Diferenças

Aqui você vê as diferenças entre duas revisões dessa página.

--- geodma [2025/04/04 12:02] – [GeoDMA - Geographic Data Mining Analyst] thales
+++ geodma [2025/08/21 17:56] (atual) – [Load Raster and Apply Segmentation] thales
@@ Linha 4: / Linha 4: @@
 {{:thales:logo2017.png?300|}}
-Testes:
-  * [[https://www.dpi.inpe.br/geodma/download/download.php?FileName=TerraView-5.7.1-win64.exe|5.7.1|Teste == Link externo dentro da dpi, com https]]
-  * [[http://www.dpi.inpe.br/geodma/download/download.php?FileName=TerraView-5.7.1-win64.exe|5.7.1|Teste == Link externo dentro da dpi, sem https]]
-  * [[@../download/download.php?FileName=geodma-2.0.5-beta-setup.exe]]
-  * [[https://download.anydesk.com/linux/anydesk_6.4.3-1_amd64.deb|Link externo fora da dpi]]
 ===== Downloads =====
 ^ Release ^ TerraView Version ^ Date ^ Download link  ^ Release Notes ^
@@ Linha 47: / Linha 42: @@
 {{:thales:geodma-2-instalation.mp4?800|}}
+===== Quick Tutorial - From Raster to Land Cover Map using GeoDMA =====
+==== Load Raster and Apply Segmentation ====
+The first step is to use TerraView software to load and visualize image (download [[https://github.com/tkorting/remote-sensing-images/raw/refs/heads/master/pan-2025-03-04.tif| using this link]]). We selected the True Color composition for our 4 bands image (Blue, Green, Red, Nir), using bands 2, 1 and 0.
+We also applied the 2% image stretching to provide a fast visualization with high contrast. With the image on the screen, we selected the Image Segmentation method to produce regions according. We have selected the minimum number of pixels as 25 to avoid very small regions, and defined a similarity threshold of 0.01, based on the minimum euclidean distance of pixels with 4 bands to be considered homogeneous.
+|Loading Raster in TerraView {{ :thales:terraview-load-raster-rgb-2p.mp4?400 |1. Load Raster, 2. Define True Color composition, 3. Apply high contrast}} | Apply image segmentation in TerraView {{ :thales:terraview-segmentation.mp4?400 |Apply segmentation using Region Growing method}} |
+==== Extract Features and Select Samples ====
+Based on the regions obtained by the segmentation, we call GeoDMA Feature Extraction tool that produces a table full of features. **This process can take several minutes** (~15 minutes in this case). We also select samples from the images to train a Decision Tree Algorithm.
+|Feature Extraction in GeoDMA {{ :thales:geodma-feature-extraction.mp4?400 |Feature Extraction (slow process)}}|Sample Selection in GeoDMA{{ :thales:geodma-sampling.mp4?400 |Sample selection, based on Land Cover classes expected to be present on the image}}|
+==== Classify and Visualize Results ====
+The final step is to load the C5.0 classification algorithm, that based on the previously selected samples, produces a Decision Tree and classify all regions. We define a visualization scheme based on the labels included on the table to visualize the final classification.
+{{ :thales:geodma-classification.mp4?600 |Applying Decision Trees and Visualizing the classification}}
 ===== Description =====
 The toolbox integrates the following techniques: