morphOps.py

# 2.12 Lab 7 object detection: a node for observing erosion/dilation
# Jacob Guggenheim 2019
# Jerry Ng 2019, 2020

import numpy as np
import cv2  # OpenCV module
import time
from tkinter import *

import math

tk = Tk()
l_h = Scale(tk, from_ = 0, to = 255, label = 'Hue, lower', orient = HORIZONTAL)
l_h.pack()
u_h = Scale(tk, from_ = 0, to = 255, label = 'Hue, upper', orient = HORIZONTAL)
u_h.pack()
u_h.set(255)
l_s = Scale(tk, from_ = 0, to = 255, label = 'Saturation, lower', orient = HORIZONTAL)
l_s.pack()
u_s = Scale(tk, from_ = 0, to = 255, label = 'Saturation, upper', orient = HORIZONTAL)
u_s.pack()
u_s.set(255)
l_v = Scale(tk, from_ = 0, to = 255, label = 'Value, lower', orient = HORIZONTAL)
l_v.pack()
u_v = Scale(tk, from_ = 0, to = 255, label = 'Value, upper', orient = HORIZONTAL)
u_v.pack()
u_v.set(255)


def main():
    # Open up the webcam
    cap = cv2.VideoCapture(0)
    while True:
        tk.update()

        # Read from the camera frame by frame
        ret, cv_image = cap.read()
        # visualize it in a cv window
        cv2.imshow("Original_Image", cv_image)
        cv2.waitKey(3)

        ################ HSV THRESHOLDING ####################
        # convert to HSV
        hsv_image = cv2.cvtColor(cv_image, cv2.COLOR_BGR2HSV)

        # get threshold values
        lower_bound_HSV = np.array([l_h.get(), l_s.get(), l_v.get()])
        upper_bound_HSV = np.array([u_h.get(), u_s.get(), u_v.get()])

        # threshold
        mask_HSV = cv2.inRange(hsv_image, lower_bound_HSV, upper_bound_HSV)

        # display image
        cv2.imshow("HSV_Thresholding", mask_HSV)
        cv2.waitKey(3)

        # kernel for all morphological operations
        #TODO: Change size of kernel
        # Also, try changing the shape of the kernel (places 1's in certain locations). Try making a circle/line/etc.
        kernel = np.ones((7,7),np.uint8)

        # EXAMPLE OF A VERTICAL LINE:
        #kernel = np.array([[0, 1, 0, 0],\
        #                [0, 1, 0, 0 ],\
        #                [1, 1, 1, 1 ],\
        #                [0, 1, 0, 0]], dtype=np.uint8)
        
        #TODO: Change number of iterations to see the effect.
        num_iterations = 3
        ################ Erosion ####################
        # erode blobs
        erosion = cv2.erode(mask_HSV,kernel,iterations = num_iterations)

        # display image
        cv2.imshow("Erosion", erosion)
        cv2.waitKey(3)

        ################ Dilation ####################
        # dilate blobs
        dilation = cv2.dilate(mask_HSV,kernel,iterations = num_iterations)

        # display image
        cv2.imshow("Dilation", dilation)
        cv2.waitKey(3)

        ################ Opening ####################
        # good for removing noise. its an erosion (to get rid of noise) followed by a dilation (to get back the original blobs you wanted to keep)
        opening = cv2.morphologyEx(mask_HSV, cv2.MORPH_OPEN, kernel, iterations = num_iterations)

        # display image
        cv2.imshow("Opening - Get rid of noise", opening)
        cv2.waitKey(3)

        ################ Closing ####################
        # good for filling small holes in blobs. its a dilation (to fill the holes) followed by an erosion (to get the object back to the right size)
        closing = cv2.morphologyEx(mask_HSV, cv2.MORPH_CLOSE, kernel, iterations = num_iterations)

        # display image
        cv2.imshow("Closing - Fill in blobs", closing)
        cv2.waitKey(3)


if __name__=='__main__':
    main()