Google Earth Engine 8: Introduction to Image Classification
HTML-код
- Опубликовано: 1 июл 2019
- Another powerful tool of the Google Earth Engine is the capability to classify images. It can be useful for a variety of cases, when you want to create a thematic map of a certain area rather than an image, or when you want to assess what land cover changes took place over a certain time period.
In the Earth Engine there are two ways to classify pixels of an image. You could do a supervised classification or an unsupervised classification.
This video will give an introduction on how to use classifiers in the GEE
If you found this video useful, consider buying me a coffee at: www.buymeacoffee.com/programsam ☕
Explaining all supervised classification classifiers: • Google Earth Engine 9:...
Previous video: • Google Earth Engine 7:... (Exporting Raster and Vector Data)
Next video: • Google Earth Engine 9:... (Clustering and Explaining All Clusterers)
GEE docs: developers.google.com/earth-e...
Thanks a lot for this useful tutorial!
Great work !
A great fan of teaching style . A video on Object-based classification of the output image please
Great work.please carry on the training program
it's very useful to me, thank you
Hi. Thank you very much for this nice tutorial
Great Tutorial. Please upload more contents. :)
You are the best, I hope you keep uploading content. Greetings from Chile // eres el mejor. Ojala sigas subiendo contenido. Saludos desde chile
Viva Chile stay strong during these rough times
thank you!
感谢！Thank u so much！
Hi, thank you very much! Have you done a land cover change analysis? Do you have any tips for that or a good link? I hope you upload more videos. Greetings from Switzerland :)
the .cart() function used to create an empty classifier is deprecated. It needs to be replaced with smileCart().
Very nice. I like your voice.
Plz make some lectures on GDAL and Raterio with phyton with Remote Sensing data.
Good day, Sir. I stumbled upon to your content while I was searching for remote sensing contents. I would like to ask if how do we know if the certain country has already data? btw, I am from PH.
it's good but it would be nice if you could put the code in the comments
Please do a supervised classification on Sentinel 2 imagery.
another question is regarding machine learning. can we train model in javascript and the extract the model for use i python based system
I've coded the entire thing but my final output is grayscale rather than the 4 color values I've given as palatte parameters.What should I do now, to remove the grey scale and have a colorful palatte
Hello, I would like to ask one thing. Can I use a geotiff images , as for the training inputs? I tried to do so, but I get error message, that user memonry exceeded. I want those data to detect burned areas with RD classification. Thanks in advance.
Thank you for this helpful tutorial! Would the whole process work the same way with, for example, Landsat 5 images from 1984?
Yes, absolutely
can we calculate the ndvi of wheat separately
and when I put Map.addLayer(selection)... it says "no bands to visualize", I put L8 instead of "selection" but it showed all the tiles for the entire country. How do I set it to my ROI / boundary only?
sir i am working over a project on forest fire, LSTM based image generation. so far i was using landsat 8 real time data set. but the problem with such satellite is image availability. can you suggest me any DATA SET SOME WHAT SIMILAR TO landsat 8 but continuous available over a detection region
Hello ! I want to make an image classification on Sentinel 2 imagery. However, I want all the bands to have spatial resolution of 10m . I need to make a resample process, however Ι have stuck. Have you any Idea, how to overcome this problem ?
I have a query. since you did not sperate the training sample into test samples, when you output confusion matrix, it is always going to classify the images with 100% accuracy, is it not? if what I told is correct, is there a way to seperate the samples and then use for confusion matrix? Thanks
That's a very good comment, but there is one caveat. Let's say you have 100 training pixels. These pixels learn the algorithm what the average pixel of that class looks like. but even then, some of those pixels might be very different from that calculated average. When the algorithm then classifies those same pixels, it is possible that for example 10 of those training pixels are very different from the average of the 100 pixels. during classification these 10 can then still be classified as another class.
When you have a small number of training data, the algorithm most likely classifies the training samples as the same category you defined it, because the characteristics of that class are for a large percentage determined by that pixel. but as pixel count increases, it can happen that some deviate more from these expected characteristics.
Does this make clear why you don't have to separate these values when creating the confusion matrix?
Hi, thank you so much for your help tutorials, I am wondering how I can assign attributes "LC" and "0" to an existing point feature that I imported into GEE, I have been trying for a while and it doesn't work.
Thanks for your help.
Have you come across this stack exchange question? gis.stackexchange.com/questions/301433/adding-properties-to-a-feature-in-google-earth-engine It seems to contain the info you need.
Thanks Sir, How to export final output classify image of land use?
You can find an instructional video on how to export images here: ruclips.net/video/bhl8dQkpVO4/видео.html
i am facing the problem of layer error and 1_1_1_0_0 is displayed in message box as error message . Can you please help me to solve it?? Thank you
Why do we use scale:30
According to the docs, scale is "A nominal scale in meters of the projection to sample in. If unspecified,the scale of the image's first band is used."
I don't think it matters much, if you leave it out it will automatically pick the resolution of the first band of the image layer. But a source can have multiple bands with multiple resolutions, and if you pick a different band it might use the wrong resolution. I believe this makes sure that each sample is 30 meters apart and you don't sample the same pixel twice.
I am getting this error after running the code.
hello great tutorial
I would like to ask how I can make a classification exporting polygons with more than 100 hectares, since when running the program it sends me an error "Computed value is too large". Greetings I hope you can help me.
you have to use an image reducer or something I think
can you share the code ?
1 + subscriber
classification: Layer error: Property 'Class' of feature '1_2_0' is missing.
first of all, thank you very much. i followed the video step by step ...it all works good but when it comes to image visualization this message appears in the console "Layer 2: Layer error: Image. visualize: Cannot provide a palette when visualizing more than one band."
When you visualize your layer you have to option to visualize one of the bands, or create a false color image by choosing a 3 bands for the red, green and blue color values. Palettes are a way to visualize one band with multiple colors, and by judging the error message you got, it is set to visualize an image using 3 bands as RGB, but it also received a color palette. at 11:09 you see the visualization parameters. Can you make sure the radio button on top is set to '1 band'? Then the palette option should be enabled.
how to perform unsupervised classification on google earth engine
There will be a video about unsupervised classification in a few weeks!
Errick bullock...