Skip to content

Latest commit

 

History

History
40 lines (29 loc) · 2.62 KB

README.md

File metadata and controls

40 lines (29 loc) · 2.62 KB
Raw

Line Following Using Image Processing

Image processing uses lot of computatiion power and it is very slow when implemented with the conventional algroithms for line following further it becomes even slower when implemented in embedded boards such as Raspberry pi and Beaglebone as the clock speed of such microprocessors is around 1 Ghz to 1.2 Ghz as of 12th May 2017. But this algorithm is different from the rest as with minimual computation the task is achieved.

Demo of the Video can be found here and also here. Implementation of the project on BeagleBone Developement demo video is given here

Asumptions Made

  1. The left hand side and the right hand side of the line is of same colour and the line colour is darker than the arena's color (hue of the line is stronger than the arena's color).
  2. The surface is not glossy and there is no bumps on the surface.
  3. The camera is positioned properly and see's only the line.

Logic of the algorithm

step 1: Take the frame in BGR value (has OpenCV takes the input as BGR value instead of RBG value)
step 2: Convert the imgae to binary image using otsu's adaptive-threasholding.
step 3: Now the image would be converted to black and white. The line will be in white where as the background will be black.
step 4: Fix the y axsis and run a linear search for finding the first white pixel and then from then on find the first black pixel. step 5: Those two corrdinates is the location of the line in the frame.
step 6: Divide the frame into half and if the both the coordinate are lying on left of the divider then the camera is moving towards right ( the camera might be mounted on the bot ) and if the both pixel coordinate is lying to the right then the camera is moving left. Change the movement of the bot accordling to bring back on line.
step 7: If the coordinates is located at either sides of the divider then the camera is said to be moving straight.
step 8: Repeat steps 1-7 for each and every frame captured.

Improvements can be made

  1. If we go for multi core-programming then the system can go even faster.
  2. If the algorithm is integrated with PID line controller algorithm then there will be less deviations.

Software Requirements

  1. OpenCV
  2. Python 2.7

NOTE: The code is developed for laptops but can be extended for bots. The only thing you would have to do is add motor control such as PWM and Direction.