Remote sensing image classification tool and visualization application of WebGIS

---

Preface

With the last blog update web edition fme Launch of data processing platform , I will study models and fme as well as webgis combination , Build a lightweight remote sensing image data automatic classification platform . The core function is very simple , The customer finishes uploading remote sensing images , And select parameters , Then you can get one shp, As well as the visualization of extracting patches on the platform .

---

One 、 Remote sensing image classification tool 2.0

Two months ago, I updated an article about deeplabv3+ Semantic segmentation model and fme Blog of automatic extraction of ground objects from remote sensing images . There were several deficiencies in the tools at that time .

1、 Poor performance , A large number of raster data are written , Consuming performance .

2、 There is no optimization algorithm , The segmented patches are irregular and uneven . Not using optimization algorithms will make the results unusable .

3、 It does not consider that multiple resolution images need to train multiple models to meet the extraction function of different scenes .

So on a secondary basis , Classification tools 2.0 Perfectly solve the above shortcomings . Convert raster data into binary streams in fme Internal calculation , Save a lot of performance . Thank you here fme The official group strives to become lazy, such as the road optimization algorithm provided by Ru Bu Dong . With the blessing of this algorithm , Some roads even go beyond human drawing . The following figure shows the road extracted by the optimization algorithm .

Two 、webgis visualization

I use mabbox As a map engine , use postgis Database and geodjango Set up the front end 、 Back end 、 The association of the three databases . Realize dynamic visualization .

1. Upload data

Select the resolution level of remote sensing image

  Select the type of extracted figure

  Select the uploaded grid file , Support jpg、png、img、tif 4 Data in two formats , Grid data needs coordinates , requirement 2000 Coordinate system .

 

2. Calculation of data

The background calculation of data is mainly fme Templates , Will be extracted after the calculation shp The data is returned to the front-end Download Interface , At the same time, import the data postgis, adopt geodjango The wkb The spot information turns to geojson Input the front-end map to form visual data .

 3. Data visualization

The extracted data is displayed in web End , The image base map is the sky map used .

  At the same time, click the spot to highlight , At the same time, display the extracted attributes .

  The following figure shows the actual effect

If you delete this item in the personal Center , The backend will be deleted at the same time postgis The corresponding data of has been downloaded shp

---

summary

Simple function implementation , Rich in a lot of knowledge . From front-end to back-end to database , take gis Data visualization . In the later stage, we can consider automatically completing the slicing and publishing of grid data , And connect to the front end , The visualization effect will be better , At the same time, it also has great requirements for the performance of the background .