gistfile1.txt

#!/usr/bin/env python
# -*- coding: utf-8 -*-

import numpy as np
import matplotlib
import argparse
import imutils
import cv2
import numpy as np
import sys
import matplotlib.pyplot as plt


def mask_range(hsv, from_hue, to_hue):
    our_range = [np.array([from_hue, 40, 40]), np.array([to_hue, 255, 255])]
    mask = cv2.inRange(hsv, our_range[0], our_range[1])
    # return mask
    return cv2.morphologyEx(mask, cv2.MORPH_OPEN, np.ones((3, 3), np.uint8))
    # kernel = x
    # mask = cv2.erode(mask, kernel, iterations=1)


def cv_hsv_2_wavelen(x):
    return 625 - int(2 * x * 175. / 240.)


def count_hues(hsv2):
    unique, counts = np.unique(np.asarray(hsv2[..., 0]), return_counts=True)
    res = dict(zip(unique, counts))
    res[0] = 0
    return res


def add_cover_mask(enter_img, mask, height, width, color_scheme):
    img = enter_img.copy()
    roi = img[0:height, 0:width]
    mask_inv = cv2.bitwise_not(mask)
    color_img = np.zeros(img.shape, img.dtype)
    color_img[:, :] = color_scheme
    red_mask = cv2.bitwise_and(color_img, color_img, mask=mask)
    img[0:height, 0:width] = cv2.add(red_mask, cv2.bitwise_and(roi, roi, mask=mask_inv))
    return img


def get_trees(image):
    green = (16, 82)
    real_green = (20, 59)
    img = cv2.imread(image)

    width = 1500
    height = 1000

    # height, width, channels = img.shape

    # Resize
    img = cv2.resize(img, (width, height))
    cv2.imwrite(image.replace(".jpg", ".resized.jpg").replace(".JPG", ".resized.jpg"), img)

    # Noise removal
    dst = cv2.fastNlMeansDenoisingColored(img, None, 5, 5)
    cv2.imwrite(image.replace(".jpg", ".blurred1.jpg").replace(".JPG", ".blurred1.jpg"), dst)

    # dst = cv2.fastNlMeansDenoisingColored(img, None, 10, 10)
    # cv2.imwrite(image.replace(".jpg", ".blurred2.jpg").replace(".JPG", ".blurred2.jpg"), dst)

    # dst = cv2.filter2D(img, -1, np.array([[1, 2, 1], [2, 8, 2], [1, 2, 1]], np.float32) / 20)
    # cv2.imwrite(image.replace(".jpg", ".blurred3.jpg").replace(".JPG", ".blurred3.jpg"), dst)

    # dst = cv2.filter2D(img, -1, np.array([[1, 1, 1], [1, 1, 1], [1, 1, 1]], np.float32) / 9)
    # cv2.imwrite(image.replace(".jpg", ".blurred4.jpg").replace(".JPG", ".blurred4.jpg"), dst)

    img = dst
    # height, width, channels = img.shape

    hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
    mask_tree = mask_range(hsv, green[0], green[1])
    mask_healthy = mask_range(hsv, real_green[0], real_green[1])

    # Render extracted tree image
    extracted = add_cover_mask(img, cv2.bitwise_not(mask_tree), height, width, (255, 255, 255))
    cv2.imwrite(image.replace(".jpg", ".extracted-mask.jpg").replace(".JPG", ".extracted-mask.jpg"), mask_tree)
    cv2.imwrite(image.replace(".jpg", ".extracted.jpg").replace(".JPG", ".extracted.jpg"), extracted)

    # Render extracted healthy image
    extracted = add_cover_mask(extracted, cv2.bitwise_not(mask_healthy), height, width, (255, 255, 255))
    cv2.imwrite(image.replace(".jpg", ".extracted-g.jpg").replace(".JPG", ".extracted-g.jpg"), extracted)
    cv2.imwrite(image.replace(".jpg", ".extracted-g-mask.jpg").replace(".JPG", ".extracted-g-mask.jpg"), mask_healthy)


    img2 = np.invert(img*mask_tree[:,:,np.newaxis])

    # HSV paletter
    hsv2 = cv2.cvtColor(img2, cv2.COLOR_BGR2HSV)


    # cv2.imwrite(image.replace(".jpg", ".extracted.jpg").replace(".JPG", ".extracted.jpg"), img2)

    # Mask extraction
    img3 = np.invert(img*mask_healthy[:,:,np.newaxis])
    # hsv2 = cv2.cvtColor(img3, cv2.COLOR_BGR2HSV)
    cv2.imwrite(image.replace(".jpg", ".extracted-g.jpg").replace(".JPG", ".extracted-g.jpg"), img3)



    #ddddddddd

    # print img.shape

    # img4 = cv2.addWeighted(redMask, 1, img, 1, 0, img)

    mask_bad = cv2.bitwise_xor(mask_tree, mask_healthy)
    img4 = add_cover_mask(img, mask_bad, height, width, (0, 0, 255))
    cv2.imwrite(image.replace(".jpg", ".extracted-bad.jpg").replace(".JPG", ".extracted-bad.jpg"), img4)







    res = count_hues(hsv2)

    active_area = np.count_nonzero(mask_tree)
    green_area = np.count_nonzero(mask_healthy)


    fig = plt.figure()
    fig.suptitle(image, fontsize=14, fontweight='bold')

    green_percent = int(green_area  * 10000. / active_area) * 1. / 100
    if green_percent > 94:
        condition = "excellent";
    elif green_percent > 79 :
        condition = "good"
    elif green_percent > 62 :
        condition = "satisfactory"
    elif green_percent > 44 :
        condition = "poor"
    else :
        condition = "terminal"

    plt.title("(active area: " + str(int(active_area  * 10000. / np.size(mask_tree)) * 1. / 100) + "%, green: " + str(green_percent) + "%, condition: " + condition + ")", fontsize=11)
    ax = fig.add_subplot(111)
    ax.set_xlabel('Wavelength')
    ax.set_ylabel('% in active area')

    plt.bar(
        range(len(res)),
        np.asarray(res.values()) * 100. / active_area,
        align='center'
    )

    wavelengths_for_axis_legend = np.asarray(map(lambda x: x * 10, range(int(green[0] / 10), int(green[1] / 10))))

    plt.xticks(wavelengths_for_axis_legend, reversed(map(cv_hsv_2_wavelen, wavelengths_for_axis_legend)))
    plt.savefig(image.replace(".jpg", ".plot.png").replace(".JPG", ".plot.png"))


if __name__ == "__main__":
    get_trees(sys.argv[1])