Merge branch 'master' of https://www.github.com/arielcatli/automated-…

…rice-grader

python2/capture.py

@@ -2,6 +2,7 @@
 import cv2
 import os
 from extraction import extract
+from extract_grains import extract_grains
 import Tkinter as tk
 import pickle
 import numpy as np
@@ -72,6 +73,7 @@ def __start_capture(self):
         camera.capture(current)
         self.whole_sample += 1
         camera.close()
+#        self.__extr = extract(current, self.sample_directory_extracted_e, self.sample_directory_extracted_p, self.__extracted_count)
         self.__extr = extract(current, self.sample_directory_extracted_e, self.sample_directory_extracted_p, self.__extracted_count)
         self.__grain_list_masked.append(self.__extr[1])
         self.__grain_list_platform.append(self.__extr[2])

python2/extract_grains.py

@@ -0,0 +1,91 @@
+# -*- coding: utf-8 -*-
+"""
+Created on Sat Apr 14 15:30:34 2018
+
+@author: Ariel Domingo Catli
+"""
+
+import cv2
+import numpy as np
+import os
+from math import ceil
+
+
+def extract_grains(sample_image_directory, sample_directory_extracted_e, sample_directory_extracted_p, start):
+    original_image = cv2.imread(sample_image_directory)
+    image = cv2.imread(sample_image_directory, 0)
+    image = cv2.medianBlur(image, 7)
+#    image = cv2.bilateralFilter(image, 20, 31, 21)
+    image = cv2.threshold(image, 0, 255, cv2.THRESH_BINARY+cv2.THRESH_OTSU)[1]
+#   image = cv2.Canny(image[1], 128, 255)
+
+    masked_image = cv2.bitwise_and(original_image, original_image, mask = image)
+#    cv2.imshow("Original", original_image)
+#    cv2.imshow("masked", image)
+#    __clean_windows()
+    contours = cv2.findContours(image.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
+    contours = np.array(contours)
+    lists = __bound(masked_image, contours[1], sample_directory_extracted_e, sample_directory_extracted_p, start)
+
+    return[len(contours[1]), lists[0], lists[1]]
+
+
+def __bound(image, contours, sample_directory_extracted_e, sample_directory_extracted_p, start):
+    images = []
+    for i,item in enumerate(contours):
+        if not (len(item)<3):
+            rectangle = cv2.minAreaRect(item)
+            box = cv2.boxPoints(rectangle)
+            box = np.int0(box)
+            rotation_matrix = cv2.getRotationMatrix2D(rectangle[0], rectangle[2], 1)
+            rotated_image = cv2.warpAffine(image.copy(), rotation_matrix, (image.shape[1], image.shape[0]))
+
+            #new_rectangle = (rectangle[0], rectangle[1], 0.0)
+            box = cv2.boxPoints(rectangle)
+            new_box = np.int0(cv2.transform(np.array([box]), rotation_matrix))[0]
+
+            cropped = rotated_image[new_box[1][1] : new_box[0][1], new_box[1][0] : new_box[2][0]]
+
+            if cropped.shape[0] < cropped.shape[1]:
+#                center = (int(cropped.shape[1]/2), int(cropped.shape[0]/2))
+#                rotation_matrix_cropped = cv2.getRotationMatrix2D(center, -90, 1)
+#                cropped = cv2.warpAffine(cropped, rotation_matrix_cropped, (cropped.shape[0], cropped.shape[1]))
+                cropped = np.rot90(cropped)
+
+            images.append(cropped)
+
+    lists = __write_images(images, sample_directory_extracted_e, sample_directory_extracted_p, start)
+    return lists
+
+def __clean_windows():
+    cv2.waitKey(0)
+    cv2.destroyAllWindows()
+
+def __write_images(images, sample_directory_extracted_e, sample_directory_extracted_p, start):
+    grain_list_masked = []
+    grain_list_platform = []
+    for i,image in enumerate(images):
+        if os.path.exists(sample_directory_extracted_e):
+            cv2.imwrite(sample_directory_extracted_e + "e" + str(i) + ".jpg", image)
+            grain_list_masked.append(sample_directory_extracted_e + "e" + str(i) + ".jpg")
+        else:
+            print("The directory: " + sample_directory_extracted_e + " does not exists.")
+
+        if os.path.exists(sample_directory_extracted_p):
+            if image.shape[0] < 128 and image.shape[1] <64:
+                platform = np.zeros([128,64,3], np.uint8)
+                vertical_offset = int(ceil((platform.shape[0] - image.shape[0])/2))
+                horizontal_offset = int(ceil((platform.shape[1] - image.shape[1])/2))
+                platform[vertical_offset:(vertical_offset+image.shape[0]), horizontal_offset:(horizontal_offset+image.shape[1])] = image
+
+                cv2.imwrite(sample_directory_extracted_p + "p" + str(i) + ".jpg", platform)
+                grain_list_platform.append(sample_directory_extracted_p + "p" + str(i) + ".jpg")
+        else:
+            print("The directory: " + sample_directory_extracted_p + " does not exists.")
+
+    return [grain_list_masked, grain_list_platform]
+
+
+if __name__ == "__main__":
+    print(extract_grains("grain.jpg", "e/", "p/", 0))
+

python3/utils/extract_grains.py

@@ -0,0 +1,91 @@
+# -*- coding: utf-8 -*-
+"""
+Created on Sat Apr 14 15:30:34 2018
+
+@author: Ariel Domingo Catli
+"""
+
+import cv2
+import numpy as np
+import os
+from math import ceil
+
+
+def extract_grains(sample_image_directory, sample_directory_extracted_e, sample_directory_extracted_p, start):
+    original_image = cv2.imread(sample_image_directory)
+    image = cv2.imread(sample_image_directory, 0)
+    image = cv2.medianBlur(image, 7)
+#    image = cv2.bilateralFilter(image, 20, 31, 21)
+    image = cv2.threshold(image, 0, 255, cv2.THRESH_BINARY+cv2.THRESH_OTSU)[1]
+#   image = cv2.Canny(image[1], 128, 255)
+
+    masked_image = cv2.bitwise_and(original_image, original_image, mask = image)
+#    cv2.imshow("Original", original_image)
+#    cv2.imshow("masked", image)
+#    __clean_windows()
+    contours = cv2.findContours(image.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
+    contours = np.array(contours)
+    lists = __bound(masked_image, contours[1], sample_directory_extracted_e, sample_directory_extracted_p, start)
+
+    return[len(contours[1]), lists[0], lists[1]]
+
+
+def __bound(image, contours, sample_directory_extracted_e, sample_directory_extracted_p, start):
+    images = []
+    for i,item in enumerate(contours):
+        if not (len(item)<3):
+            rectangle = cv2.minAreaRect(item)
+            box = cv2.boxPoints(rectangle)
+            box = np.int0(box)
+            rotation_matrix = cv2.getRotationMatrix2D(rectangle[0], rectangle[2], 1)
+            rotated_image = cv2.warpAffine(image.copy(), rotation_matrix, (image.shape[1], image.shape[0]))
+
+            #new_rectangle = (rectangle[0], rectangle[1], 0.0)
+            box = cv2.boxPoints(rectangle)
+            new_box = np.int0(cv2.transform(np.array([box]), rotation_matrix))[0]
+
+            cropped = rotated_image[new_box[1][1] : new_box[0][1], new_box[1][0] : new_box[2][0]]
+
+            if cropped.shape[0] < cropped.shape[1]:
+#                center = (int(cropped.shape[1]/2), int(cropped.shape[0]/2))
+#                rotation_matrix_cropped = cv2.getRotationMatrix2D(center, -90, 1)
+#                cropped = cv2.warpAffine(cropped, rotation_matrix_cropped, (cropped.shape[0], cropped.shape[1]))
+                cropped = np.rot90(cropped)
+
+            images.append(cropped)
+
+    lists = __write_images(images, sample_directory_extracted_e, sample_directory_extracted_p, start)
+    return lists
+
+def __clean_windows():
+    cv2.waitKey(0)
+    cv2.destroyAllWindows()
+
+def __write_images(images, sample_directory_extracted_e, sample_directory_extracted_p, start):
+    grain_list_masked = []
+    grain_list_platform = []
+    for i,image in enumerate(images):
+        if os.path.exists(sample_directory_extracted_e):
+            cv2.imwrite(sample_directory_extracted_e + "e" + str(i) + ".jpg", image)
+            grain_list_masked.append(sample_directory_extracted_e + "e" + str(i) + ".jpg")
+        else:
+            print("The directory: " + sample_directory_extracted_e + " does not exists.")
+
+        if os.path.exists(sample_directory_extracted_p):
+            if image.shape[0] < 128 and image.shape[1] <64:
+                platform = np.zeros([128,64,3], np.uint8)
+                vertical_offset = int(ceil((platform.shape[0] - image.shape[0])/2))
+                horizontal_offset = int(ceil((platform.shape[1] - image.shape[1])/2))
+                platform[vertical_offset:(vertical_offset+image.shape[0]), horizontal_offset:(horizontal_offset+image.shape[1])] = image
+
+                cv2.imwrite(sample_directory_extracted_p + "p" + str(i) + ".jpg", platform)
+                grain_list_platform.append(sample_directory_extracted_p + "p" + str(i) + ".jpg")
+        else:
+            print("The directory: " + sample_directory_extracted_p + " does not exists.")
+
+    return [grain_list_masked, grain_list_platform]
+
+
+if __name__ == "__main__":
+    print(extract_grains("grain.jpg", "e/", "p/", 0))
+