Skip to content

Latest commit

 

History

History
62 lines (42 loc) · 2.48 KB

Challenge.md

File metadata and controls

62 lines (42 loc) · 2.48 KB
Raw

Challenge 6

Unfortunately, the (Battlestar Battleshippy McBattleshipFace) and its crew have been decimated due to the enemy exploiting a buffer overflow caused by the long ship's name; That because of the vulnerability was broadcasting the ship's location in its communication array pings. Not only that all communication was decrypted because the buffer overflow burped the word "McBattleshipFace" in plain text they were able to eavesdrop on the fleet communication for weeks. On the Battlestar Battleshippy McBattleshipFace the last transmission indicated that the enemy is headed our way.

The few remaining officers have devised a plan to detonate an EMP when the enemy space jumps on top of us. Which meant frying our systems with the enemy as well but at least it means we have a fighting chance.

Few ancient defence Barretts from 2017 have been shielded to withstand the EMP blast because it was built at a time we almost destroyed humanity by nuclear fallout. You are tasked with building an automated targeting system to shoot down anything that space jumps after the EMP.

Suggested Methods

Steps for circle location detection:

  1. Load image
  2. Convert image to greyscale
  3. Do edge detection (Sobel operator)
  4. Do circle detection at various radii (Hough transform)
  5. Draw circle around the most prominent circle(s)

Convert to greyscale

Set TYPE_BYTE_GRAY

Tip: If you know the colour of the circle you are detecting, you can filter for this too.

Sobel operator

https://en.wikipedia.org/wiki/Sobel_operator

For every pixel, apply this matrix: https://upload.wikimedia.org/math/2/a/5/2a56db5dad1226aebd954d0a1e59f59d.png

The value of edge detection is then: https://upload.wikimedia.org/math/3/d/8/3d82fd5595a30ee764dc968e29101012.png

Pick a threshold at which to indicate that this is an edge.

Hough transform

https://en.wikipedia.org/wiki/Hough_transform

For every radius you're looking for
    For every pixel in edge detection (over certain threshold)
        In a new image (where each pixel is a vote), draw circle around point at radius
Count votes to find most likely (x,y) for a circle

Note: you will also need an algorithm for plotting points on the edge of a circle. You can do this with some clever sin and cos wrangling (loop round 360° and plot in the x and y). There is also something called the midpoint algorithm, which is an optimisation of this, that you might consider researching.

Possible Extensions

Make the entire process run in real-time using a gpu.