Image-Based Method Example

This method convert landslide polygon data to landslide polygon Images. These converted landslide images are used as a input to CNN for landslide classification

Import Image based module functions from Landsifier library

from image_based_model import make_ls_images
from image_based_model import classify_inventory_cnn
from image_based_model import plot_geometric_results
from image_based_model import read_shapefiles
import numpy as np

In this example, we are using six landslide inventories spread over Japan archipaelogo. Out of six-inventories, three inventories are earthquake-triggered and rest three inventories are rainfall-induced inventories.

• Earthquake-triggered inventories (Hokkaido, Iwata and Niigata region)

• Rainfall-induced inventories (Kumamoto, Fukuoka and Saka region)

Import Shapefiles of Earthquake and Rainfall triggered landslide Inventories

• Earthquake triggered inventories

earth_hokkaido_shapefile = read_shapefiles("E:/Germany Visit/Landslide Datasets/Japan Inventory/Earthquake_hokkaido_polygons.shp")
earth_iwata_shapefile = read_shapefiles("E:/Germany Visit/Landslide Datasets/Japan Inventory/Earthquake_iwata_polygons.shp")
earth_niigata_shapefile =read_shapefiles("E:/Germany Visit/Landslide Datasets/Japan Inventory/Earthquake_niigata_polygons.shp")

• Rainfall triggered inventories

rain_kumamoto_shapefile = read_shapefiles("E:/Germany Visit/Landslide Datasets/Japan Inventory/Rainfall_kumamoto_polygons.shp")
rain_fukuoka_shapefile = read_shapefiles("E:/Germany Visit/Landslide Datasets/Japan Inventory/Rainfall_fukuoka_polygons.shp")
rain_saka_shapefile = read_shapefiles("E:/Germany Visit/Landslide Datasets/Japan Inventory/Rainfall_saka_polygons.shp")

Convert landslide polygons to grayscale binary images

• Earthquake triggered inventories

feature_Earth_hokkaido=make_ls_images(earth_hokkaido_shapefile)
feature_Earth_iwata=make_ls_images(earth_iwata_shapefile)
feature_Earth_niigata = make_ls_images(earth_niigata_shapefile)

• Rainfall triggered inventories

feature_Rain_kumamoto = make_ls_images(rain_kumamoto_shapefile)
feature_Rain_fukuoka = make_ls_images(rain_fukuoka_shapefile)
feature_Rain_saka = make_ls_images(rain_saka_shapefile)

Taking one of the landslide inventory as the Testing inventory

• Case 1: Hokkaido (Earthquake -triggerd) is a testing inventory

Training the algorithm on rest five inventories and test it on Hokkaido inventory.

earthquake_inventory_features=np.vstack((feature_Earth_iwata,feature_Earth_niigata))
rainfall_inventory_features=np.vstack((feature_Rain_fukuoka,feature_Rain_saka,feature_Rain_kumamoto))
test_inventory_features=feature_Earth_hokkaido
predict_probability_hokkaido=classify_inventory_cnn(earthquake_inventory_features,rainfall_inventory_features,test_inventory_features)

Visualization of the classification results

plot_geometric_results(predict_probability_hokkaido)

• Case 2: Iwata (Earthquake -triggerd) is a testing inventory

Training the algorithm on rest five inventories and test it on Iwata inventory.

earthquake_inventory_features=np.vstack((feature_Earth_hokkaido,feature_Earth_niigata))
rainfall_inventory_features=np.vstack((feature_Rain_fukuoka,feature_Rain_saka,feature_Rain_kumamoto))
test_inventory_features=feature_Earth_iwata
predict_probability_iwata=classify_inventory_cnn(earthquake_inventory_features,rainfall_inventory_features,test_inventory_features)

Visualization of the classification results

plot_geometric_results(predict_probability_iwata)

• Case 3: Niigata (Earthquake -triggerd) is a testing inventory

Training the algorithm on rest five inventories and test it on Iwata inventory.

earthquake_inventory_features=np.vstack((feature_Earth_hokkaido,feature_Earth_iwata))
rainfall_inventory_features=np.vstack((feature_Rain_fukuoka,feature_Rain_saka,feature_Rain_kumamoto))
test_inventory_features=feature_Earth_niigata
predict_probability_niigata=classify_inventory_cnn(earthquake_inventory_features,rainfall_inventory_features,test_inventory_features)

Visualization of the classification results

plot_geometric_results(predict_probability_niigata)

• Case 4: Kumamoto (Rainfall -triggerd) is a testing inventory

Training the algorithm on rest five inventories and test it on Kumamoto inventory.

earthquake_inventory_features=np.vstack((feature_Earth_hokkaido,feature_Earth_iwata,feature_Earth_niigata))
rainfall_inventory_features=np.vstack((feature_Rain_fukuoka,feature_Rain_saka))
test_inventory_features=feature_Rain_kumamoto
predict_probability_kumamoto=classify_inventory_cnn(earthquake_inventory_features,rainfall_inventory_features,test_inventory_features)

Visualization of the classification results

plot_geometric_results(predict_probability_kumamoto)

• Case 5: Fukuoka (Rainfall -triggerd) is a testing inventory

Training the algorithm on rest five inventories and test it on Kumamoto inventory.

earthquake_inventory_features=np.vstack((feature_Earth_hokkaido,feature_Earth_iwata,feature_Earth_niigata))
rainfall_inventory_features=np.vstack((feature_Rain_kumamoto,feature_Rain_saka))
test_inventory_features=feature_Rain_fukuoka
predict_probability_fukuoka=classify_inventory_cnn(earthquake_inventory_features,rainfall_inventory_features,test_inventory_features)

Visualization of the classification results

plot_geometric_results(predict_probability_fukuoka)

• Case 6: Saka (Rainfall -triggerd) is a testing inventory

Training the algorithm on rest five inventories and test it on Saka inventory.

earthquake_inventory_features=np.vstack((feature_Earth_hokkaido,feature_Earth_iwata,feature_Earth_niigata))
rainfall_inventory_features=np.vstack((feature_Rain_kumamoto,feature_Rain_fukuoka))
test_inventory_features=feature_Rain_saka
predict_probability_saka=classify_inventory_cnn(earthquake_inventory_features,rainfall_inventory_features,test_inventory_features)

Visualization of the classification results

plot_geometric_results(predict_probability_saka)