tag:blogger.com,1999:blog-64920340951923579232012-02-17T01:55:47.922+01:00Converts research in Photogrammetry, Remote Sensing and Computer Vision into easy-to-use 3D products and software.Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.comBlogger1261100tag:blogger.com,1999:blog-6492034095192357923.post-8686947848870835962011-12-19T20:33:00.002+01:002011-12-19T20:44:01.828+01:00

Metria CityModeler version 6.9 has been released. Information and download is available here. Online help is available here.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-18809757610355239612011-05-05T12:07:00.001+02:002011-05-05T12:19:11.533+02:00

The multispectral and panchromatic images are not aligned in the top view. In this image, this is manifested e.g. as blue shadows north west of sharp edges. In the bottom image, the images are  aligned. The pixel alignment in WiMap PanSharp has been improved. The result is a much more distinct pan sharpened image with less color bleeding around the edges. Follow this link to see the difference

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-86462173507196006282011-05-04T15:43:00.001+02:002011-05-04T15:43:55.968+02:00

Today, WiMap PanSharp 1.4 was released. The new version adds support for TIFF files with multiple samples per pixel stored in contiguous format. Support for single TIFF files with different bands in separate planes as well as multiple files with  a single band per file was supported already in earlier versions. WiMap PanSharp still only accepts 16 bits unsigned integers for each sample. Support

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-90457076777516364722011-03-16T20:04:00.002+01:002011-03-17T12:58:04.794+01:00

An evaluation version of WiMap PanSharp was released today. You can read the online help about WiMap PanSharp here. The demo software will add a diagonal pattern to the pansharpened image, but works otherwise as the full version. Download the demo software installer here. Send feedback to info@citymodeler.com.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-36025695979487760572011-03-09T13:35:00.000+01:002011-12-19T20:45:40.664+01:00

Metria CityModeler version 6.4 has been released. It includes a number of bug fixes, primarily related to saving pan sharpened images to TIFF files. Online help is available here.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-77086303045019183012011-02-22T15:03:00.002+01:002011-02-22T15:21:05.517+01:00

 Original and pansharpened multispectral images. The white lines on the football field are clearly visible in the pansharpened image. WiMap PanSharp was used. Metria CityModeler v6.3 was released today. It includes the new WiMap PanSharp software. It also has an improved functionality to measure objects on top of roofs. Several defects have been fixed as well. Online help is available here.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-3979184506595954452011-02-18T09:37:00.004+01:002011-02-18T10:16:21.428+01:00

This DSM was generated with WiMap DSM from pansharpened 1.2 m WorldView 2 images. Pansharpening was of course done with the new WiMap PanSharp software. The image shows two out of several hundred X3D tiles, automatically created in the DSM generation. A couple of short videos are here and here and the two X3D tiles are here and here.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-57757903503349842792011-02-16T11:48:00.001+01:002011-02-16T11:52:39.649+01:00

Many very high resolution (VHR) remote sensing satellites and digital aerial cameras record one high resolution panchromatic image and multiple low resolution spectral bands, such as red, gree, blue, and near infrared (NIR). The panchromatic image is recorded in the visible spectrum and often extends into the NIR wavelengths. The resolution of the multispectral bands is typically one fourth of

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-1473164613423662332011-01-25T13:23:00.001+01:002011-01-25T13:24:02.116+01:00

Yet a version of Metria CityModeler was released today. Version 6.2 is mainly an update for improved stability including many fixed defects.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-11692862340487521202010-12-25T17:33:00.002+01:002010-12-25T17:47:07.256+01:00

Metria CityModeler 6.1 is released. In addition to a number of bug fixes, it includes the following new features: The file name of each image is (optionally) displayed in the image windows. The color of the bottom object can be changed. The transparency of filled polygons can be changed. Zooming to a specific zoom ratio is enabled. The help is available online here.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-2662174210911592412010-12-09T12:51:00.001+01:002010-12-09T12:57:57.644+01:00

Metria CityModeler version 6.0 includes the following new features: The height of each building can optionally be displayed. Zooming can be centered around a selected point. The control panel is now on the left for easier access, since the primary image is to the left when displayed in vertical mode.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-50902928276139789442010-12-09T12:16:00.001+01:002010-12-09T12:57:15.782+01:00

In September, I attended the ISPRS Commission III "Computer Vision Meets Photogrammetry" conference in Paris. Last week, I attended the European LiDAR Mapping Forum (ELMF) in The Hague.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-54151607110152723482010-06-10T11:53:00.001+02:002010-06-10T12:20:16.823+02:00

A LiDAR point cloud has been classified in WiMap BM. Each connected "unit" was assigned a unique color. The colors were assigned to an image and a grid/raster/DSM of the point cloud was textured with the image.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-22225044924859431952010-06-02T13:29:00.002+02:002010-06-03T11:55:09.996+02:00

3D LiDAR point cloud, part of Stockholm, Copyright: Blom Sweden AB I have started working on WiMap BM again. I have been documenting the code and planned for future development. It takes some time since it has been a while since I worked with WiMap BM. The system is 20% faster when the data is on my SSD than when the data is on the standard hard disk.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-6581478912138935442010-05-27T12:19:00.001+02:002010-05-27T17:45:13.915+02:00

I bought a Solid State Drive (SSD) to check how much faster it is than my standard hard disk drive. The SSD is an Intel X25-V. I tested an image segmentation of a panchromatic 16-bit satellite scene of more than 200 Mb. Running a segmentation with my standard hard disk drive required 4 min 35 sec. Then I changed the path for the temporary files to the SSD drive, since the image caching uses

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-3156186273344543572010-05-25T10:36:00.003+02:002010-05-25T14:07:08.799+02:00

A few Digital Surface Models (DSM), ortho photos and 3D visualizations have been created using WiMap DSM from scanned aerial photos over Botkyrka. The four top images show interactive 3D visualizations while the bottom image shows a DSM as a grayscale image. There are also three videos here, here and here.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-10579266637096490412010-05-07T23:43:00.000+02:002010-05-07T23:43:20.469+02:00

Metria CityModeler 5.5 was released today. It includes a very brief documentation for WiMapDSM.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-61278002249268039912010-05-07T10:30:00.002+02:002010-05-07T12:59:09.314+02:00

Some of the 169 X3D tiles that were automatically generated during the DSM generation over Sundsvall from two QuickBird images. Click on the images for enlargements. In the last post, I described how WiMap DSM exports browser ready 3D visualizations simultaneously with the DSM generation. Yesterday, I tested this approach on two QuickBird scenes over Sundsvall. The results were excellent, so I

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-78117438753706386052010-04-30T11:23:00.029+02:002010-05-05T21:49:53.590+02:00

Two examples of a new semi realistic 3D visualization of digital surface models using a TIN and generalized colors. Top two images: Örebro, a composite of several X3D tiles and a video of a single X3D tile generated from two aerial images (0.25 meter resolution, 2500 m flying height). The stereo model includes 70 tiles. Bottom two images: Pristina, a composite of several X3D tiles and a

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-82528551331698861862010-04-20T22:00:00.001+02:002010-04-20T22:01:28.315+02:00

I had a presentation with Metria about generation of Digital Surface Models (DSM:s) and ortho photos from satellite images at the Swedish "Kartdagarna" (Eng. "Map days"). The presentation can be downloaded here (in Swedish).

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-89733452999177943992010-04-20T09:21:00.009+02:002010-05-05T17:10:02.114+02:00

Digital surface model (DSM) over Västerås, Sweden. Top: Perspective view textured with ortho photo. Middle: Ortho photo. Bottom: DSM as gray scale image. The images are subsampled and cropped versions of the original 0.6 m DSM. The DSM covers an area of approximatelly 60 square km. Note how well e.g. bridges over the road are modeled.  I used WiMap DSM to create a DSM and ortho photo over

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-52288154165948422772010-04-19T17:54:00.002+02:002010-05-05T17:11:44.341+02:00

I made a couple of X3D visualisations of the DSM over Norrköping, created from two 1 m Ikonos images. You can view the 3D visualisations here.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-28148057139605657892010-04-17T18:40:00.000+02:002010-04-17T18:40:52.808+02:00

I have uploaded some X3D models of DSM:s that were created with WiMap DSM. You can view them here.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-12315472631013247872010-04-12T09:21:00.010+02:002010-05-05T17:11:20.174+02:00

Aerial image of Örebro, Sweden (left) and digital surface model (DSM) generated with WiMap DSM (right). The images are here subsampled to a GSD of 1 meter. A DSM generator has recently been integrated in Metria CityModeler. An enhanced version is now also available as a stand alone software - WiMap DSM. Unlike the DSM generator in Metria CityModeler, WiMap DSM can also use aerial images for DSM

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-25899226179016581682010-03-31T12:18:00.002+02:002010-03-31T16:06:31.707+02:00

Ortho photo and DSM over Ringhals, Sweden, generated with WiMap DSM using Metria CityModeler 5.2  I made a DSM over Ringhals to further test the DSM generator using Metria CityModeler 5.2. There is certainly room for improvements, but the results are quite satisfactory. The DSM was generated from a 1 [m] panchromatic Ikonos pair. The cut out area includes four segmentation tiles, where matching

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-63857834145365359592010-03-29T20:16:00.002+02:002010-03-30T08:11:22.432+02:00

Version 5.1 of Metria CityModeler was released today. It includes automatic DSM generation, ortho photo generation and building top and base height estimation. Like version 5.0 this is an alpha release to be evaluated by Metria. A stable version will soon be released.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-7372257530241141792010-03-26T08:45:00.007+01:002010-04-20T11:32:03.439+02:00

Top image: DSM generated from two images with four pan sharpened bands (blue, green, red, near infrared). Bottom: DSM generated from two images with one panchromatic band. Both DSM:s have here been cropped and converted to 8-bit images with Ground Sample Distance (GSD) of 1.2 [m]. The originals are floating point images with GSD 0.6 [m]. The DSM:s cover Sundsvall (map at Eniro.se)The DSM

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-92030289603132366732010-03-25T12:32:00.002+01:002010-03-26T08:01:59.798+01:00

Since I started using Bugzilla, I have reported 30 bugs and enhancements and fixed 14. I have now used more hours than allocated for "Service enhancement" in the ESA project, so I will have to put an end to the project. I will make a release of Metria CityModeler 5.1 for Metria to evaluate.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-66661378858106347672010-03-25T08:09:00.001+01:002010-03-25T08:12:55.169+01:00

I found a post from september 2009, that I never published, so I do it now: I had a look at some Range Imaging (RIM) data at http://www.ipf.uni-karlsruhe.de/Personen/jutzi/wgi2/tests_datasets.html. Boris Jutzi, chair of "ISPRS Working Group I / 2 - LIDAR, SAR and optical sensors for Airborne and Spaceborne Platforms", showed the equipment at the LiDAR conference recently. The range images seam

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-72434753582042910952010-03-24T20:49:00.005+01:002010-04-20T22:10:29.070+02:00

Once the building is measured in one of the images, the top and bottom are automatically computed.  Metria CityModeler can automatically set the top and bottom heights of a building. I made a short video showing the process with the same buildings as in my last post. It uses a DSM as a first approximation and improves the estimate with some matching and modeling. I have not tested it very much,

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-17042938533343049752010-03-22T16:53:00.004+01:002010-04-20T22:10:59.040+02:00

The roof and base heights have been measured in a project with two pan sharpened QuickBird scenes (Sundsvall). I made another 1-minute video on how to measure the base height of an arbitrarily shaped building in Metria CityModeler. The process is as follows: The outline of the building is measured in one of the images The height of the top is adjusted by scrolling the mouse wheel The height of

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-36018219048552352902010-03-17T12:03:00.001+01:002010-04-20T22:11:25.518+02:00

I made a 1 minute video on how to measure gable roofs in Metria CityModeler 5.0. You do not actually see how the measurements are done, since they are combinations of keys (e.g. 'w' for changing the width) and scrolling the mouse wheel, but you do get an impression of how fast it is.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-2998799532079672032010-03-16T17:23:00.001+01:002010-03-16T18:34:24.566+01:00

I have made some improvements on Metria CityModeler the last few days: Nine bugs have been fixed The coordinates are now saved in 3D Shape files. Each point in a roof polygon must now be in 3D, since roofs now may be sloping. Up to now, the single height of each 2D building polygon with flat roof was saved as an attribute in the DBF file as requested by the users. I have started to enable

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-44025555767846478542010-03-12T08:54:00.002+01:002010-03-12T09:17:49.067+01:00

I started to use Bugzilla for keeping track of bugs and enhancements in Metria CityModeler - excellent tool! I can even close and comment bugs when I commit code in CVS, so the overhead is minimal.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-56560993198149660512010-03-11T10:49:00.002+01:002010-04-20T22:12:00.179+02:00

Automatically detected polygons in 3D point cloud. There are still some bugs and some buildings are not so well modeled, but WiMap BM is coming closer to a sharp release. WiMap BM (BuildingModeler) is ReVisitors software for generation of 3D building models from LiDAR point clouds. I have done som more development on the software so that it does not require neither known 2D polygons nor

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-11481225193199795042010-02-26T22:40:00.000+01:002010-02-26T22:40:05.292+01:00

I showed my progress on the ESA project on automatic building generation to the project partners at Metria. We had a constructive discussion about how to proceed. We will concentrate on processes that speed up the measurement and enhance the product and will focus less on the most advanced algorithms for fully automatic reconstruction on buildings.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-44905459816004990102010-02-08T09:35:00.000+01:002010-02-08T09:35:52.289+01:00

Read Jeff Thurston´s interview of me in V1 Magazine here!

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-25264403382761173322010-01-20T10:07:00.002+01:002010-04-20T22:12:35.677+02:00

Digital Surface Model for Sundsvall. This image is a reduced resolution 8-bit version of the generated DSM. Click for  a larger (but not full resolution) DSM. I have created a DSM over the Sundsvall area. For the first time it was created using Metria CityModeler, not my workbench programs. The DSM has a Ground Sample Distance of 0.6 m and was created from panchromatic WorldView images of

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-6438662291369489842010-01-14T12:23:00.010+01:002010-04-20T22:13:10.145+02:00

Metria CityModeler with pansharpened images, R,G,B(top) and NIR,G,B (bottom)    I received a pair of pan sharpened images over Sundsvall from Metria. I fixed so that they can be imported in Metria CityModeler. Pan sharpening is the process to produce high resolution spectral bands from low resolution spectral bands and a high resolution panchromatic band. Color makes interpretation (manual as

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-65405120158382346182010-01-13T22:21:00.003+01:002010-01-14T14:12:21.351+01:00

I found this link at Rymdstyrelsen, which refers to a project where Metria CityModeler is used for dispersion modeling. Also, the page on Digital Surface Models and ortho photos has been updated with some images.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com1tag:blogger.com,1999:blog-6492034095192357923.post-59301248992755574022010-01-12T10:49:00.003+01:002010-04-20T22:13:40.287+02:00

Top image: Segmentation of red, green, blue and NIR bandsMiddle image: Original low resolution red, green, blue and NIR bands Bottom image: The same segmentation showing NIR, green and blue bands.   Now that it is possible to edit every aspect of a building with a gable roof, it is time to make as much as possible automatic. Given the relatively low resolution of VHR satellite scenes, automation

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-30077652812021671172009-12-18T14:19:00.002+01:002009-12-18T14:22:04.881+01:00

Now, all aspects of a building with gable roof can be adjusted in Metria CityModeler. Once the building has been measured as described before, the position (X0, Y0, Z0), size (width, length, height), rotation and roof base height can be adjusted by pressing the keyboard/num pad keys 1-7 or first letter (x,y,z,w,l,h,r,b) while scrolling the mouse or dragging the Z slider. The numeric keys are

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-90144709259016580512009-12-14T17:59:00.005+01:002010-04-20T22:14:07.089+02:00

Metria CityModeler 5.0 with some measured gable roofs. I have continued to implement measurements of complex roof types. It is now possible to adjust both the top height and the base height of a gable roof. The top height is adjusted as usual with the Z-slider and the bottom is adjusted by dragging the slider while pressing the letter h on the keyboard. The process to measure a gable roof is

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-4526606514528645522009-12-14T09:03:00.000+01:002009-12-14T09:03:23.635+01:00

I opened a Twitter account for ReVisitor today. We´ll see how much I will use it - kind of new to me.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-65211504071384897432009-12-10T13:24:00.006+01:002009-12-11T09:45:27.494+01:00

Metria CityModeler with some gable roofs.The foundation for measuring gable roofs is laid. The basic form is given by only three mouse clicks in one of the images - the end points of the ridge and any point along either roof edge. To fully define the building, it will be possible to adjust the free parameters. For a gable roof these parameters are:positionorientation in 2Dwidth, lengthheight to

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-82203299113808134482009-12-07T16:59:00.014+01:002009-12-11T09:45:00.549+01:00

Metria CityModeler using two monitors - here one for the primary image and the other for the secondary image and the control panel.Next release of Metria CityModeler can use double monitors.Most of the day, I have however planned for (and to some tiny extent implemented) the ESA project. In Metria CityModeler, measurements are made in the primary image window and then transformed to object space

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-60696666771561195032009-11-20T12:05:00.008+01:002009-12-22T08:21:17.291+01:00

Preview of Metria CityModeler 5.0. Vertical layout, scroll bars and iconized buttons (top). In-place zoom out, resized primary image, displayed object view and hidden control panel (bottom). The most important enhancement in the upcoming release of Metria CityModeler is automatic generation of Digital Surface Models (DSM:s). In February 2007, the source code for Metria CityModeler was

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-76333649223323973872009-11-06T11:06:00.011+01:002009-12-07T17:08:26.217+01:00

Recently implemented roof types: flat, shed, gable, elliptic and semi-elliptic.I have mentioned earlier that ReVisitor and Metria have received financing from ESA. The project is briefly called "Automation of 3D City Models". In spite of the general title, it is restricted to 3D extraction of buildings from Very High Resolution (VHR) satellite stereo pairs. The building extraction will be

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-37303279468857592632009-11-03T16:07:00.015+01:002009-11-15T16:02:44.223+01:00

Digital Surface Model and ortho photo over Norrköping, here subsampled to 8 x 8 meter resolution from the original 1 x 1 meter resolution (north is down). Created from two Ikonos images. Water, clouds and cloud shadows make this DSM rather difficult. No manual editing has been performed. Click on images to enlarge.The Digital Surface Models have so far been created in image space - i.e. each

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com1tag:blogger.com,1999:blog-6492034095192357923.post-86281583515599868892009-11-01T15:40:00.003+01:002009-11-01T15:44:09.012+01:00

A new version of the wireframe-to-surface software WiMap W2S has been released. The new version checks all generated surfaces for planarity and marks non-planar surfaces in the output file, see more in the online help. WiMap W2S can be downloaded here.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-87042171252737083772009-10-26T08:17:00.009+01:002009-11-01T16:07:35.207+01:00

Digital Surface Model and Ikonos image over Norrköping.I presented the current status of the DSM generation at Metria. I explained the advantages of a segmentation/feature based approach over the area based YARD approach. The presentation is here (in Swedish).

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-69877455323345981052009-10-11T20:15:00.014+02:002009-11-08T14:04:28.067+01:00

Entire Sweden will be laser scanned (see here and here) at a resolution of approximately 0.5 points per square meter. Some demo data are now available here. I had a look and made some images to visualize the data (one of the point clouds consisting of approximately 10 million points, all data (C) 2009, Lantmäteriet, www.lantmateriet.se).Colored by height. Red is lower, blue is higher.Colored by

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-24342861442888062652009-09-28T23:25:00.003+02:002009-09-28T23:32:31.075+02:00

ReVisitor has now migrated from Microsoft Visual Studio 6 to Microsoft Visual Studio 2008 , which enables a considerably more effective software development.Furthermore, ReVisitors source code is now hosted on a secure external server and not only on discs, which occassionally were brought to a bank valve.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-58667081097392730202009-09-22T12:28:00.006+02:002009-09-22T12:42:00.927+02:00

Four tiles of segmentation overlayed on an Ikonos panchromatic image. The transition between the tiles is now smoother and the borders between tiles are included.I have the last few days improved the image segmentation further:The boundaries of each tile are saved so that all pixels are included in exactly one region/segmentThe pixels close to the tile borders are better segmentedA Run Length

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-6529631407988558312009-09-18T13:36:00.004+02:002009-10-02T08:26:50.031+02:00

I installed the latest wxWidgets release, 2.9. From now on wxWidgets will only support Unicode builds. This made updating ReVisitor´s code base more challenging than usual. Since most of the code uses wxString already, that part was easy. But 3rd party libraries took a bit longer time.If you by chance will go through the same process: The redefinition of NULL as (void*)0 in libtiff was the

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-11729729232398490082009-09-12T15:59:00.005+02:002009-09-12T16:03:51.068+02:00

It is now seven years since Metria CityModeler was introduced! Since then, 31 versions have been released, the latest being Metria CityModeler 4.6.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-12766266548272368162009-09-03T12:22:00.004+02:002009-09-10T23:10:32.291+02:00

I am currently in Paris, France participating in the two ISPRS conferences Laser Scanning and 3D City Models, Road Data bases and Traffic Monitoring (CMRT09). Inspiring presentations exactly in line with ReVisitor:s business idea.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-37845110123874392742009-08-31T09:48:00.004+02:002009-08-31T10:52:46.526+02:00

I have extended ReVisitor:s CityGML library with import capabilities, so that CityGML files can be read and exported to Shape, X3D, DXF and other formats. Up to now, it was only possible to write CityGML files. The import capabilities are somewhat limited right now, but will be extended on demand. It does already keep track of Buildings, Building parts, Walls, Levels-of-Details, interior/exterior

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-21392891470796720772009-08-28T13:58:00.002+02:002009-08-28T14:05:01.420+02:00

Business or pleasure? Well, I enjoy my work so much that it is hard to tell the difference... Still, I have mixed personal posts with business posts on this blog. As it is so easy to create new blogs, I decided to create one for personal matters and only blog about ReVisitor at this blog. The presonal blog will be in Swedish. Så, välkommen till hwiman.blogspot.com för att hälsa på min personliga

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-45885893596410289822009-08-26T17:35:00.004+02:002009-12-07T17:08:46.214+01:00

In May, I prepared a research proposal together with Metria. Today we had a tele-conference with the European Space Agency (ESA) and got to know that the proposal has been accepted. I still want the contract signed before I tell you more, but it is a project I look much forward to!

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-17011082491427605662009-08-25T16:45:00.005+02:002009-08-25T16:57:05.367+02:00

I want you all to know that there is an excellent free C++ cross platform library called wxWidgets! I have used it for many years, almost exclusively for Windows development. It is of course comforting to know that it would not require too much to port my software to Mac or Linux. Even if I would never need to compile on other platforms than Windows, the wxWidgets API is so easy to work with that

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-66309515902724278472009-08-25T13:14:00.004+02:002009-09-11T12:52:09.772+02:00

I am making progress every day with my development of a DSM generator, which uses my image segmentation. There are great "logistic" challenges in keeping track of all Levels-of-Details, images, tiles, chain code boundaries etc, but more and more is coming into place. I hope that I will make my first image matchings today...

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-55533961882758733542009-08-25T12:57:00.005+02:002009-08-25T13:55:49.759+02:00

I trained 5 x 1 km intervals together with Benny, a former colleague at Readsoft. We trained at the running tracks in Tureberg, Sollentuna. It is so comfortable to run on such tracks! I had my "competition shoes" (Reebok Kuai), which I rarely use because they do not have much cusion and support, but on these tracks they work very well and make you feel almost like a real runner!The Garmin

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com1tag:blogger.com,1999:blog-6492034095192357923.post-63137174037411630842009-08-22T21:16:00.002+02:002009-08-22T21:23:14.948+02:00

I rode what has become the standard cycling tour, 51 km, 1h 34 m. We lived here for many years before we realized how nice biking environments we have just nearby. Before that, we were almost always training on Ekerö. I am really satisfied with my speed today (>32 km/h) considering that it was quite windy, I was alone and I ran 21 km yesterday!

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-75568408948171619772009-08-21T16:02:00.004+02:002009-08-21T16:33:01.100+02:00

The Swedish orienteer Martin Johansson was in the lead of world mastership in Hungary when he got a large stick in his thigh. Showing great sportsmanship the following French, Norwegian and Chech orienteers stopped, one fetched help and they all helped to carry him from the woods! Martin has a 10 cm deep wound in a muscle, but is ok. Switzerland, Russia and Finland were unaware of the accident

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com1tag:blogger.com,1999:blog-6492034095192357923.post-39744788897953554402009-08-21T11:29:00.002+02:002009-08-21T11:33:17.848+02:00

Chain codes with bounding boxes. One color per connected set of chain codes.This may look like a Printed Circuit Board, but is just a set of chain code boundaries including their bounding boxes. The bounding boxes will help to speed up the search for matching candidates.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-50188541540419481192009-08-21T10:41:00.004+02:002009-08-21T10:57:49.536+02:00

For the fourth time this summer, Sara and I started the day running in the morning. Now it is official - we will try to do this every week. I realize that it is easy to say when the sun is shining. Getting out a gloomy November morning is not quite the same thing. Well, we´ll see... We were both much more tired this time than last week. I managed to run each kilometer sub-5, but I had to push

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-45131516296564283972009-08-20T15:43:00.007+02:002009-08-20T16:41:33.472+02:00

Last Tuesday night, I was out with Sara and some of her colleagues at Metria. We went around a rainy Stockholm on the steam ship "S/S Blidösund" and listened to beautiful, soft music by Anna Stadling & Idde Schultz. It was like an extra vacation in the middle of the week! I listen to their CD right now... (you can listen to it here!)

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-68389613987143540472009-08-20T10:10:00.005+02:002009-08-20T16:54:56.878+02:00

Yesterday, I was at Gröna Lund with my daughter, Frida, and a friend of hers. We had great fun! The new attraction, Insane, has an appropriate name... You can not tell up from down most of the time, thrilling. It is a very short ride though. Fortunatelly, there were not too many people at Gröna Lund yesterday. I tried the old Extreme for the first time and just as I thought, I found it worse than

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-22125918869794475732009-08-20T09:13:00.003+02:002009-08-20T09:18:36.964+02:00

I have found it hard to use wxWidgets for windows with scroll bars, but after reading their sample code it was not that hard after all. Now e.g. the fields in a DBF record can be viewed even if they are too many to fit vertically in a dialog.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-33099065307359443452009-08-19T08:35:00.005+02:002009-08-19T21:51:54.182+02:00

My compact chain code format is done. What was earlier saved as a 170 Mb Shape file is now saved in less than 2.5 Mb. Not only are the files smaller, but the compact chain code structure is maintained in memory as well. This will be really useful, I think.When an image segmentation is tiled, one chain code file is produced per tile. These are collected to a single ZIP arhive including an XML file

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-56560841074210450032009-08-17T18:21:00.007+02:002009-08-17T19:46:04.000+02:00

A few days ago, I implemented chain coding of segmented boundaries. Now, this implementation was a bit quick-and-dirty... I actually save double precision 3D (!) coordinates for each point, rather than chain codes. Since I splice all these points from different tiles into one list, the size of the list becomes really large. A single Shape file with chain codes could become 170 Mb for large images

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-46041158220795105382009-08-17T15:16:00.006+02:002009-08-17T15:43:19.885+02:00

I have started to upload Google Earth KML files with my training to www.wiman.nu/Hakan/training. I haven´t made any HTML page for it, so they are just listed file by file. Probably, I will never organize it more than this...Yesterday, I was running intervals up and down a nearby road. Not very fun, not very fast. I was planning to go to a sport track, but I felt already the very first steps, that

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com4tag:blogger.com,1999:blog-6492034095192357923.post-1302418815192239072009-08-16T15:13:00.003+02:002009-08-16T15:19:56.211+02:00

I made a re-exam for the Digital Image Processing course at KTH. It is one year since I started as guest lecturer in photogrammetry and remote sensing at KTH, one day a week. Now, the exam is my last assignment. I simply do not have the time for that commitment, even though I enjoy lecturing. I will however continue to supervise the M.Sc Thesis on photo texturing of 3D models.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-22964612968551683072009-08-14T17:33:00.003+02:002009-08-14T17:51:45.209+02:00

Most of the day I have been thinking about how to use the image segmentation for robust, high quality DSM generation. The current DSM generator works well, but is not very fast and tries desperately to find matches even when it is impossible due to e.g. water or clouds. By using the image segmentation, homogeneous areas will be bridged, incompatible matches will be easier to find and less points

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-86542438094105998732009-08-13T13:14:00.008+02:002009-08-17T14:31:47.174+02:00

Started the day by running with my wife to Bergshamra and then continued around Edsviken - 21 km. A bit cooler than last time, so we did not stop for a bath.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com3tag:blogger.com,1999:blog-6492034095192357923.post-91912265438518892692009-08-13T12:25:00.009+02:002009-08-13T13:54:13.784+02:00

Segmentation of a subsampled panchromatic Ikonos image (GSD=16 m). Segmented image (top), Chain codes (middle), Chain codes and vectors (bottom). Click on images for enlargements. Built up areas are at this level grouped into large, highly textured segments.I tested vectorization of the chain coded boundaries. The vectorization is done with the Douglas-Peucker method with a user defined

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-58200437673242375092009-08-12T18:27:00.007+02:002009-08-13T13:28:54.938+02:00

Region and boundary descriptors can now be derived from an image segmentation and saved to DBF files.Currently implemented region descriptors are:Average gray level in each bandStandard deviation of each bandArea (i.e. number of pixels currently)PerimeterCompactness = Perimeter*Perimeter/(4*PI*Area) - 1 for circle, larger for less compact regionsCurrently implemented boundary descriptors are:

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-65550587617708479532009-08-11T12:14:00.003+02:002009-08-11T12:57:28.916+02:00

Chain coded region boundariesThe boundaries between regions after an image segmentation need to be explicit. So I have started to implement chain codes. I choose 4-connectivity, rather than 8-connectivity since I want the boundaries to follow the pixels exactly. Generalization of these boundaries can be made later through vectorization.The chain codes can be exported to Shape files. The tile

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-11984799826710262672009-08-10T12:18:00.008+02:002009-08-14T18:02:20.683+02:00

The border between two tiles is (just) visible after segmentationThere has been a limit of about 10 million pixels for the image segmentation so far. But today I have fixed so that larger images are processed in tiles and each tile is pasted back into one image. So now it is possible to segment images of any size. The tile borders are not stitched together with any smart algorithm yet so there

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-80586768312380878762009-08-08T22:05:00.008+02:002009-08-17T14:37:27.815+02:00

Biked to Vaxholm, where I met my family, who spent the day on Grinda. We had a nice dinner in Vaxholm.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com3tag:blogger.com,1999:blog-6492034095192357923.post-54639209316099183882009-08-08T13:52:00.006+02:002009-08-08T22:26:38.095+02:00

Original (top) and segmented (bottom) images. Click on the images for enlargements.I tested my image segmentation on an aerial image over some mansion in Botkyrka. The image is 1670 x 919 pixels - about 1.5 million pixels or 4.5 Mb. The segmentation required 1.7 seconds on a Dell Vostro 200 and resulted in 3343 regions - an average of 459 pixels per region. The scale parameter is set to 100 and

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-1741815715033686652009-08-08T13:33:00.008+02:002009-08-08T14:15:09.865+02:00

Stereo pair from PRJ sensor fileAt last I found out how to compose the sensor model for PRJ files! With the solution at hand, I can not see why this took so long!Now, ReVisitor will continue to create 3D building models for Botkyrka. The building models are 3D boxes, known as Level-of-Detail (LoD) 1. The 2D outlines of the buildings are assigned heights, using aerial photos with known interior

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-75586618548610349662009-08-08T09:32:00.005+02:002009-08-08T09:48:00.145+02:00

I have finalized the implementation of the sensor model read from DBF files. I have created stereo pairs in my own workbench and can measure in 3D without problems. I do not model the radial distortion, but it seems to be set to zero in the data set I have anyway. I do get small y-parallaxes in some stereo pairs, but in others the match is perfect.I have continued to implement the sensor model

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-1265834919640179932009-08-06T11:45:00.005+02:002009-08-17T14:38:01.120+02:00

Started the day running to Stockholm with my wife (who continued to work from there) and then back home alone - 22 km all together. It is a hot day today, so we stopped for a bath in Brunnsviken. A wonderful start of the day - now I am ready for some work!

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com1tag:blogger.com,1999:blog-6492034095192357923.post-41593155640152521282009-08-05T15:26:00.006+02:002009-08-05T15:33:48.675+02:00

Updated the documentation of the ReVisitor CityGML library. The library itself has not been changed, but I added some comments to make its documentation more complete.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-30622474474761492972009-08-05T14:33:00.004+02:002009-08-20T09:19:13.061+02:00

I need to parse a DBF file, where some fields are specified to be integers, but they are really floating point numbers and it makes all the difference. So now I have made a fix, which for the first record in a DBF file checks if a claimed integer really is a double and in that case reformats it with the correct number of decimals. This kind of extra trouble takes so much time from "real" work!In

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-25613110519887104642009-08-04T20:19:00.004+02:002009-08-04T20:26:29.788+02:00

I have continued to parse PRJ sensor data, but I have so far failed to set up a valid sensor model. The parameters are correctly read from the PRJ file, but when I create a stereo pair, the images are not correctly aligned. Directions of coordinate axises of image, pixel and/or object systems are usually the problems, so I will continued to test different variations...

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-59862643365612097482009-08-04T09:23:00.007+02:002009-08-04T10:10:14.420+02:00

I have tested my image segmentation on a number of different images. I find the results quite satisfactory. Above is a segmentation of "goldhill", a standard test image in image analysis. Click here for more segmentations of the same image. Each segmentation differs with a user defined scale parameter. The first segmentation after the original in the animation is an information theoretic optimum.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-74642051932192693562009-08-03T14:40:00.010+02:002009-08-03T18:53:43.028+02:00

There are a number of ways to provide the sensor parameters for an image:Include the parameters in GeoTIFF, e.g. in the 4x4 transformation matrix. Rarely used, although a clear advantage is that the parameters are always attached to the correct image.Store the parameters of a rational polynomial in a *.rpb or *rpc.txt file (very high resolution satellite scenes like Ikonos, WorldView etc.).Read

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-68911871981060004382009-08-03T12:53:00.009+02:002009-08-17T14:38:30.790+02:00

I biked three out of six days in the Laholm 6-days. Daily distances of approximately 90-120 km. Wonderful environment for biking and a really nice arrangement! Very rainy and hilly the first day. Got a flat tire early the second day so I lost contact with the "peleton", but I got some company anyway. The third day was my best - a Google Earth view of it is here.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-46248668406353185492009-08-03T09:46:00.001+02:002009-08-03T09:50:30.087+02:00

I have now transfered most of the blog that I maintained on www.citymodeler.com to blogspot. Since I did not find a way to post-date old entries, I added the dates in the titles. So as you can see, the posts on August 2nd summarize the blogs since May 2009.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-74143379910496727932009-08-03T09:33:00.000+02:002009-08-03T10:32:15.729+02:00

I have continued to develop an object based region growing algorithm. There are some issues to be solved still, but most of it is working fine. I plan to use this to bridge homogeneous areas for DSM generation, but also for building reconstruction from aerial and remote sensing images.Region growing at different scales. Click on the image or here to view full resolution images. The image is from

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-253911815235469532009-08-03T09:32:00.001+02:002009-08-03T10:32:38.081+02:00

Delivered two sample DSM:s to Metria for quality assessment. One of them was this one:A DSM example generated from two Ikonos images over a mountainous area. View X3D model here.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-57560169551814023722009-08-03T09:31:00.001+02:002009-08-03T09:31:45.152+02:00

The DSM:s have so far been regularized by a simple 5x5 median filter. I have today implemented the filter that was used in the YARD DSM generator. This filter takes the quality of each matched pixel into account in a weighted average. Using this filter, the DSM:s look smoother and more realistic. I measure the number of large changes in height between each iteration. When using the new adaptive

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-89503575382276658692009-08-03T09:30:00.001+02:002009-08-03T09:31:06.601+02:00

Had a meeting at KTH about a M.Sc. Thesis, that I will supervise. The Thesis is on texturing 3D city models, using high resolution oblique aerial images.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-22363639447768171832009-08-03T09:26:00.000+02:002009-08-03T10:32:56.033+02:00

Yet a milestone in the development of the DSM generator! After yesterdays work on mutual preference, the quality of the DSM:s has improved significantly. There are still a number of features I will reimplement from my YARD DSM generator, but already now the results are quite ok. Matching is still time consuming because of the transformations, but a simple optimization today reduced the execution

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-56173898777820124892009-08-03T09:25:00.002+02:002009-08-03T09:26:19.981+02:00

Started to implement matching with mutual preference, i.e. an object point is only accepted if it is considered the best match using Image A as template and Image B as search image and vice versa. It turns out, that 72 percent of the matched point in a Very High Resolution satellite image pair are mutually preferred by both images. Only mutually preferred points are used to create the DSM.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-53858078599858486172009-08-03T09:25:00.001+02:002009-08-03T09:25:38.448+02:00

The DSM area of my test data is not aligned with the x and y axises. Thus there are DSM grid cells where no heights are assigned. When regularizing the DSM and also when propagating the DSM to the next image pyramid level, these "NO DATA" pixels are used as if they actually had assigned heights. This influences the quality of the DSM particularily at the borders. Today, I have more or less made

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com0tag:blogger.com,1999:blog-6492034095192357923.post-49229118748679306692009-08-03T09:24:00.003+02:002009-09-11T12:53:33.796+02:00

The link to the Swedish Space Board for the DSM generator is broken because they have reorganized their web site. I will keep an eye and see when/if the description comes online again.

Håkan Wimanhttp://www.blogger.com/profile/11638151452293128628noreply@blogger.com1