BardisRenos / Object_Detection

Object_Detection

In this repo will demonstrate the ability a machine to detect different objects and in different position. This repo will shows the steps that need to follow in order to achieve that.

Reading the image and showing

As the first step. Read the image using the opencv module.

  def read_image(path):
      return cv2.imread(path)

Showing the image that I will use

  def show_image_with_opencv(image):
      cv2.imshow("Given Image", image)
      cv2.waitKey(0)
      cv2.destroyAllWindows()

Showing the stages of the images

During the image preprocessing the image passed through different stages.

  def image_preprocessing(self, given_image):
      self.convert_to_gray_scale(given_image)
      self.pure_image = given_image

      _, self.image_threshold_bw = cv2.threshold(self.image_gray_scale, 190, 255, cv2.THRESH_BINARY)
      kernel = np.ones((3, 3), np.uint8)
      self.image_morph = cv2.morphologyEx(self.image_threshold_bw, cv2.MORPH_CLOSE, kernel, iterations=9)
      self.image_mask = 255 - self.image_morph
      self.background = cv2.dilate(self.image_mask, kernel, iterations=5)
      dist_transform = cv2.distanceTransform(self.image_mask, cv2.DIST_L2, 3)
      _, self.foreground = cv2.threshold(dist_transform, 0.237 * dist_transform.max(), 255, 0)
      self.the_unknown_image = self.background - self.foreground
      self.foreground = np.uint8(self.foreground)

This piece of code plots all 6 images in order to show the stages that an image pass.

  def plot_images_stages(self):
      show_images_stages(self.pure_image, self.image_gray_scale, self.image_threshold_bw,
                         self.image_morph, self.image_mask, self.background,
                         "The input image", "Image to Gray", "Threshold Image", "Image Morph",
                         "Image with the Mask", "Background Image")

Creating new borders

From the previous chapter the image has a small thing to care of. Some pills are touching the border and as a result the pills are not well separeted and create a solid body with pills and the background.

  def creating_boundary_image(self, given_image):
      self.image_copy = given_image.copy()
      h, w = self.image_copy.shape[:2]
      cv2.circle(self.image_copy, ((w // 2) + 1, (h // 2) - 5), 622, (255, 255, 255), 350)

Here is the image and creating the white boundary to separte then pills with the rest of the image

After the applying the boundary to the image

The images below show the preprocessed images after putting the boundary.

Markers

def markers_creation(self):
    _, self.image_markers = cv2.connectedComponents(self.foreground, connectivity=8)
    self.image_markers = self.image_markers + 10
    self.image_markers[self.the_unknown_image == 255] = 0

Watershed

def watershed(self):
    self.image_markers = cv2.watershed(self.pure_image, self.image_markers)
    self.pure_image[self.image_markers == -1] = [0, 255, 0]
    self.image_label2rgb = color.label2rgb(self.image_markers, bg_label=0)

Showing the Image after passing a number of preprocessing steps

The below image shows that, the image passed through.

Showing the last result of the image preprocessing

As can been shown each pill (object) has a thin green line around. The distinction of each pill is clear by looking the the green line.