[email protected] wrote: > > Hello, > > Does anyone could confirms it ? > > On Friday, March 6, 2015 at 3:35:05 PM UTC, [email protected] wrote: > > Thanks, Jordi, > > Now I am running the CloudDetectionExample, and my goal is to > adapt this method to detect the cloud automatically, which is mean > that i should combine the classification method and spectral angle > together. > > My first problem here is that: what exactly the referencePixel > mean? > I have found one reply from a discussion several months ago which > said:"pixel components parameters are a pixel value in the input > image taken on a cloud which is used as a reference for spectral > angle computation".So whether the "a pixel value" is the DN > (Digital Number) Value of the specified pixel? Just because there > are 4 bands, so the values have 4 components, Is that right? > > If the image only have 3 bands (Like RGB colorful images, and such > kind of images are more normal and available that tiff with four > bands), So I only need to provide 3 components of a cloud pixel, > Is that right? > > Best wishes!
Hi, The reference pixel is the one with which every pixel in the image would be compared using the spectral angle. If you use the example in OTB, it assumes that the input image has 4 bands and therefore it takes 4 values as spectral bands. VectorPixelType referencePixel; referencePixel.SetSize(4); referencePixel.Fill(0.); referencePixel[0] = (atof(argv[5])); referencePixel[1] = (atof(argv[6])); referencePixel[2] = (atof(argv[7])); referencePixel[3] = (atof(argv[8])); cloudDetection->SetReferencePixel(referencePixel); You can modify the example to use as many bands as your input image has. Jordi > > On Friday, March 6, 2015 at 12:53:42 PM UTC, Jordi Inglada wrote: > > [email protected] wrote: > > > > Thanks very much for your reply,Jordi. > > > > I have read the first paper, it need cloud and cloud shadow > match, and > > the processing was more complicated than OTB. > > > > As refer to OTB cloud detection method,I have another > question:the OTB > > cloud-detect method need specify a pixel which possibly is > cloud, that > > means it can not recognize cloud automatically.If someone > adopt the > > OTB cloud detection method to process large amount of images > or run > > real-time on-board the remote sensing satellite, Obviously, > one can > > not pick a cloud pixel for the OTB program.So my question > is: > > > > Does there any techniques to determine the reference cloud > spectral > > angle automatically? Or could I just preset the reference > cloud > > spectral angle? > > > > I know that some methods such as SVM could train the cloud > spectral > > angle classifier which could act as a reference，Any other > > suggestion?? > > > > Hi, > > I simple approach for which I have seen good results is indeed > supervised classification. If you can generate a set of > training samples (small polygons) for the cloud and for the > non-cloud classes, you can use either SVM or Random Forests to > implement your cloud screening procedure. > > In this case, you don't even have to code in C++ and you can > use the OTB applications. Have a look here: > https://www.orfeo-toolbox.org/CookBook/CookBookse14.html#x66-1010004.4 > > > Good luck. > > Jordi > > > Thanks again! > > > > On Thursday, March 5, 2015 at 12:37:58 PM UTC, Jordi Inglada > wrote: > > > > [email protected] wrote: > > > > > > Hello guys, > > > > > > Now I am a beginner to use the OTB and I have some > problems when > > using > > > the Cloud Detection Example. > > > > > > As far as I know that there are several techniques to > detect > > cloud > > > from the remote sensing imagery. In the OTB, it detects > the > > cloud > > > based on spectral angle principle and assume that the > image have > > four > > > spectral bands. In my mind the parameter setting should > affect > > the > > > detection result, and the parameters must set according to > > Sensor that > > > the camera adopted.But not every cameras have four > spectral > > bands. > > > > > > So My questions are: > > > > > > Does there any papers about the OTB Cloud Detection > method? > > > > > > Does it work well using only three spectral band? > > > > > > What is the basic principle to set the parameters? > > > > > > Anyone any suggestion is welcome! > > > > > > Thanks very much. > > > > > > -- > > > > Hi, > > > > You are right about the fact that there are many methods for > the > > detection of clouds. The spectral method proposed in OTB is > a very > > simple one (spectral angle and low-pass filtering before > > thresholding), but has the advantage of not needing any > particular > > spectral band (SWIR or thermal, for instance). > > > > If you want more sophisticated approaches, you should try to > > implement something inspired from Fmask[1] for single-date > > acquisitions or from MACCS[2] for multi-temporal series. > > > > Jordi > > > > [1] Zhu, Zhe, and Curtis E. Woodcock. "Object-based cloud > and > > cloud shadow detection in Landsat imagery." Remote Sensing > of > > Environment 118 (2012): 83-94. > > > > [2] Hagolle, Olivier, et al. "A multi-temporal method for > cloud > > detection, applied to FORMOSAT-2, VENµS, LANDSAT and > SENTINEL-2 > > images." Remote Sensing of Environment 114.8 (2010): > 1747-1755. -- -- Check the OTB FAQ at http://www.orfeo-toolbox.org/FAQ.html You received this message because you are subscribed to the Google Groups "otb-users" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.com/group/otb-users?hl=en --- You received this message because you are subscribed to the Google Groups "otb-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
