How to get an image from the web cam in open condition so that segmentation can be done to identify weeds from crops ??

http://www.mathworks.in/matlabcentral/answers/14921-how-to-extract-the-rgb-values-of-pixel-in-image

I am using Logitech Quick PRO 9000 webcams for detection of weeds in the fields. The problem is coming while calibrating the image as color intensity is totally depend on sunlight exposure in the field.So how to identify green pixel in sunlight exposure with help of a web am and how to calibrate it ? Whether i have to provide artificial light or how to get a good image from video stream so that segmentation can be done by having a threshold limit of green pixels intensity or what ?? So please suggest me a best way to get out from this problem.

Answer 1

[Walter Roberson]

Have you turned off RightLight and manually adjusted the settings? How are you calibrating? Have you ensured that there is color constancy over the entire sensor? (It is common for webcams to have color bleed towards the outside of the sensor.)

Have you also turned off autofocus? How are you adjusting the focus?

If you have the resources to provide consistent artificial light to an entire field, then you have the resources to use an instrument-grade camera rather than a webcam.

[ANKUR]

Yes dear Walter i have turned off all settings leading to auto control for the calibration purpose to count the no of green pixels based on their color intensity in lab condition. While moving to the field i am facing some difficulties like intensities of rgb varies with the solar light intensity. So to use web cam as a color sensor to identify weeds the values are different than calibrating equations.

But it is impossible to provide artificial light to the entire test field. So i will try to follow your suggestion tomorrow by turning on some functions suggested by you. Thanks for your reply Mr. Walter Roberson.

[Walter Roberson]

If I recall correctly, Image Analyst sometimes suggest an illumination correction of an RGB image as

newimg = CurrentImage ./ repmat(max(CurrentImage,3), [1 1 3]);

Answer 2

[Image Analyst]

It should not be too hard to discriminate green from brown or black (dirt) in a wide variety of illumination conditions because they are fairly well separated. Of course green weeds from green "good" crops would be a lot tougher. Where did you upload your images? I can look at them with the 3D color gamut inspector to see what their gamut looks like.