Dear Krishnaveni ji,

It is absolutely possible with template matching and proper threshold selection with a little bit image processing.

Out of several processes the most customizable and the best method is of using a vision library (openCV) and an image processor or a single board computer (SBC) for building the robot.

You can try developing the robot using the following low cost SBCs

1. Beaglebone black [INR 7000 approx]

       or

2. Raspberry pi [INR 3000]

      or

3. ODROID

     and so on......

You can follow the comparison between 1 and 2 in the following link towards selecting appropriate board for  your particular application.

NB: Please select carefully keeping all the parameters in mind, especially about the requirement of any analog sensor interface.

Wish you for the best

Regards,

Bhaskar

https://beagleboard.org/black

https://www.raspberrypi.org/

http://www.hardkernel.com/main/main.php

https://www.engineersgarage.com/blogs/raspberry-pi-2-3-vs-beaglebone-black

Hi Krishnaveni,

There are two ways to interpret what you are calling "control". If you rever to the navigation system, the answer yes. One can detect the weed and diferentiate it from the other plants by a simple image processing or object recognition algorithm. there are several simple approaches you can find in the literature. There are even videos demonstrating how to implement it in several programming lenguages.

If you refered to the control system, the answer is also yes and there are also other several approaches involvin form servoing control to active target tracking control using several control approaches such as nonlinear model predictive control, slidding mode control, and so on.

If you have any questions please contact me.

Regards,

We have done some work back on multiple robot formations using camera, even though it was primitive model which used only specific color at a time may be it would be a good start to gather some basics for your work.

Article A Sketch-Based Interface for Multi-Robot Formations