How to Perform Camera Calibration Using OpenCV

In this tutorial, we will write a program to correct camera lens distortion using OpenCV. This process is known as camera calibration.

7_chessboard_input1_drawn_corners

Real-World Applications

Prerequisites

Install OpenCV

The first thing you need to do is to make sure you have OpenCV installed on your machine. If you are using Anaconda, you can type:

conda install -c conda-forge opencv

Alternatively, you can type:

pip install opencv-python

What Is Camera Calibration and Why Is It Important?

A camera’s job is to convert light that hits its image sensor into an image. An image is made up of picture elements known as pixels

In a perfect world, these pixels (i.e. what the camera sees) would look exactly like what you see in the real world. However, no camera is perfect. They all contain physical flaws that can, in some cases, cause significant distortion.

For example, look at an extreme example of image distortion below. 

1_before_distorted_houseJPG

Here is the same image after calibrating the camera and correcting for the distortion.

2_after_distorted_houseJPG

Do you see how this could lead to problems…especially in robotics where accuracy key?

Imagine you have a robotic arm that is working inside a warehouse. Its job is to pick up items off a conveyor belt and place those items into a bin. 

scara-robot-packaging-cookies

To help the robot determine the exact position of the items, there is a camera overhead. To get an accurate translation between camera pixel coordinates and real world coordinates, you need to calibrate the camera to remove any distortion.

The two most important forms of distortion are radial distortion and tangential distortion.

Radial distortion occurs when straight lines appear curved. Here is an example below of a type of radial distortion known as barrel distortion. Notice the bulging in the image that is making straight lines (in the real world) appear curved.

600px-Barrel_distortion

Tangential distortion occurs when the camera lens is not exactly aligned parallel to the camera sensor. Tangential distortion can make the real world look stretched or tilted. It can also make items appear closer than they are in real life.

To get an accurate representation of the real world on a camera, we have to calibrate the camera. We want to know that when we see an object of a certain size in the real world, this size translates to a known size in camera pixel coordinates. 

By calibrating a camera, we can enable a robotic arm, for example, to have better “hand-eye” coordination. You also enable a self-driving car to know the location of pedestrians, dogs, and other objects relative to the car. 

Calibrating the camera is the process of using a known real-world pattern (e.g. a chessboard) to estimate the extrinsic parameters (rotation and translation vectors) and intrinsic parameters (e.g. focal length, optical center, etc.) of a camera’s lens and image sensor to reduce distortion error caused by the camera’s imperfections. 

These parameters include:

  • Focal length 
  • Image (i.e. optical) center (it is usually not exactly at (width/2, height/2))
  • Scaling factors for the pixels along the rows and columns
  • Skew factor
  • Lens distortion

Why Use a Chessboard?

3-chessboard

Chessboard calibration is a standard technique for performing camera calibration and estimating the values of the unknown parameters I mentioned in the previous section. 

OpenCV has a chessboard calibration library that attempts to map points in 3D on a real-world chessboard to 2D camera coordinates. This information is then used to correct distortion. 

Note that any object could have been used (a book, a laptop computer, a car, etc.), but a chessboard has unique characteristics that make it well-suited for the job of correcting camera distortions:

  • It is flat, so you don’t need to deal with the z-axis (z=0), only the x and y-axis. All the points on the chessboard lie on the same plane.
  • There are clear corners and points, making it easy to map points in the 3D real world coordinate system to points on the camera’s 2D pixel coordinate system.
  • The points and corners all occur on straight lines.

Perform Camera Calibration Using OpenCV

The official tutorial from OpenCV is here on their website, but let’s go through the process of camera calibration slowly, step by step. 

Print a Chessboard

The first step is to get a chessboard and print it out on regular A4 size paper. You can download this pdf, which is the official chessboard from the OpenCV documentation, and just print that out. 

Measure Square Length

Measure the length of the side of one of the squares. In my case, I measured 2.3 cm (0.023 m).

Take Photos of the Chessboard From Different Distances and Orientations

We need to take at least 10 photos of the chessboard from different distances and orientations. I’ll take 19 photos so that my algorithm will have a lot of input images from which to perform camera calibration.

Tape the chessboard to a flat, solid object.

Take the photos, and then move them to a directory on your computer.

Here are examples of some of the photos I took: 

6-chessboard-calibration-angles

Draw the Corners

The first thing we need to do is to find and then draw the corners on the image.

Make sure you have NumPy installed, a scientific computing library for Python.

If you’re using Anaconda, you can type:

conda install numpy

Alternatively, you can type:

pip install numpy

Write the following code in Python. You can copy and paste this code into your favorite IDE. Pick one of the chessboard images as a test case. I’ll name the file draw_corners.py. This code will draw the corners on an input chessboard image and then save the drawing to an image file.

import cv2 # Import the OpenCV library to enable computer vision
import numpy as np # Import the NumPy scientific computing library

# Author: Addison Sears-Collins
# https://automaticaddison.com
# Description: Detect corners on a chessboard

filename = 'chessboard_input1.jpg'

# Chessboard dimensions
number_of_squares_X = 10 # Number of chessboard squares along the x-axis
number_of_squares_Y = 7  # Number of chessboard squares along the y-axis
nX = number_of_squares_X - 1 # Number of interior corners along x-axis
nY = number_of_squares_Y - 1 # Number of interior corners along y-axis

def main():
	
  # Load an image
  image = cv2.imread(filename)

  # Convert the image to grayscale
  gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)  

  # Find the corners on the chessboard
  success, corners = cv2.findChessboardCorners(gray, (nY, nX), None)
	
  # If the corners are found by the algorithm, draw them
  if success == True:

    # Draw the corners
    cv2.drawChessboardCorners(image, (nY, nX), corners, success)

    # Create the output file name by removing the '.jpg' part
    size = len(filename)
    new_filename = filename[:size - 4]
    new_filename = new_filename + '_drawn_corners.jpg'		
    
    # Save the new image in the working directory
    cv2.imwrite(new_filename, image)

    # Display the image 
    cv2.imshow("Image", image) 
	
    # Display the window until any key is pressed
    cv2.waitKey(0) 
	
    # Close all windows
    cv2.destroyAllWindows() 
	
main()

Run the code.

Here is the output:

7_chessboard_input1_drawn_corners-1

Write the Python Code for Camera Calibration

Now that we know how to identify corners on a chessboard, let’s write the code to perform camera calibration.

Here is the code. I put my distorted image inside a folder named ‘distorted’ inside the working directory. This folder is where the output undistorted saves to.

Don’t be scared at how long the code is. I put a lot of comments in here in order to make each line easier for you to understand. Just copy and paste the code into your favorite text editor or IDE. I named the file camera_calibration.py.

# Author: Addison Sears-Collins
# https://automaticaddison.com
# Description: Perform camera calibration using a chessboard.

import cv2 # Import the OpenCV library to enable computer vision
import numpy as np # Import the NumPy scientific computing library
import glob # Used to get retrieve files that have a specified pattern

# Path to the image that you want to undistort
distorted_img_filename = 'distorted/chessboard_input12.jpg'

# Chessboard dimensions
number_of_squares_X = 10 # Number of chessboard squares along the x-axis
number_of_squares_Y = 7  # Number of chessboard squares along the y-axis
nX = number_of_squares_X - 1 # Number of interior corners along x-axis
nY = number_of_squares_Y - 1 # Number of interior corners along y-axis
square_size = 0.023 # Length of the side of a square in meters

# Store vectors of 3D points for all chessboard images (world coordinate frame)
object_points = []

# Store vectors of 2D points for all chessboard images (camera coordinate frame)
image_points = []

# Set termination criteria. We stop either when an accuracy is reached or when
# we have finished a certain number of iterations.
criteria = (cv2.TERM_CRITERIA_EPS + cv2.TERM_CRITERIA_MAX_ITER, 30, 0.001)

# Define real world coordinates for points in the 3D coordinate frame
# Object points are (0,0,0), (1,0,0), (2,0,0) ...., (5,8,0)
object_points_3D = np.zeros((nX * nY, 3), np.float32) 		

# These are the x and y coordinates												 
object_points_3D[:,:2] = np.mgrid[0:nY, 0:nX].T.reshape(-1, 2) 

object_points_3D = object_points_3D * square_size

def main():
	
  # Get the file path for images in the current directory
  images = glob.glob('*.jpg')
	
  # Go through each chessboard image, one by one
  for image_file in images:
 
    # Load the image
    image = cv2.imread(image_file)  

    # Convert the image to grayscale
    gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)  

    # Find the corners on the chessboard
    success, corners = cv2.findChessboardCorners(gray, (nY, nX), None)
	
    # If the corners are found by the algorithm, draw them
    if success == True:

      # Append object points
      object_points.append(object_points_3D)

      # Find more exact corner pixels		
      corners_2 = cv2.cornerSubPix(gray, corners, (11,11), (-1,-1), criteria)		
      
	  # Append image points
      image_points.append(corners_2)

      # Draw the corners
      cv2.drawChessboardCorners(image, (nY, nX), corners_2, success)

      # Display the image. Used for testing.
      #cv2.imshow("Image", image) 
	
      # Display the window for a short period. Used for testing.
      #cv2.waitKey(200) 
																													
  # Now take a distorted image and undistort it 
  distorted_image = cv2.imread(distorted_img_filename)

  # Perform camera calibration to return the camera matrix, distortion coefficients, rotation and translation vectors etc 
  ret, mtx, dist, rvecs, tvecs = cv2.calibrateCamera(object_points, 
                                                    image_points, 
                                                    gray.shape[::-1], 
                                                    None, 
                                                    None)

  # Get the dimensions of the image	
  height, width = distorted_image.shape[:2]
	
  # Refine camera matrix
  # Returns optimal camera matrix and a rectangular region of interest
  optimal_camera_matrix, roi = cv2.getOptimalNewCameraMatrix(mtx, dist, 
                                                            (width,height), 
                                                            1, 
                                                            (width,height))

  # Undistort the image
  undistorted_image = cv2.undistort(distorted_image, mtx, dist, None, 
                                    optimal_camera_matrix)
  
  # Crop the image. Uncomment these two lines to remove black lines
  # on the edge of the undistorted image.
  #x, y, w, h = roi
  #undistorted_image = undistorted_image[y:y+h, x:x+w]
	
  # Display key parameter outputs of the camera calibration process
  print("Optimal Camera matrix:") 
  print(optimal_camera_matrix) 

  print("\n Distortion coefficient:") 
  print(dist) 
  
  print("\n Rotation Vectors:") 
  print(rvecs) 
  
  print("\n Translation Vectors:") 
  print(tvecs) 

  # Create the output file name by removing the '.jpg' part
  size = len(distorted_img_filename)
  new_filename = distorted_img_filename[:size - 4]
  new_filename = new_filename + '_undistorted.jpg'
	
  # Save the undistorted image
  cv2.imwrite(new_filename, undistorted_image)

  # Close all windows
  cv2.destroyAllWindows() 
	
main()

Output

Here is the original distorted image.

chessboard_input12

Here is the undistorted image, which is the output. Note that the distorted image is the same dimensions as the undistorted image. Both images are 600 x 450 pixels.

chessboard_input12_undistorted

The correction is quite subtle, so it might be hard to see (my camera must be pretty good!). It is more noticeable when I flip back and forth between both images as you see in the gif below.

8-distorted-vs-undistorted

Saving Parameters Using Pickle

If you want to save the camera calibration parameters to a file, you can use a package like Pickle to do it. It will encode these parameters into a text file that you can later upload into a program.

The three key parameters you want to make sure to save are mtx, dist, and optimal_camera_matrix.

Here is a list of some tutorials on how to use Pickle:

Assuming you have import pickle at the top of your program, the Python code for saving the parameters to a pickle file would be as follows:

# Save the camera calibration results.
calib_result_pickle = {}
calib_result_pickle["mtx"] = mtx
calib_result_pickle["optimal_camera_matrix"] = optimal_camera_matrix
calib_result_pickle["dist"] = dist
calib_result_pickle["rvecs"] = rvecs
calib_result_pickle["tvecs"] = tvecs
pickle.dump(calib_result_pickle, open("camera_calib_pickle.p", "wb" )) 

Then if at a later time, you wanted to load the parameters into a new program, you would use this code:

calib_result_pickle = pickle.load(open("camera_calib_pickle.p", "rb" ))
mtx = calib_result_pickle["mtx"]
optimal_camera_matrix = calib_result_pickle["optimal_camera_matrix"]
dist = calib_result_pickle["dist"]

Then, given an input image or video frame (i.e. distorted_image), we can undistort it using the following lines of code:

undistorted_image = cv2.undistort(distorted_image, mtx, dist, None, 
                                    optimal_camera_matrix)

That’s it for this tutorial. Hope you enjoyed it. Now you know how to calibrate a camera using OpenCV. 

Keep building!

How to Detect Pedestrians in Images and Video Using OpenCV

In this tutorial, we will write a program to detect pedestrians in a photo and a video using a technique called the Histogram of Oriented Gradients (HOG). We will use the OpenCV computer vision library, which has a built-in pedestrian detection method that is based on the original research paper on HOG

I won’t go into the details and the math behind HOG (you don’t need to know the math in order to implement the algorithm), but if you’re interested in learning about what goes on under the hood, check out that research paper.

pedestrians_2_detect

Real-World Applications

  • Self-Driving Cars

Prerequisites

Install OpenCV

The first thing you need to do is to make sure you have OpenCV installed on your machine. If you’re using Anaconda, you can type this command into the terminal:

conda install -c conda-forge opencv

Alternatively, you can type:

pip install opencv-python

Detect Pedestrians in an Image

Get an image that contains pedestrians and put it inside a directory somewhere on your computer. 

Here are a couple images that I will use as test cases:

pedestrians_1
pedestrians_2

Inside that same directory, write the following code. I will save my file as detect_pedestrians_hog.py:

import cv2 # Import the OpenCV library to enable computer vision

# Author: Addison Sears-Collins
# https://automaticaddison.com
# Description: Detect pedestrians in an image using the 
#   Histogram of Oriented Gradients (HOG) method

# Make sure the image file is in the same directory as your code
filename = 'pedestrians_1.jpg'

def main():

  # Create a HOGDescriptor object
  hog = cv2.HOGDescriptor()
	
  # Initialize the People Detector
  hog.setSVMDetector(cv2.HOGDescriptor_getDefaultPeopleDetector())
	
  # Load an image
  image = cv2.imread(filename)
		
  # Detect people
  # image: Source image
  # winStride: step size in x and y direction of the sliding window
  # padding: no. of pixels in x and y direction for padding of sliding window
  # scale: Detection window size increase coefficient	
  # bounding_boxes: Location of detected people
  # weights: Weight scores of detected people
  (bounding_boxes, weights) = hog.detectMultiScale(image, 
                                                   winStride=(4, 4),
                                                   padding=(8, 8), 
                                                   scale=1.05)

  # Draw bounding boxes on the image
  for (x, y, w, h) in bounding_boxes: 
    cv2.rectangle(image, 
                  (x, y),  
                  (x + w, y + h),  
                  (0, 0, 255), 
                   4)
					
  # Create the output file name by removing the '.jpg' part
  size = len(filename)
  new_filename = filename[:size - 4]
  new_filename = new_filename + '_detect.jpg'
	
  # Save the new image in the working directory
  cv2.imwrite(new_filename, image)
	
  # Display the image 
  cv2.imshow("Image", image) 
	
  # Display the window until any key is pressed
  cv2.waitKey(0) 
	
  # Close all windows
  cv2.destroyAllWindows() 

main()

Run the code.

python detect_pedestrians_hog.py

Image Results

Here is the output for the first image:

pedestrians_1_detect-1

Here is the output for the second image:

pedestrians_2_detect-1

Detect Pedestrians in a Video

Now that we know how to detect pedestrians in an image, let’s detect pedestrians in a video.

Find a video file that has pedestrians in it. You can check free video sites like Pixabay if you don’t have any videos.

The video should have dimensions of 1920 x 1080 for this implementation. If you have a video of another size, you will have to tweak the parameters in the code.

Place the video inside a directory.

Now let’s download a library that will apply a fancy mathematical technique called non-maxima suppression to take multiple overlapping bounding boxes and compress them into just one bounding box.

pip install --upgrade imutils

Also, make sure you have NumPy installed, a scientific computing library for Python.

If you’re using Anaconda, you can type:

conda install numpy

Alternatively, you can type:

pip install numpy

Inside that same directory, write the following code. I will save my file as detect_pedestrians_video_hog.py. We will save the output as an .mp4 video file:

# Author: Addison Sears-Collins
# https://automaticaddison.com
# Description: Detect pedestrians in a video using the 
#   Histogram of Oriented Gradients (HOG) method

import cv2 # Import the OpenCV library to enable computer vision
import numpy as np # Import the NumPy scientific computing library
from imutils.object_detection import non_max_suppression # Handle overlapping

# Make sure the video file is in the same directory as your code
filename = 'pedestrians_on_street_1.mp4'
file_size = (1920,1080) # Assumes 1920x1080 mp4
scale_ratio = 1 # Option to scale to fraction of original size. 

# We want to save the output to a video file
output_filename = 'pedestrians_on_street.mp4'
output_frames_per_second = 20.0 

def main():

  # Create a HOGDescriptor object
  hog = cv2.HOGDescriptor()
	
  # Initialize the People Detector
  hog.setSVMDetector(cv2.HOGDescriptor_getDefaultPeopleDetector())
	
  # Load a video
  cap = cv2.VideoCapture(filename)

  # Create a VideoWriter object so we can save the video output
  fourcc = cv2.VideoWriter_fourcc(*'mp4v')
  result = cv2.VideoWriter(output_filename,  
                           fourcc, 
                           output_frames_per_second, 
                           file_size) 
	
  # Process the video
  while cap.isOpened():
		
    # Capture one frame at a time
    success, frame = cap.read() 
		
    # Do we have a video frame? If true, proceed.
    if success:
		
    	# Resize the frame
      width = int(frame.shape[1] * scale_ratio)
      height = int(frame.shape[0] * scale_ratio)
      frame = cv2.resize(frame, (width, height))
			
      # Store the original frame
      orig_frame = frame.copy()
			
      # Detect people
      # image: a single frame from the video
      # winStride: step size in x and y direction of the sliding window
      # padding: no. of pixels in x and y direction for padding of 
      #	sliding window
      # scale: Detection window size increase coefficient	
      # bounding_boxes: Location of detected people
      # weights: Weight scores of detected people
      # Tweak these parameters for better results
      (bounding_boxes, weights) = hog.detectMultiScale(frame, 
                                                       winStride=(16, 16),
                                                       padding=(4, 4), 
                                                       scale=1.05)

      # Draw bounding boxes on the frame
      for (x, y, w, h) in bounding_boxes: 
            cv2.rectangle(orig_frame, 
            (x, y),  
            (x + w, y + h),  
            (0, 0, 255), 
             2)
						
      # Get rid of overlapping bounding boxes
      # You can tweak the overlapThresh value for better results
      bounding_boxes = np.array([[x, y, x + w, y + h] for (
                                x, y, w, h) in bounding_boxes])
			
      selection = non_max_suppression(bounding_boxes, 
                                      probs=None, 
                                      overlapThresh=0.45)
		
      # draw the final bounding boxes
      for (x1, y1, x2, y2) in selection:
        cv2.rectangle(frame, 
                     (x1, y1), 
                     (x2, y2), 
                     (0, 255, 0), 
                      4)
		
      # Write the frame to the output video file
      result.write(frame)
			
      # Display the frame 
      cv2.imshow("Frame", frame) 	

      # Display frame for X milliseconds and check if q key is pressed
      # q == quit
      if cv2.waitKey(25) & 0xFF == ord('q'):
        break
		
    # No more video frames left
    else:
      break
			
  # Stop when the video is finished
  cap.release()
	
  # Release the video recording
  result.release()
	
  # Close all windows
  cv2.destroyAllWindows() 

main()

Run the code.

python detect_pedestrians_hog.py

Video Results

Here is a video of the output:

How To Convert a Quaternion Into Euler Angles in Python

Given a quaternion of the form  (x, y, z, w) where w is the scalar (real) part and x, y, and z are the vector parts, how do we convert this quaternion into the three Euler angles:

  • Rotation about the x axis = roll angle = α
  • Rotation about the y-axis = pitch angle = β
  • Rotation about the z-axis = yaw angle = γ
yaw_pitch_rollJPG

Doing this operation is important because ROS2 (and ROS) uses quaternions as the default representation for the orientation of a robot in 3D space. Roll, pitch, and yaw angles are a lot easier to understand and visualize than quaternions.

Here is the Python code:

import math

def euler_from_quaternion(x, y, z, w):
		"""
		Convert a quaternion into euler angles (roll, pitch, yaw)
		roll is rotation around x in radians (counterclockwise)
		pitch is rotation around y in radians (counterclockwise)
		yaw is rotation around z in radians (counterclockwise)
		"""
		t0 = +2.0 * (w * x + y * z)
		t1 = +1.0 - 2.0 * (x * x + y * y)
		roll_x = math.atan2(t0, t1)
    
		t2 = +2.0 * (w * y - z * x)
		t2 = +1.0 if t2 > +1.0 else t2
		t2 = -1.0 if t2 < -1.0 else t2
		pitch_y = math.asin(t2)
    
		t3 = +2.0 * (w * z + x * y)
		t4 = +1.0 - 2.0 * (y * y + z * z)
		yaw_z = math.atan2(t3, t4)
    
		return roll_x, pitch_y, yaw_z # in radians

Example

Suppose a robot is on a flat surface. It has the following quaternion:

Quaternion [x,y,z,w] = [0, 0, 0.7072, 0.7072]

What is the robot’s orientation in Euler Angle representation in radians?

quaternion_to_euler_1JPG

The program shows that the roll, pitch, and yaw angles in radians are (0.0, 0.0, 1.5710599372799763).

quaternion_to_euler_2

Which is the same as:

Euler Angle (roll, pitch, yaw) = (0.0, 0.0, π/2)

And in Axis-Angle Representation, the angle is:

Axis-Angle {[x, y, z], angle} = { [ 0, 0, 1 ], 1.571 }

So we see that the robot is rotated π/2 radians (90 degrees) around the z axis (going counterclockwise). 

And that’s all there is to it folks. That’s how you convert a quaternion into Euler angles.

You can use the code in this tutorial for your work in ROS2 since, as of this writing, the tf.transformations.euler_from_quaternion method isn’t available for ROS2 yet.