#!/usr/bin/env python

import cv2

import pandas as pd

import numpy as np

import argparse,sys

import math

import glob

import subprocess

from openpyxl import Workbook, load_workbook

from json_tricks.np import dump, load

from utilities import Utilities

def filter_rawMatches(kp1, kp2, matches, ratio = 0.75):

return p1,p2,kp_pairs

def rankingList(index,image_id,n_inliers,percent):

resList[index][3] = percent

if __name__ == '__main__':

parser = argparse.ArgumentParser()

parser.add_argument('-img1', action='append', dest='im1',

help='Query Image')

parser.add_argument('-nF', action='store', dest='nFeatures', type=int,

default=0,help='Number of Features to retain <0-5000>')

parser.add_argument('-nL', action='store', dest='nOctaveLayers', type=int,

default=3,help='Number of Octave Layers <3-6>')

parser.add_argument('-cT', action='store',dest='contrastThres', type=float,

default=0.08,help='Contrast Threshold weak feature filter factor')

parser.add_argument('-eT', action='store',dest='edgeThres', type=int,

default=10,help='Edge Threshold edge-like feature factor')

parser.add_argument('-kpfixed', action='store_true',default=False,

dest='kpfixed',help='Fixed Keypoint Colour')

parser.add_argument('-v', action='store_true',default=False,

dest='v',help='Save image to file')

parser.add_argument('-o', action='store_true',default=False,

dest='o',help='Save data to CSV')

arguments = parser.parse_args()

if arguments.im1:

img1Path = str(arguments.im1)[2:-2]

else:

parser.print_help()

print("-img1: Query Image")

sys.exit(1)

util = Utilities()

#image counter

n = 0

## Prepare Dataset ##

dataset = []

listImages = glob.glob('dataset/*.jpg')

for i in listImages:

dataset.append(i.split('/')[-1])

#creating a list of (<image>,#inliers) pairs

resList = np.zeros( len(dataset) , [('idx', 'int16'), ('imageId', 'a28'), ('inliers', 'int16'), ('percent', 'float') ])

print("\n================")

print("Features", arguments.nFeatures)

print("Octave", arguments.nOctaveLayers)

print("Contrast Thres", arguments.contrastThres)

print("Edge Threshold", arguments.edgeThres)

print("================")

## SIFT features and descriptor

sift = cv2.xfeatures2d.SIFT_create(arguments.nFeatures,arguments.nOctaveLayers,arguments.contrastThres,arguments.edgeThres,1.6)

## #----------------- # ##

## Read, Resize, Grayscale Query Image ##

img1 = cv2.resize(cv2.imread(img1Path, 1), (480, 640))

img1Gray = cv2.cvtColor(img1,cv2.COLOR_RGB2GRAY )

kp1, d1 = sift.detectAndCompute(img1Gray, None)

for img2Path in dataset:

print("\nProcessing..")

print("Test Image:%s (%d/%d) \n" % (img2Path,n+1,len(dataset)))

## #----------------- # ##

## Read, Resize, Grayscale Test Image ##

img2 = cv2.resize(cv2.imread("dataset/" + img2Path, 1), (480, 640))

img2Gray = cv2.cvtColor(img2,cv2.COLOR_RGB2GRAY )

kp2, d2 = sift.detectAndCompute(img2Gray, None)

## # Use BFMatcher, Euclidian distance, Eliminate Multiples # ##

bf = cv2.BFMatcher(cv2.NORM_L2,crossCheck=True)

raw_matches = bf.match(d1,d2)

src_points, dst_points, kp_pairs = filter_rawMatches(kp1,kp2,raw_matches)

print('Matching tentative points in image1: %d, image2: %d' % (len(src_points), len(dst_points)))

## # ----------------# ##

## # Homography # ##

print('#----------------#')

print('Homography')

print('#----------------#')

if len(kp_pairs) > 4:

Homography, status = cv2.findHomography(src_points, dst_points, cv2.RANSAC, 5.0)

inliers = np.count_nonzero(status)

percent = float(inliers) / len(kp_pairs)

print("# Inliers %d out of %d tentative pairs" % (inliers,len(kp_pairs)))

print('Score:' + '{percent:.2%}\n'.format(percent= percent))

rankingList(n,img2Path,inliers,percent)

else:

rankingList(n,img2Path,0,0)

print("Not enough correspondenses")

n = n+1

## Verbose Results

if arguments.v:

img1kp = img1

img2kp = img2

if arguments.kpfixed:

img1kp = util.drawKeypoint(img1kp,kp1)

img2kp = util.drawKeypoint(img2kp,kp2)

else:

cv2.drawKeypoints(img1kp,kp1,img1kp,None,flags=cv2.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS)

cv2.drawKeypoints(img2kp,kp2,img2kp,None,flags=cv2.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS)

print("Default Keypoints")

cv2.imshow('Query',img1kp)

cv2.imshow('Test',img2kp)

imgTentMatches = cv2.drawMatches(img1kp,kp1,img2kp,kp2,raw_matches,None, flags=2)

cv2.imshow('Tentative Matches',imgTentMatches)

try:

h1, w1, z1 = img1.shape[:3]

h2, w2, z2 = img2.shape[:3]

img3 = np.zeros((max(h1, h2), w1+w2,z1), np.uint8)

img3[:h1, :w1, :z1] = cv2.resize(cv2.imread(img1Path, 1), (480, 640))

img3[:h2, w1:w1+w2, :z2] = cv2.resize(cv2.imread("dataset/" + img2Path, 1), (480, 640))

p1 = np.int32([kpp[0].pt for kpp in kp_pairs])

p2 = np.int32([kpp[1].pt for kpp in kp_pairs]) + (w1, 0)

for (x1, y1), (x2, y2), inlier in zip(p1,p2, status):

if inlier:

cv2.circle(img3, (x1, y1), 2, (0,250,0), 5)

cv2.circle(img3, (x2, y2), 2, (0,250,0), 5)

cv2.line(img3, (x1, y1), (x2, y2), (255,100,0),2)

else:

cv2.line(img3, (x1-2, y1-2), (x1+2, y1+2), (0, 0, 255), 3)

cv2.line(img3, (x1-2, y1+2), (x1+2, y1-2), (0, 0, 255), 3)

cv2.line(img3, (x2-2, y2-2), (x2+2, y2+2), (0, 0, 255), 3)

cv2.line(img3, (x2-2, y2+2), (x2+2, y2-2), (0, 0, 255), 3)

except (RuntimeError, TypeError, NameError):

print("Not enough Inliers")

cv2.imshow('SIFT Match + Inliers',img3)

cv2.waitKey(0)

cv2.destroyAllWindows()

#Output CSV

if arguments.o:

outFolder = "sift_experiments/"

util.initWrite()

util.writeFile(kp2,d2,img1Path,img2Path,inliers,percent,len(kp2))

util.closeWrite(outFolder,img2Path,'sift')

print("\n#### Ranking ####")

rList = np.sort(resList, order= 'inliers')[::-1]

for bestPair in range(10):

print('#%d: %s -> Inliers: %d') % (bestPair + 1, rList[bestPair][1], rList[bestPair][2])

print('{percent:.2%}'.format(percent= rList[bestPair][3]))

## # Results and Experimental Values Logging # ##

if arguments.o:

subprocess.check_output(["sed -e '!d' sift_experiments/sift*.csv >> sift_experiments/sift_" + img1Path[8:-4] +"_merge.csv"], shell=True)

jList = rList.reshape((n,1))

with open('results.json','w') as resultFile:

dump({'Results': jList },resultFile)