Optical flow for indoor positioning in Drone Programming - Time & Space Complexity

Time Complexity: Optical flow for indoor positioning

O(n^2)

Understanding Time Complexity

When using optical flow for indoor positioning, the drone processes many image frames to estimate movement. Understanding how the processing time grows as more image data is analyzed helps us know how fast the drone can react.

We want to know: how does the time to calculate position change as the number of pixels or frames increases?

Scenario Under Consideration

Analyze the time complexity of the following code snippet.

function calculateOpticalFlow(currentFrame, previousFrame) {
  let flowVectors = []
  for (let y = 0; y < currentFrame.height; y++) {
    for (let x = 0; x < currentFrame.width; x++) {
      let flow = computePixelFlow(currentFrame.getPixel(x, y), previousFrame.getPixel(x, y))
      flowVectors.push(flow)
    }
  }
  return flowVectors
}

This code compares each pixel in the current frame to the previous frame to find movement vectors, which help the drone understand its position indoors.

Identify Repeating Operations

Identify the loops, recursion, array traversals that repeat.

• Primary operation: Nested loops over image pixels to compute flow for each pixel.
• How many times: Once for every pixel in the frame, which is width x height times.

How Execution Grows With Input

As the image size grows, the number of pixels grows, so the work grows with the total pixels.

Input Size (pixels)	Approx. Operations
10 x 10 = 100	100
100 x 100 = 10,000	10,000
1000 x 1000 = 1,000,000	1,000,000

Pattern observation: Doubling the width and height multiplies the total operations by about four times, since total pixels grow with area.

Final Time Complexity

Time Complexity: O(n^2)

This means the time to calculate optical flow grows directly with the number of pixels processed.

Common Mistake

[X] Wrong: "The time to process optical flow depends only on the number of frames, not the image size."

[OK] Correct: Each frame has many pixels, and the code checks every pixel. So bigger images take more time even if the frame count stays the same.

Interview Connect

Understanding how image size affects processing time is a useful skill. It shows you can think about how real sensors and data impact drone performance, which is important in many tech roles.

Self-Check

"What if we only computed optical flow for every other pixel instead of every pixel? How would the time complexity change?"

Practice

1. What is the main purpose of using optical flow in indoor drone positioning?

easy

A. To track the drone's movement by analyzing camera images

B. To connect the drone to GPS satellites

C. To control the drone's battery usage

D. To measure the drone's altitude using sound waves

Solution

1. Step 1: Understand optical flow concept

   Optical flow uses camera images to detect movement by comparing changes between frames.

2. Step 2: Identify its use in indoor positioning

   Since GPS doesn't work well indoors, optical flow helps track position by visual movement detection.

3. Final Answer:

   To track the drone's movement by analyzing camera images -> Option A

4. Quick Check:

   Optical flow = movement tracking indoors [OK]

Hint: Optical flow tracks movement visually, not with GPS indoors [OK]

Common Mistakes:

• Confusing optical flow with GPS tracking
• Thinking optical flow measures altitude
• Assuming optical flow controls battery

2. Which of the following is the correct Python syntax to calculate optical flow using OpenCV's calcOpticalFlowPyrLK function?

easy

A. flow = cv2.calcOpticalFlowPyrLK(prev_img, next_img, prev_pts, None)

B. flow = cv2.calcOpticalFlowPyrLK(prev_pts, prev_img, next_img, None)

C. flow = cv2.calcOpticalFlowPyrLK(next_img, prev_img, None, prev_pts)

D. flow = cv2.calcOpticalFlowPyrLK(None, prev_img, prev_pts, next_img)

Solution

1. Step 1: Recall function parameter order

   The function calcOpticalFlowPyrLK expects parameters in order: previous image, next image, previous points, and next points (usually None to compute).

2. Step 2: Match parameters to options

   flow = cv2.calcOpticalFlowPyrLK(prev_img, next_img, prev_pts, None) matches this order correctly; others mix parameter positions incorrectly.

3. Final Answer:

   flow = cv2.calcOpticalFlowPyrLK(prev_img, next_img, prev_pts, None) -> Option A

4. Quick Check:

   Correct parameter order = flow = cv2.calcOpticalFlowPyrLK(prev_img, next_img, prev_pts, None) [OK]

Hint: Remember: prev_img, next_img, prev_pts, None order [OK]

Common Mistakes:

• Swapping image and points parameters
• Passing None in wrong position
• Confusing previous and next images

3. Given the following Python snippet for updating drone position using optical flow displacement:

dx, dy = 5, -3
position = [10, 10]
position[0] += dx
position[1] += dy
print(position)

What will be the output?

medium

A. [15, -3]

B. [5, 13]

C. [10, 10]

D. [15, 7]

Solution

1. Step 1: Understand position update

   Initial position is [10, 10]. Adding dx=5 to x gives 15. Adding dy=-3 to y gives 7.

2. Step 2: Calculate final position

   Updated position is [15, 7].

3. Final Answer:

   [15, 7] -> Option D

4. Quick Check:

   10+5=15 and 10-3=7 [OK]

Hint: Add dx to x and dy to y for new position [OK]

Common Mistakes:

• Mixing x and y updates
• Forgetting to add dy as negative
• Printing initial position instead of updated

4. Identify the error in this Python code snippet for calculating optical flow displacement:

prev_pts = np.array([[100, 150]], dtype=np.float32)
next_pts = np.array([[105, 145]], dtype=np.float32)
dx = next_pts[0][0] - prev_pts[0]
dy = next_pts[0][1] - prev_pts[1]

medium

A. Wrong data type for numpy arrays

B. Missing import statement for numpy

C. Incorrect indexing of prev_pts causing IndexError

D. Using subtraction instead of addition

Solution

1. Step 1: Check indexing of prev_pts

   prev_pts is a 2D array; prev_pts[0] is [100, 150], but prev_pts[1] does not exist (index error).

2. Step 2: Identify cause of error

   Accessing prev_pts[1] causes an IndexError; correct indexing should be prev_pts[0][1].

3. Final Answer:

   Incorrect indexing of prev_pts causing IndexError -> Option C

4. Quick Check:

   Indexing 2D arrays needs two indices [OK]

Hint: Use two indices for 2D arrays like prev_pts[0][1] [OK]

Common Mistakes:

• Using single index on 2D numpy arrays
• Confusing dx and dy calculations
• Ignoring numpy import errors

5. You want to improve indoor drone positioning by combining optical flow displacement with altitude data from a barometer. Which approach best integrates these two data sources in Python?

hard

A. Store altitude as a separate variable and never update position

B. Create a dictionary with keys 'x', 'y', 'altitude' updated each frame

C. Use optical flow data only and ignore altitude for simplicity

D. Replace optical flow with barometer readings entirely

Solution

1. Step 1: Understand data integration needs

   Combining horizontal movement (optical flow) and vertical position (altitude) requires a unified data structure.

2. Step 2: Choose appropriate data structure

   A dictionary with keys 'x', 'y', and 'altitude' allows updating all position components together each frame.

3. Final Answer:

   Create a dictionary with keys 'x', 'y', 'altitude' updated each frame -> Option B

4. Quick Check:

   Combine data in one structure for full 3D position [OK]

Hint: Use one dictionary to track x, y, and altitude together [OK]

Common Mistakes:

• Ignoring altitude data
• Keeping altitude separate without integration
• Replacing optical flow instead of combining