Binarization

0. Import package

%matplotlib inline

import cv2
import collections

import openalea.phenomenal.image as phm_img
import openalea.phenomenal.data as phm_data
import openalea.phenomenal.display as phm_display
from openalea.phenotyping_data.fetch import fetch_all_data

1. Prerequisites

1. Load images

Here, we load all the image of the dataset to simply the toturial, commodly you load just the images you want process one after the other.

data_dir = fetch_all_data("plant_6/raw").parent
raw_images = phm_data.raw_images(data_dir)

Downloading file 'plant_6/raw/side/0.png' from 'https://raw.githubusercontent.com/openalea/phenotyping_data/main/data/plant_6/raw/side/0.png' to '/home/docs/.cache/phenotyping_data'.
Downloading file 'plant_6/raw/side/120.png' from 'https://raw.githubusercontent.com/openalea/phenotyping_data/main/data/plant_6/raw/side/120.png' to '/home/docs/.cache/phenotyping_data'.
Downloading file 'plant_6/raw/side/150.png' from 'https://raw.githubusercontent.com/openalea/phenotyping_data/main/data/plant_6/raw/side/150.png' to '/home/docs/.cache/phenotyping_data'.
Downloading file 'plant_6/raw/side/180.png' from 'https://raw.githubusercontent.com/openalea/phenotyping_data/main/data/plant_6/raw/side/180.png' to '/home/docs/.cache/phenotyping_data'.
Downloading file 'plant_6/raw/side/210.png' from 'https://raw.githubusercontent.com/openalea/phenotyping_data/main/data/plant_6/raw/side/210.png' to '/home/docs/.cache/phenotyping_data'.
Downloading file 'plant_6/raw/side/240.png' from 'https://raw.githubusercontent.com/openalea/phenotyping_data/main/data/plant_6/raw/side/240.png' to '/home/docs/.cache/phenotyping_data'.
Downloading file 'plant_6/raw/side/270.png' from 'https://raw.githubusercontent.com/openalea/phenotyping_data/main/data/plant_6/raw/side/270.png' to '/home/docs/.cache/phenotyping_data'.
Downloading file 'plant_6/raw/side/30.png' from 'https://raw.githubusercontent.com/openalea/phenotyping_data/main/data/plant_6/raw/side/30.png' to '/home/docs/.cache/phenotyping_data'.
Downloading file 'plant_6/raw/side/300.png' from 'https://raw.githubusercontent.com/openalea/phenotyping_data/main/data/plant_6/raw/side/300.png' to '/home/docs/.cache/phenotyping_data'.
Downloading file 'plant_6/raw/side/330.png' from 'https://raw.githubusercontent.com/openalea/phenotyping_data/main/data/plant_6/raw/side/330.png' to '/home/docs/.cache/phenotyping_data'.
Downloading file 'plant_6/raw/side/60.png' from 'https://raw.githubusercontent.com/openalea/phenotyping_data/main/data/plant_6/raw/side/60.png' to '/home/docs/.cache/phenotyping_data'.
Downloading file 'plant_6/raw/side/90.png' from 'https://raw.githubusercontent.com/openalea/phenotyping_data/main/data/plant_6/raw/side/90.png' to '/home/docs/.cache/phenotyping_data'.
Downloading file 'plant_6/raw/top/0.png' from 'https://raw.githubusercontent.com/openalea/phenotyping_data/main/data/plant_6/raw/top/0.png' to '/home/docs/.cache/phenotyping_data'.

1.3. Display images

You can view each image according to this date and angle of view like this : Note : Angle top view is represented by negative number (-1)

phm_display.show_images(
    [raw_images["side"][120], raw_images["side"][30], raw_images["top"][0]]
)

2. Binarization

2.1. Define a binarization routines

masks=fetch_all_data("plant_6/mask")

Downloading file 'plant_6/mask/mask_hsv.png' from 'https://raw.githubusercontent.com/openalea/phenotyping_data/main/data/plant_6/mask/mask_hsv.png' to '/home/docs/.cache/phenotyping_data'.
Downloading file 'plant_6/mask/mask_mean_shift.png' from 'https://raw.githubusercontent.com/openalea/phenotyping_data/main/data/plant_6/mask/mask_mean_shift.png' to '/home/docs/.cache/phenotyping_data'.

def routine_side_binarization(image, mean_img):
    maks = phm_data.tutorial_data_binarization_mask(data_dir)

    threshold = 0.3
    dark_background = False

    hsv_min = (30, 11, 0)
    hsv_max = (129, 254, 141)

    # Convert image on HSV representation
    hsv_image = cv2.cvtColor(image, cv2.COLOR_BGR2HSV)
    # Threshold the image with HSV min and max value
    binary_hsv_image = phm_img.threshold_hsv(hsv_image, hsv_min, hsv_max, maks[0])

    # Threshold the image with difference between image and mean_image
    binary_mean_shift_image = phm_img.threshold_meanshift(
        image, mean_img, threshold, dark_background, maks[1]
    )

    # Add the two image
    result = cv2.add(binary_hsv_image, binary_mean_shift_image)

    # Erode and dilate the image to remove possible noise
    result = cv2.medianBlur(result, 3)

    return result


def routine_top_binarization(image):
    hsv_min = (42, 75, 28)
    hsv_max = (80, 250, 134)
    median_blur_size = 9
    iterations = 5

    # Convert image on HSV representation
    hsv_image = cv2.cvtColor(image, cv2.COLOR_BGR2HSV)
    # Apply a median blur on the image
    hsv_image = cv2.medianBlur(hsv_image, ksize=median_blur_size)

    # Threshold the image with HSV min and max value
    bin_img = phm_img.threshold_hsv(hsv_image, hsv_min, hsv_max)
    # dilate and erode the image to remove possible noise
    bin_img = phm_img.dilate_erode(bin_img, kernel_shape=(3, 3), iterations=iterations)

    return bin_img

2.2. Binarize images

# Compute the mean image of the side view image

mean_img = phm_img.mean_image(list(raw_images["side"].values()))

routine_binarization = {
    "side": lambda im: routine_side_binarization(im, mean_img),
    "top": lambda im: routine_top_binarization(im),
}

bin_images = collections.defaultdict(dict)
for id_camera in raw_images:
    for angle in raw_images[id_camera]:
        bin_images[id_camera][angle] = routine_binarization[id_camera](
            raw_images[id_camera][angle]
        )

2.3. Display images binarize

id_camera, angle = "side", 120
phm_display.show_images(
    [raw_images[id_camera][angle], mean_img, bin_images[id_camera][angle]]
)

id_camera, angle = "top", 0
phm_display.show_images([raw_images[id_camera][angle], bin_images[id_camera][angle]])