Introduction

TronGisPy aims to automate the whole GIS process on raster data using python interface. To get start, please see GettingStarted.ipynb. The main module are listed below:

• Raster: This module is Main class in TronGisPy. Use ras = tgp.read_raster('<file_path>') to read the file as Raster object. A Raster object contains all required attribute for a gis raster file such as .tif or .geotiff file including digital number for each pixel (ras.data), number of rows (ras.rows), number of columns (ras.cols), number of bands (ras.bands), geo_transform (ras.geo_transform), projection (ras.projection), no_data_value and metadata. The Raster object can also be plot with GeoDataFrame(shapefile) on the same canvas using ras.plot(). Functions like ras.reproject(), ras.remap() and ras.refine_resolution() are useful functions.

• CRS: Convert the projection sys between well known text (WKT) and epsg(tgp.epsg_to_wkt, tgp.wkt_to_epsg). Convert the indexing sys tem between numpy index and coordinate system(tgp.coords_to_npidxs, tgp.npidxs_to_coords).

• ShapeGrid: Interaction between raster and vector data including tgp.ShapeGrid.rasterize_layer, tgp.ShapeGrid.rasterize_layer_by_ref_raster, tgp.ShapeGrid.vectorize_layer, tgp.ShapeGrid.clip_raster_with_polygon and tgp.ShapeGrid.clip_raster_with_extent.

• DEMProcessor: General dem processing functions including tgp.DEMProcessor.dem_to_hillshade, tgp.DEMProcessor.dem_to_slope, tgp.DEMProcessor.dem_to_aspect, tgp.DEMProcessor.dem_to_TRI, tgp.DEMProcessor.dem_to_TPI and tgp.DEMProcessor.dem_to_roughness. normalizer.

• Interpolation: Interpolation for raster data on specific cells which are usually nan cells. Once majority or mean value in the filter (convolution) are prefered value for interpolation, tgp.Interpolation.majority_interpolation, tgp.Interpolation.mean_interpolation are written in numba to speed up the process. If Inverse Distance Weight (IDW) method is appropriate, tgp.Interpolation.gdal_fillnodata impolemented by GDAL can be called.

• Normalizer: Normalize the Image data for model training or plotting. Normalizer can be initialize from normalizer = tgp.Normalizer(). Function normalizer.fit_transform() can help to normalize the data. Function normalizer.clip_by_percentage can be used to clip the head and tail of the data to avoid the outlier affecting plotting.

• SplittedImage: Split raster images for machine learning model training. Use splitted_image = tgp.SplittedImage(raster, box_size, step_size=step_size) to initialize SplittedImage object. SplittedImage object have n_steps_h, n_steps_w, padded_rows, padded_cols, shape, n_splitted_images, padded_image attributes. Function splitted_image.apply() can be used to process all splitted images using the funtion. Function splitted_image.get_geo_attribute() helps to get the vector of all splitted images and return GeoDataFrame object. When the prediction on each image is done, splitted_image.write_splitted_images() can be called to combine the prediction results on each splitted images to have the same size as original raster image.

• TypeCast: Mapping the data type betyween gdal and numpy, and convert the gdal data type from integer to readable string. Because gdal use integer to represent defferent data types, tgp.get_gdaldtype_name() helps to convert the integer to its data type name in string. Also, once converting the data type between numpy and gdal is required, tgp.gdaldtype_to_npdtype and tgp.npdtype_to_gdaldtype can help.

• io: Create, read and update the raster from the raster file. Use tgp.read_raster to read raster file as Raster object. Functions tgp.get_raster_info and tgp.get_raster_extent can be used when you don't want to read all digital value of the raster into the memory. Function tgp.update_raster_info can used to update the infomation of the raster file such as projection and geo_transform. Finally, if you want to get the testing file, tgp.get_testing_fp can help.

Contributor

Author

• GoatWang/王選仲, AI Engineer from Thinktron
• DavidHuang/黃梓育(聯絡人), AI Engineer from Thinktron
• Funskie/方子齊, AI Engineer from Thinktron

Domain Instructor

• 沈哲緯, Tech Lead of Thinktron
• YuHsiang/王禹翔(聯絡人), Remote Sensing Engineer, Section Manager from Thinktron
• 周立生, RD Engineer, Section Manager from Thinktron
• 鄧澤揚, RD Engineer from Thinktron

Getting Started

To get start, please see GettingStarted.ipynb.

Install

Python Version

Python3.6 and Python3.7 is tested.

Windows

1. Install preinstalls from pre-build wheel package

   • Python3.6
     • GDAL==3.0.4
     • Fiona==1.8.13
     • Shapely==1.6.4.post2
     • geopandas==0.7.0
     • Rtree==0.9.4
     • opencv_python>=4.1.2
   • Python3.7
     • GDAL==3.3.1
     • Fiona==1.8.20
     • Shapely==1.7.1
     • geopandas==0.9.0
     • Rtree==0.9.7
     • opencv_python==4.5.3

2. Install TronGisPy

   pip install TronGisPy
   

Linux

• Python3.6

  1. Build GDAL==3.0.4 by yourself
  2. Build opencv by yourself
  3. install other preinstalls from public pypi server

  pip install GDAL==3.0.4 Fiona==1.8.13 Shapely==1.6.4.post2 geopandas==0.7.0 Rtree>=0.9.4
  

  4. Install TronGisPy

  pip install TronGisPy
  

• Python3.7

  1. Build GDAL==3.3.1 by yourself
  2. Build opencv by yourself
  3. install other preinstalls from public pypi server

  pip install GDAL==3.3.1 Fiona==1.8.20 Shapely==1.1.1 geopandas==0.9.0 Rtree==0.9.7
  

  4. Install TronGisPy

  pip install TronGisPy
  

Docker

sudo docker run -it --rm jeremy4555/trongispy:latest

For Developer

Build

python setup.py sdist bdist_wheel

Docker Build

sudo docker build -t <dockerhub_id>/trongispy:latest -< Dockerfile

Reference

1. Logo

For Thinktron Worker

Install on Windows

1. Install preinstall thinktron pypi server

# python36
pip install -U --index-url http://192.168.0.128:28181/simple --trusted-host 192.168.0.128 GDAL>=3.0.4 Fiona>=1.8.13 Shapely>=1.6.4.post2 geopandas>=0.7.0 Rtree>=0.9.4 opencv_python>=4.1.2

# python37
pip install pyproj
pip install -U --index-url http://192.168.0.128:28181/simple --trusted-host 192.168.0.128 GDAL Fiona Shapely geopandas Rtree opencv_python

2. Install TronGisPy from thinktron pypi server (Windows)

pip install TronGisPy