The satellite data provided need to have different frequency of the EM wave associated with that band. It would be really helpful to understand and comprehend the data if the frequency of each band is provided.
", "messages": [ { "id": "151620", "postDate": "12/21/2016 15:05:13", "content": "The satellite data provided need to have different frequency of the EM wave associated with that band. It would be really helpful to understand and comprehend the data if the frequency of each band is provided.
", "rawMarkdown": "The satellite data provided need to have different frequency of the EM wave associated with that band. It would be really helpful to understand and comprehend the data if the frequency of each band is provided.", "votes": null }, { "id": "151628", "postDate": "12/21/2016 15:33:45", "content": "Agreed. We know it is the WorldView 3 satellite, and can lookup the specs for that. But I can't figure out which bands in the images correspond to the satellite wavelengths.
", "rawMarkdown": "Agreed. We know it is the WorldView 3 satellite, and can lookup the specs for that. But I can't figure out which bands in the images correspond to the satellite wavelengths.", "votes": null }, { "id": "151665", "postDate": "12/21/2016 18:38:31", "content": "http://satimagingcorp.s3.amazonaws.com/site/pdf/WorldView3-DS-WV3-Web.pdf\nhas the band-by-band frequencies.
\n\nThe header of the GeoTIFFS contain the band ordering.
\n\ngdalinfo sixteen_band/6150_3_2_P.tif
\nFiles: sixteen_band/6150_3_2_P.tif
\nSize is 3403, 3348 --- bandList = [ 1; ]
Files: sixteen_band/6150_3_2_M.tif
\nSize is 851, 837 --- bandList = [ 6; 2; 3; 7; 4; 8; 5; 9;]
Files: sixteen_band/6150_3_2_A.tif
\nSize is 137, 134 --- bandList = [ 10; 11; 12; 13; 14; 15; 16; 17;]
So are they like this then? I'm not sure if this is correct. If I load an M image into QGIS and set RGB to bands 6,4,3 as below, it does not look like a true color image.
\n\nP Images\n1 : Panchromatic 400-800 nm
\n\nM images\n2 : Coastal 400-450
\n3 : Blue 450-510
\n4 : Green 510-580
\n5 : Yellow 585-625
\n6 : Red 630-690
\n7 : Red edge 705-745
\n8 : Near-IR1 770-895
\n9 : Near-IR2 860-1040
@wendy kan Can you please look into it. \nI am attaching the figure of 6010_0_0_M with merging the bands as described by timbabweSz. \nI don't think this is the correct labels for bands.
", "rawMarkdown": "wendy kan Can you please look into it. \r\nI am attaching the figure of 6010_0_0_M with merging the bands as described by timbabweSz. \r\nI don't think this is the correct labels for bands.", "votes": null }, { "id": "151933", "postDate": "12/22/2016 15:57:47", "content": "I'm looking at 6100_2_2_M.tif. The GeoTiff header gives the scrambled order, but comparing the band means for the 3-band RGB with 6100_2_2_M.tif, the fifth/third/second M bands match up, more or less.
\n\nThe band order given in the header looks incorrect.
\n\nFilename: /net/tmp/dstl/three_band/6100_2_2.tif
\n\nDims: Full Scene (11,356,459 points)
\n\nBasic Stats Min Max Mean Stdev
\n\n Band 1 76 1717 198.610207 77.636097\n\n Band 2 215 1949 328.566782 61.056596\n\n Band 3 154 1664 229.807967 34.622226\nFilename: /net/tmp/dstl/sixteen_band/6100_2_2_M.tif
\n\nDims: Full Scene (709,776 points)
\n\nBasic Stats Min Max Mean Stdev
\n\n Band 1 216 1134 247.462656 21.151406\n\n Band 2 155 1700 229.574581 35.088444\n\n Band 3 195 2047 328.013562 61.678726\n\n Band 4 186 2047 317.227769 94.644968\n\n Band 5 13 2047 197.929383 78.598665\n\n Band 6 184 2047 450.883749 75.600431\n\n Band 7 202 2047 625.945739 133.913580\n\n Band 8 166 2047 538.216023 114.308670\nattaching the file of visualizing with red= band 5, green=3, blue =band 2. But how to get info regarding other bands.
", "rawMarkdown": "attaching the file of visualizing with red= band 5, green=3, blue =band 2. But how to get info regarding other bands.", "votes": null }, { "id": "151974", "postDate": "12/22/2016 17:54:12", "content": "@Attila, I've contacted dstl and they will look into this.
", "rawMarkdown": "Attila, I've contacted dstl and they will look into this.", "votes": null }, { "id": "152012", "postDate": "12/22/2016 21:58:13", "content": "@shawn,\nIn the \"Layer Properties\" dialog, to choose bands 6,4,3 as RGB you have to actually select 5, 3, 2 because the numbering is shifted (the first band displayed in the select box as \"Band 1\" is actually the first band in the multispectral file which is Band 2). It looks correct to me.
", "rawMarkdown": "shawn,\r\nIn the \"Layer Properties\" dialog, to choose bands 6,4,3 as RGB you have to actually select 5, 3, 2 because the numbering is shifted (the first band displayed in the select box as \"Band 1\" is actually the first band in the multispectral file which is Band 2). It looks correct to me.", "votes": null }, { "id": "153017", "postDate": "12/29/2016 14:35:01", "content": "@amaia, yes those do look correct for RGB. It would still be very helpful to have the official band designations from Wendy though.
", "rawMarkdown": "amaia, yes those do look correct for RGB. It would still be very helpful to have the official band designations from Wendy though.", "votes": null }, { "id": "153197", "postDate": "12/30/2016 11:34:20", "content": "Based on the WorldView online bandwidth info and QGis extracted bandlist metadata and data description
\n\nSensor-Bands WV3-Type WV3-label WV3 range Image-ID Notes
\n\nThanks for the breakdown, @FPP_UK.
\n\nSo \"P\" stands for \"Panchromatic,\" \"M\" stands for \"Multi-spectral\"; does anyone know what the \"A\" stands for?
", "rawMarkdown": "Thanks for the breakdown, @FPP_UK.\r\n\r\nSo \"P\" stands for \"Panchromatic,\" \"M\" stands for \"Multi-spectral\"; does anyone know what the \"A\" stands for?", "votes": null }, { "id": "154498", "postDate": "01/06/2017 14:32:02", "content": "@FPP_UK can you explain a bit more. \nDo the bands follow the same sequence as given in the meta description of following ? \nSo if the description in file states that the order is [ 6; 2; 3; 7; 4; 8; 5; 9;] , does it means that the red band is first followed by the costal, blue, red edge, greeen, NIR-1, yellow, NIR-2 ? \nI am still having trouble with visualizing this image.
", "rawMarkdown": "FPP_UK can you explain a bit more. \r\nDo the bands follow the same sequence as given in the meta description of following ? \r\nSo if the description in file states that the order is [ 6; 2; 3; 7; 4; 8; 5; 9;] , does it means that the red band is first followed by the costal, blue, red edge, greeen, NIR-1, yellow, NIR-2 ? \r\nI am still having trouble with visualizing this image.", "votes": null }, { "id": "154522", "postDate": "01/06/2017 16:18:13", "content": "@Attila: In my understanding yes, 6 (Red) would be the first band - best to open the original images in QGIS or another GUI, compute / look at the histograms for each of the bands.
\n\nAt that point you can also create yourself some single / multi-band images for visualisation e.g. for the SWIR ( 1,4,6) and try out some contrast enhancements & band cuts.
", "rawMarkdown": "Attila: In my understanding yes, 6 (Red) would be the first band - best to open the original images in QGIS or another GUI, compute / look at the histograms for each of the bands.\r\n\r\nAt that point you can also create yourself some single / multi-band images for visualisation e.g. for the SWIR ( 1,4,6) and try out some contrast enhancements & band cuts.", "votes": null }, { "id": "154560", "postDate": "01/06/2017 19:51:59", "content": "@fpp_uk well I already tried all the visualisation of given images with the order of band as suggested by you.(normalizing the values of 2% -98% to range 0 -1 ) even tried various other combinations. The image from bands from M spectrum don't look like a visual RGB image. If you can look into some particular image and try to construct RGB image from multispectral image and post your approach, it will clear many doubts.
", "rawMarkdown": "fpp_uk well I already tried all the visualisation of given images with the order of band as suggested by you.(normalizing the values of 2% -98% to range 0 -1 ) even tried various other combinations. The image from bands from M spectrum don't look like a visual RGB image. If you can look into some particular image and try to construct RGB image from multispectral image and post your approach, it will clear many doubts.", "votes": null }, { "id": "154562", "postDate": "01/06/2017 20:02:09", "content": "[quote=Attila;154560]
\n\n@fpp_uk well I already tried all the visualisation of given images with the order of band as suggested by you.(normalizing the values of 2% -98% to range 0 -1 ) even tried various other combinations. The image from bands from M spectrum don't look like a visual RGB image. If you can look into some particular image and try to construct RGB image from multispectral image and post your approach, it will clear many doubts.
\n\n[/quote]
\n\nSee here: https://www.kaggle.com/aamaia/dstl-satellite-imagery-feature-detection/rgb-using-m-bands-example/
", "rawMarkdown": "[quote=Attila;154560]\r\n\r\n@fpp_uk well I already tried all the visualisation of given images with the order of band as suggested by you.(normalizing the values of 2% -98% to range 0 -1 ) even tried various other combinations. The image from bands from M spectrum don't look like a visual RGB image. If you can look into some particular image and try to construct RGB image from multispectral image and post your approach, it will clear many doubts. \r\n\r\n[/quote]\r\n\r\nSee here: https://www.kaggle.com/aamaia/dstl-satellite-imagery-feature-detection/rgb-using-m-bands-example/", "votes": null }, { "id": "155390", "postDate": "01/10/2017 23:52:20", "content": "Some additional links provided by Dstl:
\n\nWorldView-3 spacecraft information and specifications:\nhttps://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/95/DG_WorldView3_DS_forWeb_0.pdf
\n\nAbsolute Radiometric Calibration Adjustment Factors:\nhttps://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/209/DGConstellationAbsRadCalAdjustmentFactors_2015v2.pdf
\n\nWorldView-3 Relative Radiometric Response Curves (scroll down to find SWIR bands):\nhttps://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/208/WV03_technote_raduse_AttachmentA.pdf
\n\nRadiometric Use of WorldView-3:\nhttps://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/207/Radiometric_Use_of_WorldView-3_v2.pdf
", "rawMarkdown": "Some additional links provided by Dstl:\n \nWorldView-3 spacecraft information and specifications:\n[https://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/95/DG_WorldView3_DS_forWeb_0.pdf][1]\n \nAbsolute Radiometric Calibration Adjustment Factors:\n[https://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/209/DGConstellationAbsRadCalAdjustmentFactors_2015v2.pdf][2]\n \nWorldView-3 Relative Radiometric Response Curves (scroll down to find SWIR bands):\n[https://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/208/WV03_technote_raduse_AttachmentA.pdf][3]\n \nRadiometric Use of WorldView-3:\n[https://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/207/Radiometric_Use_of_WorldView-3_v2.pdf][4]\n\n\n [1]: https://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/95/DG_WorldView3_DS_forWeb_0.pdf\n [2]: https://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/209/DGConstellationAbsRadCalAdjustmentFactors_2015v2.pdf\n [3]: https://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/208/WV03_technote_raduse_AttachmentA.pdf\n [4]: https://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/207/Radiometric_Use_of_WorldView-3_v2.pdf", "votes": null }, { "id": "156115", "postDate": "01/14/2017 10:03:16", "content": "@Attila + @amaia: \nIf you look at the TIFF metadata you get the following info:
\n\n[3-bands] 6nnn_n_n: TIFFTAG_IMAGEDESCRIPTION={ bandlist = [ 4; 3; 2;] }
\n\n[16-bands] 6nnn_n_n_M: TIFFTAG_IMAGEDESCRIPTION={ bandlist = [ 6; 2; 3; 7; 4; 8; 5; 9;] }
\n\n[SWIR] 6nnn_n_n_A: TIFFTAG_IMAGEDESCRIPTION={ bandlist = [ 10; 11; 12; 13; 14; 15; 16; 17;] }
\n\n[update]:\nThe band lists included in the 3-band + 16-band TIFF meta data are seemingly wrong!
\n\nIf you apply the WV-3 band list (included in my previous comment) and just use the bands in the images sequentially it seems you are getting the right results:
\n\nexample:
\n\nRGB from 16-bands -> use band 5 from image for Red (WV-3-Band-no:6), band 3 from image for green (WV-3-Band-no:4), band 2 from image for blue (WV-3-Band-no:3).
", "rawMarkdown": "Attila + @amaia: \r\nIf you look at the TIFF metadata you get the following info:\r\n\r\n[3-bands] 6nnn_n_n: TIFFTAG_IMAGEDESCRIPTION={ bandlist = [ 4; 3; 2;] }\r\n\r\n[16-bands] 6nnn_n_n_M: TIFFTAG_IMAGEDESCRIPTION={ bandlist = [ 6; 2; 3; 7; 4; 8; 5; 9;] }\r\n\r\n[SWIR] 6nnn_n_n_A: TIFFTAG_IMAGEDESCRIPTION={ bandlist = [ 10; 11; 12; 13; 14; 15; 16; 17;] }\r\n\r\n[update]:\r\nThe band lists included in the 3-band + 16-band TIFF meta data are seemingly wrong!\r\n\r\nIf you apply the WV-3 band list (included in my previous comment) and just use the bands in the images sequentially it seems you are getting the right results:\r\n\r\nexample: \r\n\r\nRGB from 16-bands -> use band 5 from image for Red (WV-3-Band-no:6), band 3 from image for green (WV-3-Band-no:4), band 2 from image for blue (WV-3-Band-no:3).", "votes": null }, { "id": "158896", "postDate": "01/30/2017 17:10:57", "content": "Post removed by author
", "rawMarkdown": "Post removed by author", "votes": null }, { "id": "161396", "postDate": "02/13/2017 15:21:08", "content": "@ Wendy.. First of all thanks a lot for helping us a lot during the competition at every step. You guys at kaggle make data science very interesting field. :)
\n\nFor: @Wendy and @FPP_Uk \nThe calibration guide provided requires the parameter of abscalfactor and effective bandwith along with the solar zenith angle and date-time of the image capture ( for calculating the earth sun distance as described in the file.) These values are described in the IMD file of worldview 3. \nIf we can calibrate our image properly we can achieve good results. Without proper calibration it is hard to get good results.. Thanks a lot for help. :)
", "rawMarkdown": "Wendy.. First of all thanks a lot for helping us a lot during the competition at every step. You guys at kaggle make data science very interesting field. :) \n\nFor: @Wendy and @FPP_Uk \nThe calibration guide provided requires the parameter of abscalfactor and effective bandwith along with the solar zenith angle and date-time of the image capture ( for calculating the earth sun distance as described in the file.) These values are described in the IMD file of worldview 3. \nIf we can calibrate our image properly we can achieve good results. Without proper calibration it is hard to get good results.. Thanks a lot for help. :)", "votes": null } ], "comments": [ { "id": 151628, "author_name": "shawn775", "author_url": "", "post_date": "12/21/2016 15:33:45", "content": "Agreed. We know it is the WorldView 3 satellite, and can lookup the specs for that. But I can't figure out which bands in the images correspond to the satellite wavelengths.
", "votes": null, "replies": [] }, { "id": 151665, "author_name": "fragglerox", "author_url": "", "post_date": "12/21/2016 18:38:31", "content": "http://satimagingcorp.s3.amazonaws.com/site/pdf/WorldView3-DS-WV3-Web.pdf\nhas the band-by-band frequencies.
\n\nThe header of the GeoTIFFS contain the band ordering.
\n\ngdalinfo sixteen_band/6150_3_2_P.tif
\nFiles: sixteen_band/6150_3_2_P.tif
\nSize is 3403, 3348 --- bandList = [ 1; ]
Files: sixteen_band/6150_3_2_M.tif
\nSize is 851, 837 --- bandList = [ 6; 2; 3; 7; 4; 8; 5; 9;]
Files: sixteen_band/6150_3_2_A.tif
\nSize is 137, 134 --- bandList = [ 10; 11; 12; 13; 14; 15; 16; 17;]
So are they like this then? I'm not sure if this is correct. If I load an M image into QGIS and set RGB to bands 6,4,3 as below, it does not look like a true color image.
\n\nP Images\n1 : Panchromatic 400-800 nm
\n\nM images\n2 : Coastal 400-450
\n3 : Blue 450-510
\n4 : Green 510-580
\n5 : Yellow 585-625
\n6 : Red 630-690
\n7 : Red edge 705-745
\n8 : Near-IR1 770-895
\n9 : Near-IR2 860-1040
@wendy kan Can you please look into it. \nI am attaching the figure of 6010_0_0_M with merging the bands as described by timbabweSz. \nI don't think this is the correct labels for bands.
", "votes": null, "replies": [] }, { "id": 151933, "author_name": "fragglerox", "author_url": "", "post_date": "12/22/2016 15:57:47", "content": "I'm looking at 6100_2_2_M.tif. The GeoTiff header gives the scrambled order, but comparing the band means for the 3-band RGB with 6100_2_2_M.tif, the fifth/third/second M bands match up, more or less.
\n\nThe band order given in the header looks incorrect.
\n\nFilename: /net/tmp/dstl/three_band/6100_2_2.tif
\n\nDims: Full Scene (11,356,459 points)
\n\nBasic Stats Min Max Mean Stdev
\n\n Band 1 76 1717 198.610207 77.636097\n\n Band 2 215 1949 328.566782 61.056596\n\n Band 3 154 1664 229.807967 34.622226\nFilename: /net/tmp/dstl/sixteen_band/6100_2_2_M.tif
\n\nDims: Full Scene (709,776 points)
\n\nBasic Stats Min Max Mean Stdev
\n\n Band 1 216 1134 247.462656 21.151406\n\n Band 2 155 1700 229.574581 35.088444\n\n Band 3 195 2047 328.013562 61.678726\n\n Band 4 186 2047 317.227769 94.644968\n\n Band 5 13 2047 197.929383 78.598665\n\n Band 6 184 2047 450.883749 75.600431\n\n Band 7 202 2047 625.945739 133.913580\n\n Band 8 166 2047 538.216023 114.308670\nattaching the file of visualizing with red= band 5, green=3, blue =band 2. But how to get info regarding other bands.
", "votes": null, "replies": [] }, { "id": 151974, "author_name": "wendykan", "author_url": "", "post_date": "12/22/2016 17:54:12", "content": "@Attila, I've contacted dstl and they will look into this.
", "votes": null, "replies": [] }, { "id": 152012, "author_name": "aamaia", "author_url": "", "post_date": "12/22/2016 21:58:13", "content": "@shawn,\nIn the \"Layer Properties\" dialog, to choose bands 6,4,3 as RGB you have to actually select 5, 3, 2 because the numbering is shifted (the first band displayed in the select box as \"Band 1\" is actually the first band in the multispectral file which is Band 2). It looks correct to me.
", "votes": null, "replies": [] }, { "id": 153017, "author_name": "shawn775", "author_url": "", "post_date": "12/29/2016 14:35:01", "content": "@amaia, yes those do look correct for RGB. It would still be very helpful to have the official band designations from Wendy though.
", "votes": null, "replies": [] }, { "id": 153197, "author_name": "fppkaggle", "author_url": "", "post_date": "12/30/2016 11:34:20", "content": "Based on the WorldView online bandwidth info and QGis extracted bandlist metadata and data description
\n\nSensor-Bands WV3-Type WV3-label WV3 range Image-ID Notes
\n\nThanks for the breakdown, @FPP_UK.
\n\nSo \"P\" stands for \"Panchromatic,\" \"M\" stands for \"Multi-spectral\"; does anyone know what the \"A\" stands for?
", "votes": null, "replies": [] }, { "id": 154498, "author_name": "attila007", "author_url": "", "post_date": "01/06/2017 14:32:02", "content": "@FPP_UK can you explain a bit more. \nDo the bands follow the same sequence as given in the meta description of following ? \nSo if the description in file states that the order is [ 6; 2; 3; 7; 4; 8; 5; 9;] , does it means that the red band is first followed by the costal, blue, red edge, greeen, NIR-1, yellow, NIR-2 ? \nI am still having trouble with visualizing this image.
", "votes": null, "replies": [] }, { "id": 154522, "author_name": "fppkaggle", "author_url": "", "post_date": "01/06/2017 16:18:13", "content": "@Attila: In my understanding yes, 6 (Red) would be the first band - best to open the original images in QGIS or another GUI, compute / look at the histograms for each of the bands.
\n\nAt that point you can also create yourself some single / multi-band images for visualisation e.g. for the SWIR ( 1,4,6) and try out some contrast enhancements & band cuts.
", "votes": null, "replies": [] }, { "id": 154560, "author_name": "attila007", "author_url": "", "post_date": "01/06/2017 19:51:59", "content": "@fpp_uk well I already tried all the visualisation of given images with the order of band as suggested by you.(normalizing the values of 2% -98% to range 0 -1 ) even tried various other combinations. The image from bands from M spectrum don't look like a visual RGB image. If you can look into some particular image and try to construct RGB image from multispectral image and post your approach, it will clear many doubts.
", "votes": null, "replies": [ { "id": 154562, "author_name": "aamaia", "author_url": "", "post_date": "01/06/2017 20:02:09", "content": "[quote=Attila;154560]
\n\n@fpp_uk well I already tried all the visualisation of given images with the order of band as suggested by you.(normalizing the values of 2% -98% to range 0 -1 ) even tried various other combinations. The image from bands from M spectrum don't look like a visual RGB image. If you can look into some particular image and try to construct RGB image from multispectral image and post your approach, it will clear many doubts.
\n\n[/quote]
\n\nSee here: https://www.kaggle.com/aamaia/dstl-satellite-imagery-feature-detection/rgb-using-m-bands-example/
", "votes": null, "replies": [] }, { "id": 155390, "author_name": "wendykan", "author_url": "", "post_date": "01/10/2017 23:52:20", "content": "Some additional links provided by Dstl:
\n\nWorldView-3 spacecraft information and specifications:\nhttps://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/95/DG_WorldView3_DS_forWeb_0.pdf
\n\nAbsolute Radiometric Calibration Adjustment Factors:\nhttps://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/209/DGConstellationAbsRadCalAdjustmentFactors_2015v2.pdf
\n\nWorldView-3 Relative Radiometric Response Curves (scroll down to find SWIR bands):\nhttps://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/208/WV03_technote_raduse_AttachmentA.pdf
\n\nRadiometric Use of WorldView-3:\nhttps://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/207/Radiometric_Use_of_WorldView-3_v2.pdf
", "votes": null, "replies": [] }, { "id": 161396, "author_name": "attila007", "author_url": "", "post_date": "02/13/2017 15:21:08", "content": "@ Wendy.. First of all thanks a lot for helping us a lot during the competition at every step. You guys at kaggle make data science very interesting field. :)
\n\nFor: @Wendy and @FPP_Uk \nThe calibration guide provided requires the parameter of abscalfactor and effective bandwith along with the solar zenith angle and date-time of the image capture ( for calculating the earth sun distance as described in the file.) These values are described in the IMD file of worldview 3. \nIf we can calibrate our image properly we can achieve good results. Without proper calibration it is hard to get good results.. Thanks a lot for help. :)
", "votes": null, "replies": [] } ] }, { "id": 156115, "author_name": "fppkaggle", "author_url": "", "post_date": "01/14/2017 10:03:16", "content": "@Attila + @amaia: \nIf you look at the TIFF metadata you get the following info:
\n\n[3-bands] 6nnn_n_n: TIFFTAG_IMAGEDESCRIPTION={ bandlist = [ 4; 3; 2;] }
\n\n[16-bands] 6nnn_n_n_M: TIFFTAG_IMAGEDESCRIPTION={ bandlist = [ 6; 2; 3; 7; 4; 8; 5; 9;] }
\n\n[SWIR] 6nnn_n_n_A: TIFFTAG_IMAGEDESCRIPTION={ bandlist = [ 10; 11; 12; 13; 14; 15; 16; 17;] }
\n\n[update]:\nThe band lists included in the 3-band + 16-band TIFF meta data are seemingly wrong!
\n\nIf you apply the WV-3 band list (included in my previous comment) and just use the bands in the images sequentially it seems you are getting the right results:
\n\nexample:
\n\nRGB from 16-bands -> use band 5 from image for Red (WV-3-Band-no:6), band 3 from image for green (WV-3-Band-no:4), band 2 from image for blue (WV-3-Band-no:3).
", "votes": null, "replies": [] }, { "id": 158896, "author_name": "philldonn", "author_url": "", "post_date": "01/30/2017 17:10:57", "content": "Post removed by author
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "151620": "The satellite data provided need to have different frequency of the EM wave associated with that band. It would be really helpful to understand and comprehend the data if the frequency of each band is provided.", "151628": "Agreed. We know it is the WorldView 3 satellite, and can lookup the specs for that. But I can't figure out which bands in the images correspond to the satellite wavelengths.", "151665": "http://satimagingcorp.s3.amazonaws.com/site/pdf/WorldView3-DS-WV3-Web.pdf\r\nhas the band-by-band frequencies.\r\n\r\nThe header of the GeoTIFFS contain the band ordering. \r\n\r\ngdalinfo sixteen_band/6150_3_2_P.tif \r\nFiles: sixteen_band/6150_3_2_P.tif \r\nSize is 3403, 3348 --- bandList = [ 1; ] \r\n*************************\r\n Files: sixteen_band/6150_3_2_M.tif \r\nSize is 851, 837 --- bandList = [ 6; 2; 3; 7; 4; 8; 5; 9;] \r\n***********************************\r\nFiles: sixteen_band/6150_3_2_A.tif \r\nSize is 137, 134 --- bandList = [ 10; 11; 12; 13; 14; 15; 16; 17;]", "151712": "So are they like this then? I'm not sure if this is correct. If I load an M image into QGIS and set RGB to bands 6,4,3 as below, it does not look like a true color image. \r\n\r\nP Images\r\n1 : Panchromatic 400-800 nm \r\n\r\nM images\r\n2 : Coastal 400-450 \r\n3 : Blue 450-510 \r\n4 : Green 510-580 \r\n5 : Yellow 585-625 \r\n6 : Red 630-690 \r\n7 : Red edge 705-745 \r\n8 : Near-IR1 770-895 \r\n9 : Near-IR2 860-1040", "151886": "wendy kan Can you please look into it. \r\nI am attaching the figure of 6010_0_0_M with merging the bands as described by timbabweSz. \r\nI don't think this is the correct labels for bands.", "151933": "I'm looking at 6100_2_2_M.tif. The GeoTiff header gives the scrambled order, but comparing the band means for the 3-band RGB with 6100_2_2_M.tif, the fifth/third/second M bands match up, more or less. \r\n\r\nThe band order given in the header looks incorrect.\r\n*****************************************************\r\nFilename: /net/tmp/dstl/three_band/6100_2_2.tif\r\n\r\nDims: Full Scene (11,356,459 points)\r\n\r\nBasic Stats Min Max Mean Stdev\r\n\r\n Band 1 76 1717 198.610207 77.636097\r\n\r\n Band 2 215 1949 328.566782 61.056596\r\n\r\n Band 3 154 1664 229.807967 34.622226\r\n\r\n\r\n*******************************************\r\nFilename: /net/tmp/dstl/sixteen_band/6100_2_2_M.tif\r\n\r\nDims: Full Scene (709,776 points)\r\n\r\n\r\nBasic Stats Min Max Mean Stdev\r\n\r\n Band 1 216 1134 247.462656 21.151406\r\n\r\n Band 2 155 1700 229.574581 35.088444\r\n\r\n Band 3 195 2047 328.013562 61.678726\r\n\r\n Band 4 186 2047 317.227769 94.644968\r\n\r\n Band 5 13 2047 197.929383 78.598665\r\n\r\n Band 6 184 2047 450.883749 75.600431\r\n\r\n Band 7 202 2047 625.945739 133.913580\r\n\r\n Band 8 166 2047 538.216023 114.308670", "151957": "attaching the file of visualizing with red= band 5, green=3, blue =band 2. But how to get info regarding other bands.", "151974": "Attila, I've contacted dstl and they will look into this.", "152012": "shawn,\r\nIn the \"Layer Properties\" dialog, to choose bands 6,4,3 as RGB you have to actually select 5, 3, 2 because the numbering is shifted (the first band displayed in the select box as \"Band 1\" is actually the first band in the multispectral file which is Band 2). It looks correct to me.", "153017": "amaia, yes those do look correct for RGB. It would still be very helpful to have the official band designations from Wendy though.", "153197": "Based on the WorldView online bandwidth info and QGis extracted bandlist metadata and data description\r\n\r\n**Sensor-Bands\tWV3-Type\tWV3-label\tWV3 range\tImage-ID\tNotes**\r\n\r\n 1. Panchromatic\t Panchromatic\t450-800nm\t6xxx_n_n_P\t\t0.31m resolution\r\n 2. Multi-spectral\tCoastal\t400-450nm\t6xxx_n_n_M, RGB\t\t1.24m resolution\r\n 3. Multi-spectral\tBlue\t450-510nm\t6xxx_n_n_M, RGB\t\t1.24m resolution\r\n 4. Multi-spectral\tGreen\t510-580nm\t6xxx_n_n_M, RGB\t\t1.24m resolution\r\n 5. Multi-spectral\tYellow\t585-625nm\t6xxx_n_n_M\t\t1.24m resolution\r\n 6. Multi-spectral\tRed\t630-690nm\t6xxx_n_n_M\t\t1.24m resolution\r\n 7. Multi-spectral\tRed Edge\t705-745nm\t6xxx_n_n_M\t\t1.24m resolution\r\n 8. Multi-spectral\tNear-IR1\t770-895nm\t6xxx_n_n_M\t\t1.24m resolution\r\n 9. Multi-spectral\tNear-IR2\t860-1040nm\t6xxx_n_n_M\t\t1.24m resolution\r\n 10. SWIR\tSWIR-1\t1195-1225nm\t6xxx_n_n_A\t\tresolution reduced from 3.7m to 7.5m\r\n 11. SWIR\tSWIR-2\t1550-1590nm\t6xxx_n_n_A\t\tresolution reduced from 3.7m to 7.5m\r\n 12. SWIR\tSWIR-3\t1640-1680nm\t6xxx_n_n_A\t\tresolution reduced from 3.7m to 7.5m\r\n 13. SWIR\tSWIR-4\t1710-1750nm\t6xxx_n_n_A\t\tresolution reduced from 3.7m to 7.5m\r\n 14. SWIR\tSWIR-5\t2145-2185nm\t6xxx_n_n_A\t\tresolution reduced from 3.7m to 7.5m\r\n 15. SWIR\tSWIR-6\t2185-2225nm\t6xxx_n_n_A\t\tresolution reduced from 3.7m to 7.5m\r\n 16. SWIR\tSWIR-7\t2235-2285nm\t6xxx_n_n_A\t\tresolution reduced from 3.7m to 7.5m\r\n 17. SWIR\tSWIR-8\t2295-2365nm\t6xxx_n_n_A\t\tresolution reduced from 3.7m to 7.5m", "154387": "Thanks for the breakdown, @FPP_UK.\r\n\r\nSo \"P\" stands for \"Panchromatic,\" \"M\" stands for \"Multi-spectral\"; does anyone know what the \"A\" stands for?", "154498": "FPP_UK can you explain a bit more. \r\nDo the bands follow the same sequence as given in the meta description of following ? \r\nSo if the description in file states that the order is [ 6; 2; 3; 7; 4; 8; 5; 9;] , does it means that the red band is first followed by the costal, blue, red edge, greeen, NIR-1, yellow, NIR-2 ? \r\nI am still having trouble with visualizing this image.", "154522": "Attila: In my understanding yes, 6 (Red) would be the first band - best to open the original images in QGIS or another GUI, compute / look at the histograms for each of the bands.\r\n\r\nAt that point you can also create yourself some single / multi-band images for visualisation e.g. for the SWIR ( 1,4,6) and try out some contrast enhancements & band cuts.", "154560": "fpp_uk well I already tried all the visualisation of given images with the order of band as suggested by you.(normalizing the values of 2% -98% to range 0 -1 ) even tried various other combinations. The image from bands from M spectrum don't look like a visual RGB image. If you can look into some particular image and try to construct RGB image from multispectral image and post your approach, it will clear many doubts.", "154562": "[quote=Attila;154560]\r\n\r\n@fpp_uk well I already tried all the visualisation of given images with the order of band as suggested by you.(normalizing the values of 2% -98% to range 0 -1 ) even tried various other combinations. The image from bands from M spectrum don't look like a visual RGB image. If you can look into some particular image and try to construct RGB image from multispectral image and post your approach, it will clear many doubts. \r\n\r\n[/quote]\r\n\r\nSee here: https://www.kaggle.com/aamaia/dstl-satellite-imagery-feature-detection/rgb-using-m-bands-example/", "155390": "Some additional links provided by Dstl:\n \nWorldView-3 spacecraft information and specifications:\n[https://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/95/DG_WorldView3_DS_forWeb_0.pdf][1]\n \nAbsolute Radiometric Calibration Adjustment Factors:\n[https://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/209/DGConstellationAbsRadCalAdjustmentFactors_2015v2.pdf][2]\n \nWorldView-3 Relative Radiometric Response Curves (scroll down to find SWIR bands):\n[https://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/208/WV03_technote_raduse_AttachmentA.pdf][3]\n \nRadiometric Use of WorldView-3:\n[https://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/207/Radiometric_Use_of_WorldView-3_v2.pdf][4]\n\n\n [1]: https://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/95/DG_WorldView3_DS_forWeb_0.pdf\n [2]: https://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/209/DGConstellationAbsRadCalAdjustmentFactors_2015v2.pdf\n [3]: https://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/208/WV03_technote_raduse_AttachmentA.pdf\n [4]: https://dg-cms-uploads-production.s3.amazonaws.com/uploads/document/file/207/Radiometric_Use_of_WorldView-3_v2.pdf", "156115": "Attila + @amaia: \r\nIf you look at the TIFF metadata you get the following info:\r\n\r\n[3-bands] 6nnn_n_n: TIFFTAG_IMAGEDESCRIPTION={ bandlist = [ 4; 3; 2;] }\r\n\r\n[16-bands] 6nnn_n_n_M: TIFFTAG_IMAGEDESCRIPTION={ bandlist = [ 6; 2; 3; 7; 4; 8; 5; 9;] }\r\n\r\n[SWIR] 6nnn_n_n_A: TIFFTAG_IMAGEDESCRIPTION={ bandlist = [ 10; 11; 12; 13; 14; 15; 16; 17;] }\r\n\r\n[update]:\r\nThe band lists included in the 3-band + 16-band TIFF meta data are seemingly wrong!\r\n\r\nIf you apply the WV-3 band list (included in my previous comment) and just use the bands in the images sequentially it seems you are getting the right results:\r\n\r\nexample: \r\n\r\nRGB from 16-bands -> use band 5 from image for Red (WV-3-Band-no:6), band 3 from image for green (WV-3-Band-no:4), band 2 from image for blue (WV-3-Band-no:3).", "158896": "Post removed by author", "161396": "Wendy.. First of all thanks a lot for helping us a lot during the competition at every step. You guys at kaggle make data science very interesting field. :) \n\nFor: @Wendy and @FPP_Uk \nThe calibration guide provided requires the parameter of abscalfactor and effective bandwith along with the solar zenith angle and date-time of the image capture ( for calculating the earth sun distance as described in the file.) These values are described in the IMD file of worldview 3. \nIf we can calibrate our image properly we can achieve good results. Without proper calibration it is hard to get good results.. Thanks a lot for help. :)" }, "source": "meta" }