perfectly but two nearby objects may have a total interest of
0.8.
Objects with interest values exceeding the total interest threshold
are
merged.

(3) Double threshold merging... MODE applies a "higher" threshold to
define
the objects (conv_thresh) and lower threshold to apply this merging
method
(merge_thresh).  If two objects defined by conv_thresh fall within the
same
object defined by merge_thresh, then they are merged together.

Selecting the desired logic is controlled by the merge_flag and
match_flag
in the MODE configuration file.

Does that answer your questions?  If not, just let me know.

Thanks,
John Halley Gotway
met_help at ucar.edu

On Mon, Sep 12, 2016 at 3:00 PM, Markel, Joshua via RT
<met_help at ucar.edu>
wrote:

>
> Mon Sep 12 15:00:12 2016: Request 77876 was acted upon.
> Transaction: Ticket created by joshua.markel at und.edu
>        Queue: met_help
>      Subject: Merging/Splitting of Simple Objects in MODE
>        Owner: Nobody
>   Requestors: joshua.markel at und.edu
>       Status: new
>  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=77876 >
>
>
> I was wondering if you could give me some more information on how
MODE
> handles the numbering of simple objects when there is  merging or
splitting
> of existing simple objects.
>
>
> I understand the fuzzy logic engine and how that's applied to the
simple
> and single objects to get cluster and matched objects.  I am more
concerned
> with the merging and splitting of simple objects before they are
clustered
> or merged.
>
>
> Thanks, Josh
>
>


------------------------------------------------
Subject: Merging/Splitting of Simple Objects in MODE
From: John Halley Gotway
Time: Tue Sep 13 09:46:37 2016

Josh,

Ah OK, I understand now.  The MTD tool first defines objects across
all
times steps and then it numbers them.  If two objects are separate at
time
T but merge together at time T+1, they will have the same object
number.
Likewise if one object at time T splits into multiple objects at time
T+1,
they will all have the same object number.

I've attached an animated GIF illustrating this.  These are plots of
the
"fcst_object_id" variable in the NetCDF output of MTD, and I'm showing
22
time slices.  While there are several 2D objects at certain times,
there
are only 6 3D space-time objects.

Hope that helps clarify.

Thanks,
John



On Tue, Sep 13, 2016 at 8:46 AM, Markel, Joshua via RT
<met_help at ucar.edu>
wrote:

>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=77876 >
>
> Hi John,
>
>
> Thanks for your timely feedback, but I should have specified in my
> original email that I am working with MODE Time Domain and am
referring to
> objects and their evolution over time.
>
>
> For example, if a simple object is one polygon then splits into two
> polygons the next forecast time, how does MTD handle the numbering?
Are
> the objects in the later time step still considered one object or is
an
> additional object created?
>
>
> Thanks, Josh
>
> ________________________________
> From: John Halley Gotway via RT <met_help at ucar.edu>
> Sent: Monday, September 12, 2016 4:57:57 PM
> To: Markel, Joshua
> Subject: Re: [rt.rap.ucar.edu #77876] Merging/Splitting of Simple
Objects
> in MODE
>
> Hello Josh,
>
> I see you have questions about the merging of simple objects in
MODE.  To
> be clear, MODE does absolutely no "splitting" of simple objects.  It
simply
> (1) applies a convolution filter to smooth the field, (2) applies a
> threshold, (3) defines simple objects as being contiguous grid
squares
> which meet the threshold criteria, and (4) applies matching/merging
logic.
>
> If you're wondering about the "numbering" of objects... I believe
MODE just
> works through the rows of data from the upper-left grid square to
the
> lower-right grid square.  And the simple objects are numbered in the
order
> that they are encountered.
>
> MODE supports 3 types of merging that are controlled by options set
in the
> configuration file:
>
> (1) Merging while matching... MODE compares simple forecast objects
to
> simple observation objects.  Any pair exceeding the total interest
> threshold (default = 0.7) constitute a match.  Any two objects in
one field
> matching the SAME object in the other field are merged.  So if
forecast
> objects 1 and 3 both match observation object 7, 1 and 3 are merged.
>
> (2) Engine merging... MODE can compare the forecast objects to
themselves
> and the observation objects to themselves using the fuzzy logic
approach.
> Then apply the logic described in (1).  Obviously each object will
match
> itself perfectly but two nearby objects may have a total interest of
0.8.
> Objects with interest values exceeding the total interest threshold
are
> merged.
>
> (3) Double threshold merging... MODE applies a "higher" threshold to
define
> the objects (conv_thresh) and lower threshold to apply this merging
method
> (merge_thresh).  If two objects defined by conv_thresh fall within
the same
> object defined by merge_thresh, then they are merged together.
>
> Selecting the desired logic is controlled by the merge_flag and
match_flag
> in the MODE configuration file.
>
> Does that answer your questions?  If not, just let me know.
>
> Thanks,
> John Halley Gotway
> met_help at ucar.edu
>
> On Mon, Sep 12, 2016 at 3:00 PM, Markel, Joshua via RT
<met_help at ucar.edu>
> wrote:
>
> >
> > Mon Sep 12 15:00:12 2016: Request 77876 was acted upon.
> > Transaction: Ticket created by joshua.markel at und.edu
> >        Queue: met_help
> >      Subject: Merging/Splitting of Simple Objects in MODE
> >        Owner: Nobody
> >   Requestors: joshua.markel at und.edu
> >       Status: new
> >  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=77876 >
> >
> >
> > I was wondering if you could give me some more information on how
MODE
> > handles the numbering of simple objects when there is  merging or
> splitting
> > of existing simple objects.
> >
> >
> > I understand the fuzzy logic engine and how that's applied to the
simple
> > and single objects to get cluster and matched objects.  I am more
> concerned
> > with the merging and splitting of simple objects before they are
> clustered
> > or merged.
> >
> >
> > Thanks, Josh
> >
> >
>
>
>

------------------------------------------------
Subject: Merging/Splitting of Simple Objects in MODE
From: Markel, Joshua
Time: Tue Sep 13 10:06:22 2016

Hi John,


Thanks for the clarification, I had analyzed one of my runs and I
thought that's how the numbering worked, I just wanted to make sure.



________________________________
From: John Halley Gotway via RT <met_help at ucar.edu>
Sent: Tuesday, September 13, 2016 10:46:37 AM
To: Markel, Joshua
Subject: Re: [rt.rap.ucar.edu #77876] Merging/Splitting of Simple
Objects in MODE

Josh,

Ah OK, I understand now.  The MTD tool first defines objects across
all
times steps and then it numbers them.  If two objects are separate at
time
T but merge together at time T+1, they will have the same object
number.
Likewise if one object at time T splits into multiple objects at time
T+1,
they will all have the same object number.

I've attached an animated GIF illustrating this.  These are plots of
the
"fcst_object_id" variable in the NetCDF output of MTD, and I'm showing
22
time slices.  While there are several 2D objects at certain times,
there
are only 6 3D space-time objects.

Hope that helps clarify.

Thanks,
John



On Tue, Sep 13, 2016 at 8:46 AM, Markel, Joshua via RT
<met_help at ucar.edu>
wrote:

>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=77876 >
>
> Hi John,
>
>
> Thanks for your timely feedback, but I should have specified in my
> original email that I am working with MODE Time Domain and am
referring to
> objects and their evolution over time.
>
>
> For example, if a simple object is one polygon then splits into two
> polygons the next forecast time, how does MTD handle the numbering?
Are
> the objects in the later time step still considered one object or is
an
> additional object created?
>
>
> Thanks, Josh
>
> ________________________________
> From: John Halley Gotway via RT <met_help at ucar.edu>
> Sent: Monday, September 12, 2016 4:57:57 PM
> To: Markel, Joshua
> Subject: Re: [rt.rap.ucar.edu #77876] Merging/Splitting of Simple
Objects
> in MODE
>
> Hello Josh,
>
> I see you have questions about the merging of simple objects in
MODE.  To
> be clear, MODE does absolutely no "splitting" of simple objects.  It
simply
> (1) applies a convolution filter to smooth the field, (2) applies a
> threshold, (3) defines simple objects as being contiguous grid
squares
> which meet the threshold criteria, and (4) applies matching/merging
logic.
>
> If you're wondering about the "numbering" of objects... I believe
MODE just
> works through the rows of data from the upper-left grid square to
the
> lower-right grid square.  And the simple objects are numbered in the
order
> that they are encountered.
>
> MODE supports 3 types of merging that are controlled by options set
in the
> configuration file:
>
> (1) Merging while matching... MODE compares simple forecast objects
to
> simple observation objects.  Any pair exceeding the total interest
> threshold (default = 0.7) constitute a match.  Any two objects in
one field
> matching the SAME object in the other field are merged.  So if
forecast
> objects 1 and 3 both match observation object 7, 1 and 3 are merged.
>
> (2) Engine merging... MODE can compare the forecast objects to
themselves
> and the observation objects to themselves using the fuzzy logic
approach.
> Then apply the logic described in (1).  Obviously each object will
match
> itself perfectly but two nearby objects may have a total interest of
0.8.
> Objects with interest values exceeding the total interest threshold
are
> merged.
>
> (3) Double threshold merging... MODE applies a "higher" threshold to
define
> the objects (conv_thresh) and lower threshold to apply this merging
method
> (merge_thresh).  If two objects defined by conv_thresh fall within
the same
> object defined by merge_thresh, then they are merged together.
>
> Selecting the desired logic is controlled by the merge_flag and
match_flag
> in the MODE configuration file.
>
> Does that answer your questions?  If not, just let me know.
>
> Thanks,
> John Halley Gotway
> met_help at ucar.edu
>
> On Mon, Sep 12, 2016 at 3:00 PM, Markel, Joshua via RT
<met_help at ucar.edu>
> wrote:
>
> >
> > Mon Sep 12 15:00:12 2016: Request 77876 was acted upon.
> > Transaction: Ticket created by joshua.markel at und.edu
> >        Queue: met_help
> >      Subject: Merging/Splitting of Simple Objects in MODE
> >        Owner: Nobody
> >   Requestors: joshua.markel at und.edu
> >       Status: new
> >  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=77876 >
> >
> >
> > I was wondering if you could give me some more information on how
MODE
> > handles the numbering of simple objects when there is  merging or
> splitting
> > of existing simple objects.
> >
> >
> > I understand the fuzzy logic engine and how that's applied to the
simple
> > and single objects to get cluster and matched objects.  I am more
> concerned
> > with the merging and splitting of simple objects before they are
> clustered
> > or merged.
> >
> >
> > Thanks, Josh
> >
> >
>
>
>


------------------------------------------------


More information about the Met_help mailing list