Color Detection¶

This project will add a color detection algorithm to the previous Computer Vision project.

[PDF]Color Cards

Note

The printed colors may have a slightly different hue from the Python color models due to printer toner differences, or the printed medium, such as a tan-colored paper. This can cause a less accurate color recognition.

Run the Code

Note

This project requires access to the Raspberry Pi desktop to view the footage taken by the camera module.
You can connect a screen to the PiCar-X or refer to the tutorial Remote Desktop to access it with VNC or XRDP.
Once inside the Raspberry Pi desktop, open Terminal and type the following command to run it, or just open and run it with a Python editor.

cd /home/pi/picar-x/example
sudo python3 color_detect.py

When the code is run, if PiCar-X captures a red object, it will frame it out. You can also change the 'red' in the code to another color for detection.

Code

import cv2
from picamera.array import PiRGBArray
from picamera import PiCamera
import numpy as np
import time

color_dict = {'red':[0,4],'orange':[5,18],'yellow':[22,37],'green':[42,85],'blue':[92,110],'purple':[115,165],'red_2':[165,180]}  #Here is the range of H in the HSV color space represented by the color

kernel_5 = np.ones((5,5),np.uint8) #Define a 5×5 convolution kernel with element values of all 1.

def color_detect(img,color_name):

    # The blue range will be different under different lighting conditions and can be adjusted flexibly.  H: chroma, S: saturation v: lightness
    resize_img = cv2.resize(img, (160,120), interpolation=cv2.INTER_LINEAR)  # In order to reduce the amount of calculation, the size of the picture is reduced to (160,120)
    hsv = cv2.cvtColor(resize_img, cv2.COLOR_BGR2HSV)              # Convert from BGR to HSV
    color_type = color_name

    mask = cv2.inRange(hsv,np.array([min(color_dict[color_type]), 60, 60]), np.array([max(color_dict[color_type]), 255, 255]) )           # inRange()：Make the ones between lower/upper white, and the rest black
    if color_type == 'red':
            mask_2 = cv2.inRange(hsv, (color_dict['red_2'][0],0,0), (color_dict['red_2'][1],255,255))
            mask = cv2.bitwise_or(mask, mask_2)

    morphologyEx_img = cv2.morphologyEx(mask, cv2.MORPH_OPEN, kernel_5,iterations=1)              # Perform an open operation on the image

    # Find the contour in morphologyEx_img, and the contours are arranged according to the area from small to large.
    _tuple = cv2.findContours(morphologyEx_img,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE)
    # compatible with opencv3.x and openc4.x
    if len(_tuple) == 3:
        _, contours, hierarchy = _tuple
    else:
        contours, hierarchy = _tuple

    color_area_num = len(contours) # Count the number of contours

    if color_area_num > 0:
        for i in contours:    # Traverse all contours
            x,y,w,h = cv2.boundingRect(i)      # Decompose the contour into the coordinates of the upper left corner and the width and height of the recognition object

            # Draw a rectangle on the image (picture, upper left corner coordinate, lower right corner coordinate, color, line width)
            if w >= 8 and h >= 8: # Because the picture is reduced to a quarter of the original size, if you want to draw a rectangle on the original picture to circle the target, you have to multiply x, y, w, h by 4.
                x = x * 4
                y = y * 4
                w = w * 4
                h = h * 4
                cv2.rectangle(img,(x,y),(x+w,y+h),(0,255,0),2)  # Draw a rectangular frame
                cv2.putText(img,color_type,(x,y), cv2.FONT_HERSHEY_SIMPLEX, 1,(0,0,255),2)# Add character description

    return img,mask,morphologyEx_img

with PiCamera() as camera:
    print("start color detect")
    camera.resolution = (640,480)
    camera.framerate = 24
    rawCapture = PiRGBArray(camera, size=camera.resolution)
    time.sleep(2)

    for frame in camera.capture_continuous(rawCapture, format="bgr",use_video_port=True):# use_video_port=True
        img = frame.array
        img,img_2,img_3 =  color_detect(img,'red')  # Color detection function
        cv2.imshow("video", img)    # OpenCV image show
        cv2.imshow("mask", img_2)    # OpenCV image show
        cv2.imshow("morphologyEx_img", img_3)    # OpenCV image show
        rawCapture.truncate(0)   # Release cache

        k = cv2.waitKey(1) & 0xFF
        # 27 is the ESC key, which means that if you press the ESC key to exit
        if k == 27:
            break

    print('quit ...')
    cv2.destroyAllWindows()
    camera.close()

How it works?

First, the range of H in the HSV color space is defined as a dictionary, which is convenient for the following color judgment algorithm:

color_dict = {'red':[0,4],'orange':[5,18],'yellow':[22,37],'green':[42,85],'blue':[92,110],'purple':[115,165],'red_2':[165,180]}

Then, a convolution kernel of size 5x5 is defined, which will be used for morphological operations, like filtering.

kernel_5 = np.ones((5,5),np.uint8)

Next, the color_detect() function will processes pictures in four steps:

Extract the data of the target color as a new binary image (array).
Performs advanced morphological transformations.
Finds contours in a binary image.
Draws a frame for the recognized object on the image.

def color_detect(img,color_name):

    # The blue range will be different under different lighting conditions and can be adjusted flexibly.  H: chroma, S: saturation v: lightness
    resize_img = cv2.resize(img, (160,120), interpolation=cv2.INTER_LINEAR)  # In order to reduce the amount of calculation, the size of the picture is reduced to (160,120)
    hsv = cv2.cvtColor(resize_img, cv2.COLOR_BGR2HSV)              # Convert from BGR to HSV
    color_type = color_name

    mask = cv2.inRange(hsv,np.array([min(color_dict[color_type]), 60, 60]), np.array([max(color_dict[color_type]), 255, 255]) )           # inRange()：Make the ones between lower/upper white, and the rest black
    if color_type == 'red':
            mask_2 = cv2.inRange(hsv, (color_dict['red_2'][0],0,0), (color_dict['red_2'][1],255,255))
            mask = cv2.bitwise_or(mask, mask_2)

    morphologyEx_img = cv2.morphologyEx(mask, cv2.MORPH_OPEN, kernel_5,iterations=1)              # Perform an open operation on the image

    # Find the contour in morphologyEx_img, and the contours are arranged according to the area from small to large.
    _tuple = cv2.findContours(morphologyEx_img,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE)
    # compatible with opencv3.x and openc4.x
    if len(_tuple) == 3:
        _, contours, hierarchy = _tuple
    else:
        contours, hierarchy = _tuple

    color_area_num = len(contours) # Count the number of contours

    if color_area_num > 0:
        for i in contours:    # Traverse all contours
            x,y,w,h = cv2.boundingRect(i)      # Decompose the contour into the coordinates of the upper left corner and the width and height of the recognition object

            # Draw a rectangle on the image (picture, upper left corner coordinate, lower right corner coordinate, color, line width)
            if w >= 8 and h >= 8: # Because the picture is reduced to a quarter of the original size, if you want to draw a rectangle on the original picture to circle the target, you have to multiply x, y, w, h by 4.
                x = x * 4
                y = y * 4
                w = w * 4
                h = h * 4
                cv2.rectangle(img,(x,y),(x+w,y+h),(0,255,0),2)  # Draw a rectangular frame
                cv2.putText(img,color_type,(x,y), cv2.FONT_HERSHEY_SIMPLEX, 1,(0,0,255),2)# Add character description

    return img,mask,morphologyEx_img

The img , mask , and morphologyEx_img are displayed in three windows to directly observe the processing results of each step.

For more information on morphology and contouring, please reference the following resources: