Tutorial on UTM coordinates?
in [DSP]
|
Prev: Assembly optimized error
Next: RLC circuit tutorial
From: Rune Allnor on 27 Nov 2009 06:58 Hi all. I am working with some data that contain UTM positions. The problem is that I need the UTM coordinates on centimeter precision, which means I need some 8 or 9 significant digits in the UTM coordinates; a few digits too many to make the coordinates fit into a single-precision floating point number. On the other hand, the position data I work with vary only on the order of kilometers, so the single-precision floating point format is more than precise enough to do the stuff I want. And yes. Savings in both speed and memory footprint are significant motivations for making the effort to make due with single precision. Along with the data, there comes a UTM Zone designation. The question is if I can use this UTN Zone designation to somehow derive a reference UTM position, and use relative offsets from this reference position instead of the full positions? Rune
From: Tim Wescott on 27 Nov 2009 11:50 On Fri, 27 Nov 2009 03:58:12 -0800, Rune Allnor wrote: > Hi all. > > I am working with some data that contain UTM positions. The problem is > that I need the UTM coordinates on centimeter precision, which means I > need some 8 or 9 significant digits in the UTM coordinates; a few digits > too many to make the coordinates fit into a single-precision floating > point number. > > On the other hand, the position data I work with vary only on the order > of kilometers, so the single-precision floating point format is more > than precise enough to do the stuff I want. And yes. Savings in both > speed and memory footprint are significant motivations for making the > effort to make due with single precision. > > Along with the data, there comes a UTM Zone designation. > > The question is if I can use this UTN Zone designation to somehow derive > a reference UTM position, and use relative offsets from this reference > position instead of the full positions? > > Rune Hey Rune: You need to drink your coffee -- I think this is easier than you think it is, but I can't say because it's not entirely clear what you're doing. You mention data in centimeter precision, then you mention data in kilometer precision. Are you trying to say that you're being presented data in cm, but you only need to keep data in km? If so, what about simple truncation? I just skimmed the Wikipedia entry on UTM, and it looks (in principal at least) to be dead easy as long as you know what the reference meridians are for each longitudnal zone. Are there subtleties that I'm not seeing (which there shouldn't be for a system designed to make the math easy)? Or do you need to drink another cuppa? -- www.wescottdesign.com
From: Rune Allnor on 27 Nov 2009 12:08 On 27 Nov, 17:50, Tim Wescott <t...(a)seemywebsite.com> wrote: > On Fri, 27 Nov 2009 03:58:12 -0800, Rune Allnor wrote: > > Hi all. > > > I am working with some data that contain UTM positions. The problem is > > that I need the UTM coordinates on centimeter precision, which means I > > need some 8 or 9 significant digits in the UTM coordinates; a few digits > > too many to make the coordinates fit into a single-precision floating > > point number. > > > On the other hand, the position data I work with vary only on the order > > of kilometers, so the single-precision floating point format is more > > than precise enough to do the stuff I want. And yes. Savings in both > > speed and memory footprint are significant motivations for making the > > effort to make due with single precision. > > > Along with the data, there comes a UTM Zone designation. > > > The question is if I can use this UTN Zone designation to somehow derive > > a reference UTM position, and use relative offsets from this reference > > position instead of the full positions? > > > Rune > > Hey Rune: > > You need to drink your coffee -- I think this is easier than you think it > is, but I can't say because it's not entirely clear what you're doing. > > You mention data in centimeter precision, then you mention data in > kilometer precision. Are you trying to say that you're being presented > data in cm, but you only need to keep data in km? If so, what about > simple truncation? I'm working in cm precision in areas of size km x km. > Or do you need to drink another cuppa? Probably. Another couple of cans... of the espresso strength stuff. Rune
From: Tim Wescott on 27 Nov 2009 12:28 On Fri, 27 Nov 2009 09:08:23 -0800, Rune Allnor wrote: > On 27 Nov, 17:50, Tim Wescott <t...(a)seemywebsite.com> wrote: >> On Fri, 27 Nov 2009 03:58:12 -0800, Rune Allnor wrote: >> > Hi all. >> >> > I am working with some data that contain UTM positions. The problem >> > is that I need the UTM coordinates on centimeter precision, which >> > means I need some 8 or 9 significant digits in the UTM coordinates; a >> > few digits too many to make the coordinates fit into a >> > single-precision floating point number. >> >> > On the other hand, the position data I work with vary only on the >> > order of kilometers, so the single-precision floating point format is >> > more than precise enough to do the stuff I want. And yes. Savings in >> > both speed and memory footprint are significant motivations for >> > making the effort to make due with single precision. >> >> > Along with the data, there comes a UTM Zone designation. >> >> > The question is if I can use this UTN Zone designation to somehow >> > derive a reference UTM position, and use relative offsets from this >> > reference position instead of the full positions? >> >> > Rune >> >> Hey Rune: >> >> You need to drink your coffee -- I think this is easier than you think >> it is, but I can't say because it's not entirely clear what you're >> doing. >> >> You mention data in centimeter precision, then you mention data in >> kilometer precision. Are you trying to say that you're being presented >> data in cm, but you only need to keep data in km? If so, what about >> simple truncation? > > I'm working in cm precision in areas of size km x km. It makes sense now. I could see that if you have a _bunch_ of data all within one area then you could define a one-km-square zone and do everything relative to the center of that then you'd be doing pretty well. I could even see just extending the UTM concept to zones of any size. Of course, then you have to arrange your data so that should your zone change you can represent it unambiguously -- but that's more of a programming problem than a signal processing one. -- www.wescottdesign.com
From: Rich Webb on 27 Nov 2009 13:10
On Fri, 27 Nov 2009 03:58:12 -0800 (PST), Rune Allnor <allnor(a)tele.ntnu.no> wrote: >Hi all. > >I am working with some data that contain UTM positions. >The problem is that I need the UTM coordinates on centimeter >precision, which means I need some 8 or 9 significant digits >in the UTM coordinates; a few digits too many to make the >coordinates fit into a single-precision floating point number. > >On the other hand, the position data I work with vary only on >the order of kilometers, so the single-precision floating point >format is more than precise enough to do the stuff I want. >And yes. Savings in both speed and memory footprint are >significant motivations for making the effort to make due >with single precision. > >Along with the data, there comes a UTM Zone designation. > >The question is if I can use this UTN Zone designation to >somehow derive a reference UTM position, and use relative >offsets from this reference position instead of the full positions? If you haven't already run across it, one very hand app for working with UTM and similar coordinate systems is the GEOTRANS tool from the NGA. Comes with source and a BSD-style (use it but give credit) license. http://earth-info.nga.mil/GandG/geotrans/ I haven't had to muck-about with anything but good old lat/lon recently but when I was, this toolset was great. It probably doesn't specifically address your question but could be a good independent check of your results. -- Rich Webb Norfolk, VA |