0

I have 80 LANDSAT 8 Images. I can able to create composite image using Arcgis and then Pansharpen it to produce 15m multi-spectral image as a input for classification process. My requirement is to extract only roads vector(Polygon) data from LANDSAT 8 Images using ecognition developer. After segmentation, I have tried NDVI,NDWI ruleset in eCognition to unclassify vegetation and water areas(so that I can get road) but it doesn't seem possible to classify road data completely.

Is there any specific ruleset available for extracting roads?

I have licensed Arcgis 10.2.1, Ecognition developer 9 as of now.

1

As a rule of thumb using multi-spectral images with more than 30% more spatial extent than Pan image will lead to serious case of mixed pixels. This should be avoided at all cost. As is your case, this value is 50%.As such i suggest you increase resolution of the multi-spectral images. Also Landsat itself are a mid resolution images, I don't expect you to get road network of all classes. Maybe highways with 4 lanes or greater will be easily classified but that is pushing your luck too much. Best case would be to digitize it yourself.

  • I agree. Switching to S-2A/S-2B could be an alternative for this purpose. – aldo_tapia Sep 14 '17 at 10:56
  • @overarch Yeah! Can you guide me how to extract primary roads from LANDSAT 8 scenes? – Sarath SRK Sep 16 '17 at 5:25
  • @Gis_Boy you can publish your results here then we can help you. Can't hand hold you through the entire exercise. – OverArch Sep 25 '17 at 4:35
  • I have tried in eCognition and got primary roads only.Rulesets I have used are i) NDVI to unclassify vegetation ii) NDWI to unclassify water iii) BAI and REI to get road structures and used Density parameter also. This is what I have got! Looking for better results – Sarath SRK Sep 26 '17 at 4:14
  • @aldo_tapia I have checked S2A 10m resolution Image It gives result as same as LANDSAT – Sarath SRK Sep 26 '17 at 4:20
0

you should also try textural information for extracting roads, when editing the NN feature space

We're looking for long answers that provide some explanation and context. Don't just give a one-line answer; explain why your answer is right, ideally with citations. Answers that don't include explanations may be removed.

  • I dont understand, Could you explain bit clear? – Sarath SRK Sep 16 '17 at 5:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.