Loading ...
Sorry, an error occurred while loading the content.
 

Novel Machine Vision Solution For Determining Kite State

Expand Messages
  • dave santos
    Minute changes in kite line angle and line curvature at the surface deterministically reflect the position, orientation, velocity, and pull of a kite
    Message 1 of 6 , Dec 18, 2012
      Minute changes in kite line angle and line curvature at the surface deterministically reflect the position, orientation, velocity, and pull of a kite aloft. Crude mechanical line position encoders are common on AWES developmental platforms, but prone to damage and uncertainty (especially during transient line slackness). Elaborate multisensing at the kite is commonly specified to avoid the uncertainty mechanical encoders suffer from. A separate line-tension sensor is usual. A better line encoding method is needed to help resolve kite state.

      To adequately monitor kite state for control purposes, it may suffice for kiteline geometry at the anchor point to be closely imaged within a dark fabric "boot" by stereo microvideo (or en-plein-air by micro laser scanner). The kitelines in the boot would be artificially lit for an easily processed image. A stereo pair requirement could be net with just one camera and mirror(s) in the view field. 

      The system could be sensitive to events mechanical encoders miss. Even twisted line states can in principle be disambiguated by this sort of close videogrammetry. A look-up table (database) of local line states would output the detected kite state. Markov chains, Baysean inference, or model predictive control could interpret and act on this state accordingly.

      To complete a minimalist sensor suite, an encoded reel can keep track of line length. Acoustic monitoring of line noise could be a semi-redundant data source (topic for a separate post). Avionics on the kite remains an option, working even better without mechanical line encoder uncertainty to worry about. Meteorological sensors and METAR data would be a given, helping factor out hodographic twist, for example. A simple limit switch/sensor could react to, or confirm, a kite returned to a cradle.

      The proposed machine vision kite state detection method would be immune to many normal failings of outdoor vision systems, such as sun glare, night lighting, fog, and bird droppings or raindrops on a lens. It would eliminate the inherent flaws of mechanical line position encoders. It could be an ultimately cheap solution as well.

      CC BY-NC-SA


    • roderickjosephread
      Cover the lines, kites, joins, blocks, bits and pieces in RFID tags http://www.foodproductiondaily.com/Supply-Chain/Radar-reader-spots-RFID-
      Message 2 of 6 , Dec 19, 2012
        Cover the lines, kites, joins, blocks, bits and pieces in RFID tags
        http://www.foodproductiondaily.com/Supply-Chain/Radar-reader-spots-RFID-tag-location-at-a-distance 
        Map them with radar.

        or

        like the boot idea but further range
        encode tethers and control lines with individual reflection patterns (a line identifier and length number)
        what can be seen by ladar can be read and a best guess system state generated...
        This way overall curve can be determined by the relative positions of two or more points on the line.



        --- In AirborneWindEnergy@yahoogroups.com, dave santos wrote:
        >
        > Minute changes in kite line angle and line curvature at the surface deterministically reflect the position, orientation, velocity, and pull of a kite aloft.�Crude mechanical line position encoders are common on AWES developmental platforms, but prone to damage and uncertainty (especially during transient line slackness). Elaborate multisensing at the kite is commonly specified to avoid the uncertainty mechanical encoders suffer from.�A separate line-tension sensor is usual.�A better line encoding method is needed to help resolve kite state.
        >
        > To adequately monitor kite state for control purposes, it may suffice for kiteline geometry at the anchor point to be closely imaged within a dark fabric "boot" by stereo microvideo (or en-plein-air by micro laser scanner).�The kitelines in the boot would be artificially lit for an easily processed image.�A�stereo pair requirement could be net with just one camera and mirror(s) in the view field.�
        >
        > The system could be sensitive to events mechanical encoders miss.�Even twisted line states can in principle be disambiguated by this sort of close videogrammetry.�A look-up table (database) of local line states would output the detected kite state. Markov chains, Baysean inference, or model predictive control could interpret and act on this state accordingly.
        >
        > To complete a minimalist sensor suite, an encoded reel can keep track of line length.�Acoustic monitoring of line noise could be a semi-redundant data source (topic for a separate post). Avionics on the kite remains an option, working even better without mechanical line encoder uncertainty to worry about. Meteorological sensors and METAR data would be a given, helping factor out hodographic twist, for example. A simple limit switch/sensor could react to, or confirm, a kite returned to a cradle.
        >
        > The proposed machine vision kite state detection method would be immune to many normal failings of outdoor vision systems, such as sun glare, night lighting, fog, and bird droppings or raindrops on a lens. It would eliminate the inherent flaws of mechanical line position encoders. It could be an ultimately cheap solution as well.
        >
        > CC BY-NC-SA
        >
      • christopher carlin
        Good idea. Make an unobservable system observable or at least improve it. Simplifies control problem immensely. It would be interesting to see what is being
        Message 3 of 6 , Dec 19, 2012
          Good idea. Make an unobservable system observable or at least improve it. Simplifies control problem immensely. It would be interesting to see what is being done these days in the space community on large solar arrays in orbit. When I was involved many years ago our major concern was how to steer large very flexible arrays in orbit with out inducing instabilities. Having a large number of distributed feedbacks would help in both cases.


          Regards,,

          Chris
          On Dec 19, 2012, at 11:11 AM, roderickjosephread wrote:

           

          Cover the lines, kites, joins, blocks, bits and pieces in RFID tags
          http://www.foodproductiondaily.com/Supply-Chain/Radar-reader-spots-RFID-tag-location-at-a-distance 
          Map them with radar.

          or

          like the boot idea but further range
          encode tethers and control lines with individual reflection patterns (a line identifier and length number)
          what can be seen by ladar can be read and a best guess system state generated...
          This way overall curve can be determined by the relative positions of two or more points on the line.



          --- In AirborneWindEnergy@yahoogroups.com, dave santos wrote:
          >
          > Minute changes in kite line angle and line curvature at the surface deterministically reflect the position, orientation, velocity, and pull of a kite aloft.�Crude mechanical line position encoders are common on AWES developmental platforms, but prone to damage and uncertainty (especially during transient line slackness). Elaborate multisensing at the kite is commonly specified to avoid the uncertainty mechanical encoders suffer from.�A separate line-tension sensor is usual.�A better line encoding method is needed to help resolve kite state.
          >
          > To adequately monitor kite state for control purposes, it may suffice for kiteline geometry at the anchor point to be closely imaged within a dark fabric "boot" by stereo microvideo (or en-plein-air by micro laser scanner).�The kitelines in the boot would be artificially lit for an easily processed image.�A�stereo pair requirement could be net with just one camera and mirror(s) in the view field.�
          >
          > The system could be sensitive to events mechanical encoders miss.�Even twisted line states can in principle be disambiguated by this sort of close videogrammetry.�A look-up table (database) of local line states would output the detected kite state. Markov chains, Baysean inference, or model predictive control could interpret and act on this state accordingly.
          >
          > To complete a minimalist sensor suite, an encoded reel can keep track of line length.�Acoustic monitoring of line noise could be a semi-redundant data source (topic for a separate post). Avionics on the kite remains an option, working even better without mechanical line encoder uncertainty to worry about. Meteorological sensors and METAR data would be a given, helping factor out hodographic twist, for example. A simple limit switch/sensor could react to, or confirm, a kite returned to a cradle.
          >
          > The proposed machine vision kite state detection method would be immune to many normal failings of outdoor vision systems, such as sun glare, night lighting, fog, and bird droppings or raindrops on a lens. It would eliminate the inherent flaws of mechanical line position encoders. It could be an ultimately cheap solution as well.
          >
          > CC BY-NC-SA
          >


        • dbmurr@ymail.com
          During the late 90 s my practice undertook the task of doing building surveys on many of the government owned historical (100 - 300 yrs. old) buildings in
          Message 4 of 6 , Dec 19, 2012
            During the late '90's my practice undertook the task of doing building surveys on many of the government owned historical (100 - 300 yrs. old) buildings in Bermuda for the purpose of life safety upgrades. We used the latest Lieca mapping equipment (that we could afford) back then. Around four or five years ago a mayor of a small historic town asked me to price the latest 3D capture system as a favour so he could see if the county's GIS department would be interested in a shared system for the purpose of adding building fabric detail to the existing digital model. Our best choice then was a system by Faro (http://www.faro.com/product2. aspx?ct=us&content=pro). The trouble was the price tag was around $130k for what we thought we needed then...

            My wife's new http://www.fitbit.com/  showed up in the mail yesterday & won't be available for kite experimentation use until the 26th, but after reading the last few posts I thought of the following idea.

            If a hub was designed (in the spirit of Pierre's hard plastic component on http://flygenkite.com/) that would hold the fewest number of "last year's" smart phones in a fashion that would have overlapping fields of view in all directions, this could be held in a small transparent, beach ball like, housing. If this was held at the center of mass in the Mothra catenary structure it would have a better mapping vantage point than down below on the ground (needs study). A second layer of information could be got from "Fitbit" like devices, painted for best visibility, & clipped to as many primary nodes & sail corners as seems reasonable for thorough real time mapping. I would not be surprised if computational lag time, for these types of equipment, will inhibit good control of the array, but "next year's" smart phones may work well enough. Possibly this sort of system can be made to work for just a few hundred dollars added to the expense. 

            DaveB


            --- In AirborneWindEnergy@yahoogroups.com, christopher carlin wrote:
            >
            > Good idea. Make an unobservable system observable or at least improve
            > it. Simplifies control problem immensely. It would be interesting to
            > see what is being done these days in the space community on large
            > solar arrays in orbit. When I was involved many years ago our major
            > concern was how to steer large very flexible arrays in orbit with out
            > inducing instabilities. Having a large number of distributed feedbacks
            > would help in both cases.
            >
            >
            > Regards,,
            >
            > Chris
            > On Dec 19, 2012, at 11:11 AM, roderickjosephread wrote:
            >
            > > Cover the lines, kites, joins, blocks, bits and pieces in RFID tags
            > > http://www.foodproductiondaily.com/Supply-Chain/Radar-reader-spots-RFID-tag-location-at-a-distance
            > > Map them with radar.
            > >
            > > or
            > >
            > > like the boot idea but further range
            > > encode tethers and control lines with individual reflection patterns
            > > (a line identifier and length number)
            > > what can be seen by ladar can be read and a best guess system state
            > > generated...
            > > This way overall curve can be determined by the relative positions
            > > of two or more points on the line.
            > >
            > >
            > >
            > > --- In AirborneWindEnergy@yahoogroups.com, dave santos wrote:
            > > >
            > > > Minute changes in kite line angle and line curvature at the
            > > surface deterministically reflect the position, orientation,
            > > velocity, and pull of a kite aloft.�Crude mechanical line position
            > > encoders are common on AWES developmental platforms, but prone to
            > > damage and uncertainty (especially during transient line slackness).
            > > Elaborate multisensing at the kite is commonly specified to avoid
            > > the uncertainty mechanical encoders suffer from.�A separate line-
            > > tension sensor is usual.�A better line encoding method is needed
            > > to help resolve kite state.
            > > >
            > > > To adequately monitor kite state for control purposes, it may
            > > suffice for kiteline geometry at the anchor point to be closely
            > > imaged within a dark fabric "boot" by stereo microvideo (or en-plein-
            > > air by micro laser scanner).�The kitelines in the boot would be
            > > artificially lit for an easily processed image.�A�stereo pair
            > > requirement could be net with just one camera and mirror(s) in the
            > > view field.�
            > > >
            > > > The system could be sensitive to events mechanical encoders
            > > miss.�Even twisted line states can in principle be disambiguated
            > > by this sort of close videogrammetry.�A look-up table (database)
            > > of local line states would output the detected kite state. Markov
            > > chains, Baysean inference, or model predictive control could
            > > interpret and act on this state accordingly.
            > > >
            > > > To complete a minimalist sensor suite, an encoded reel can keep
            > > track of line length.�Acoustic monitoring of line noise could be a
            > > semi-redundant data source (topic for a separate post). Avionics on
            > > the kite remains an option, working even better without mechanical
            > > line encoder uncertainty to worry about. Meteorological sensors and
            > > METAR data would be a given, helping factor out hodographic twist,
            > > for example. A simple limit switch/sensor could react to, or
            > > confirm, a kite returned to a cradle.
            > > >
            > > > The proposed machine vision kite state detection method would be
            > > immune to many normal failings of outdoor vision systems, such as
            > > sun glare, night lighting, fog, and bird droppings or raindrops on a
            > > lens. It would eliminate the inherent flaws of mechanical line
            > > position encoders. It could be an ultimately cheap solution as well.
            > > >
            > > > CC BY-NC-SA
            > > >
            > >
            > >
            >
          • roderickjosephread
            Yeah, we could get some folk who know more about it to look at it... http://www.saiip-vision.org/tyang/papers/TaoYang_CVPR05.pdf
            Message 5 of 6 , Dec 19, 2012
              Yeah, we could get some folk who know more about it to look at it...

              http://www.saiip-vision.org/tyang/papers/TaoYang_CVPR05.pdf

              http://cvrc.ece.utexas.edu/Publications/Object_tracking_in_an_outdoor_enviroment.pdf

              http://robotics.caltech.edu/~jerma/research_papers/fast_robust_tracking.pdf

              RFID is maybe still slow looking from these studies ... or prone to signal deterioration.
              Anyone here handy with Ladar vision, as used by autonomous vehicles?


              --- In AirborneWindEnergy@yahoogroups.com, "dbmurr@..." wrote:
              >
              > During the late '90's my practice undertook the task of doing building
              > surveys on many of the government owned historical (100 - 300 yrs. old)
              > buildings in Bermuda for the purpose of life safety upgrades. We used
              > the latest Lieca mapping equipment (that we could afford) back then.
              > Around four or five years ago a mayor of a small historic town asked me
              > to price the latest 3D capture system as a favour so he could see if the
              > county's GIS department would be interested in a shared system for the
              > purpose of adding building fabric detail to the existing digital model.
              > Our best choice then was a system by Faro
              > (http://www.faro.com/product2.aspx?ct=us&content=pro
              > ). The trouble was
              > the price tag was around $130k for what we thought we needed then...
              > My wife's new http://www.fitbit.com/ showed up
              > in the mail yesterday & won't be available for kite experimentation use
              > until the 26th, but after reading the last few posts I thought of the
              > following idea.
              > If a hub was designed (in the spirit of Pierre's hard plastic component
              > on http://flygenkite.com/ ) that would hold the
              > fewest number of "last year's" smart phones in a fashion that would have
              > overlapping fields of view in all directions, this could be held in a
              > small transparent, beach ball like, housing. If this was held at the
              > center of mass in the Mothra catenary structure it would have a better
              > mapping vantage point than down below on the ground (needs study). A
              > second layer of information could be got from "Fitbit" like devices,
              > painted for best visibility, & clipped to as many primary nodes & sail
              > corners as seems reasonable for thorough real time mapping. I would not
              > be surprised if computational lag time, for these types of equipment,
              > will inhibit good control of the array, but "next year's" smart phones
              > may work well enough. Possibly this sort of system can be made to work
              > for just a few hundred dollars added to the expense.
              > DaveBhttp://flyinground.com/
              > http://agronautics.com/
              >
              > --- In AirborneWindEnergy@yahoogroups.com, christopher carlin wrote:
              > >
              > > Good idea. Make an unobservable system observable or at least improve
              > > it. Simplifies control problem immensely. It would be interesting to
              > > see what is being done these days in the space community on large
              > > solar arrays in orbit. When I was involved many years ago our major
              > > concern was how to steer large very flexible arrays in orbit with out
              > > inducing instabilities. Having a large number of distributed feedbacks
              > > would help in both cases.
              > >
              > >
              > > Regards,,
              > >
              > > Chris
              > > On Dec 19, 2012, at 11:11 AM, roderickjosephread wrote:
              > >
              > > > Cover the lines, kites, joins, blocks, bits and pieces in RFID tags
              > > >
              > http://www.foodproductiondaily.com/Supply-Chain/Radar-reader-spots-RFID-\
              > tag-location-at-a-distance
              > > > Map them with radar.
              > > >
              > > > or
              > > >
              > > > like the boot idea but further range
              > > > encode tethers and control lines with individual reflection patterns
              > > > (a line identifier and length number)
              > > > what can be seen by ladar can be read and a best guess system state
              > > > generated...
              > > > This way overall curve can be determined by the relative positions
              > > > of two or more points on the line.
              > > >
              > > >
              > > >
              > > > --- In AirborneWindEnergy@yahoogroups.com, dave santos wrote:
              > > > >
              > > > > Minute changes in kite line angle and line curvature at the
              > > > surface deterministically reflect the position, orientation,
              > > > velocity, and pull of a kite aloft.�Crude mechanical line
              > position
              > > > encoders are common on AWES developmental platforms, but prone to
              > > > damage and uncertainty (especially during transient line slackness).
              > > > Elaborate multisensing at the kite is commonly specified to avoid
              > > > the uncertainty mechanical encoders suffer from.�A separate
              > line-
              > > > tension sensor is usual.�A better line encoding method is
              > needed
              > > > to help resolve kite state.
              > > > >
              > > > > To adequately monitor kite state for control purposes, it may
              > > > suffice for kiteline geometry at the anchor point to be closely
              > > > imaged within a dark fabric "boot" by stereo microvideo (or
              > en-plein-
              > > > air by micro laser scanner).�The kitelines in the boot would be
              > > > artificially lit for an easily processed image.�A�stereo
              > pair
              > > > requirement could be net with just one camera and mirror(s) in the
              > > > view field.�
              > > > >
              > > > > The system could be sensitive to events mechanical encoders
              > > > miss.�Even twisted line states can in principle be
              > disambiguated
              > > > by this sort of close videogrammetry.�A look-up table
              > (database)
              > > > of local line states would output the detected kite state. Markov
              > > > chains, Baysean inference, or model predictive control could
              > > > interpret and act on this state accordingly.
              > > > >
              > > > > To complete a minimalist sensor suite, an encoded reel can keep
              > > > track of line length.�Acoustic monitoring of line noise could
              > be a
              > > > semi-redundant data source (topic for a separate post). Avionics on
              > > > the kite remains an option, working even better without mechanical
              > > > line encoder uncertainty to worry about. Meteorological sensors and
              > > > METAR data would be a given, helping factor out hodographic twist,
              > > > for example. A simple limit switch/sensor could react to, or
              > > > confirm, a kite returned to a cradle.
              > > > >
              > > > > The proposed machine vision kite state detection method would be
              > > > immune to many normal failings of outdoor vision systems, such as
              > > > sun glare, night lighting, fog, and bird droppings or raindrops on a
              > > > lens. It would eliminate the inherent flaws of mechanical line
              > > > position encoders. It could be an ultimately cheap solution as well.
              > > > >
              > > > > CC BY-NC-SA
              > > > >
              > > >
              > > >
              > >
              >
            • dave santos
              Having followed RFID over decades, its only barely practical for our needs. One problem is embedding or attaching little radios to a pure kiteline and then
              Message 6 of 6 , Dec 19, 2012
                Having followed RFID over decades, its only barely practical for our needs. One problem is embedding or attaching little radios to a pure kiteline and then running them at high duty thru fairleads, around pulleys, and wound onto reels. You can put them on the kite, but range or power are problems. Its rather wasteful to be constantly pinging a large airspace with enough EM RF energy for a passive transponder to then re-radiate in all directions so the base gets a signal. The base antenna would have to be a phased array to pinpoint the RFID location. Active RFID devices need a dedicated power supply. Wait a few more years to avoid the bleeding edge of marginal feasibility.

                The machine vision line encoding method described here is not perfect, its just what seems practical at this time, given cheap mature videogrammetry. One could market it as a terrahertz frequency device doing "TFID" of kiteline state. How cool is that?  :)

              Your message has been successfully submitted and would be delivered to recipients shortly.