# -*- coding: utf-8 -*-

"""

Spyder Editor

Date: 9-Dec-2015

Developer: Senthil Chidambaram

Ref:Geo Spatial Analysis, Joel

Scope:

Change detection using co-registered satellite images - 2 img on same location

Simple image difference techique

Logic:

[1] Check dimension,extent, projection for 2 images are same

[2] Convert image in to array and subtract the numpy array-> AfterImage - BeforeImage

[3] Get the difference and Classify/group the pixel

[4] color each class/group and mask insiginifcant changes

Note: This code works for 3 band Satellite images [Blue,Green,Red]

Since the input file is limited hardcoded some dimension [3,2727,4614] OR [3,705,1776]

Please change the dimension if you try with your own image

Code has to be refactored

"""

import gdal,gdalnumeric

import numpy as np

#Input satellite images - same extent,resolution,location

defaultChoice = True

#Selective area - Pipeline - dimension [3,705,1776]

if defaultChoice :

satImgBefore= "T:/Selective2010.tif"

satImgAfter="T:/Selective2011.tif"

else:

#Whole image -more computing -dimension [3,2727,4614]

satImgBefore="T:/Machine Learning/R/2010.tif" #"T:/before.tif"

satImgAfter= "T:/Machine Learning/R/2011.tif" #"T:/after.tif"

# convert pixel array with 3 layer info

arrayPipelineImg=gdalnumeric.LoadFile(satImgBefore).astype(np.int8)

arrayNoPipelineImg=gdalnumeric.LoadFile(satImgAfter).astype(np.int8)

#since 2 image has differenct dimension , take common rows and col

#diff=arrayImg2[0:704,0:1775]-arrayImg1[0:704,1:1775]

# image difference betwee img2-img1

#AfterImage - before image

#below commented code to check are we getiig black [0,0,0] image since no difference

#diff=arrayPipelineImgPreProcessed- arrayPipelineImgPreProcessed

if (arrayPipelineImg.shape ==arrayNoPipelineImg.shape):

diff=arrayNoPipelineImg-arrayPipelineImg

else:

#below code works for the entire satelite image with the dimension [3,2727,4614]

#arrayNoPipelineImgPreProcessed= arrayNoPipelineImg[:3,:2727,:4614]

#arrayPipelineImgPreProcessed= arrayPipelineImg[:3,:2727,:4614]

arrayPipelineImgPreProcessed= arrayPipelineImg[:3,:705,:1776]

diff=arrayNoPipelineImg-arrayPipelineImgPreProcessed

print (diff.shape)

#Before image - after image to see the diff

#diff=arrayPipelineImgPreProcessed - arrayNoPipelineImg

# classify/group the pixels into 5 group pixels -128 to 127

pixelgroups=np.histogram(diff,bins=4)[1]

#print ("class :" , (classes))

# Color Table to color each class and mask insigificant changes as just black B,G,R

lookup=[ [0,0,0],[0,0,0],[0,0,0],[0,0,0],[0,255,0],[0,0,255]]

#tried different color table to see the difference output

#lookup=[ [0,0,0],[0,255,0],[0,0,255],[0,0,0],[0,255,0],[0,0,255]]

start=1

#print ("diff shape", diff.shape[0])

#setup output imag array with size same as diff - 3L,mRows,nCol and all values set to "0"(Black color)

rgbDifference=np.zeros((diff.shape[0],diff.shape[1],diff.shape[2]), np.int8)

print (rgbDifference.shape)

#process all classes and assign colors

#

for i in range(len(pixelgroups)):

mask= np.logical_and(start <=diff, diff <=pixelgroups[i])

for j in range(len(lookup[i])):

rgbDifference = gdalnumeric.numpy.choose( mask, (rgbDifference[j], lookup[i][j]))

#rgbDifference = gdalnumeric.numpy.choose( mask, (rgbDifference[j], 255))-FULL WHITE IMAGE

start = pixelgroups[i]+1

gdalnumeric.SaveArray(rgbDifference, "T:/ChangeDetection.tif", format="GTiff", prototype=satImgAfter)