Re: detection RGB color on palm oil fruit picture to detect the maturity
- From: "Dave Robinson" <dave.robinson@xxxxxxxxxxxxx>
- Date: Tue, 24 Feb 2009 10:16:01 +0000 (UTC)
noridayuothman@xxxxxxxxx wrote in message <316d65d7-0c03-4150-960d-ae904d16f724@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>...
actually i still run my last year project in university. but i don't
know the technique to detect the maturity of the palm oil fruit. my
research is detection based on RGB color. from one picture i snap,
then the system can generate the mean value of red, green, blue
intensity from one surface of picture. i snap 2 surface of palm oil
fruit picture. the mean value RGB for both surface will be calculated
and determined the ripeness of palm oil fruit.Anyone can help me
please. i begging to anyone for help me ;(
I generated a commercial application almost identical to yours, mine was to do with the drying of tea leaves.
First RGB isn't the best colour space to work in, the RGB values will vary significantly with illumination and in the case of fruit, where you have a curved surface, the RGB values will change with the angle the fruit surface normal makes with the optical axis of your camera/light source.
My own choice here was to switch to using Normalized RGB, which removes the illumination component, and gives you a really good space to work in - note however some image processing experts will advocate using something like Hue and Saturation.
If you follow my process, then once you have normalized your RGB signal you can ignore the colour plane which has less discrimination in your task. For my application, where the leaves were turning from green to brown, I selected to use the Redness, Greenness planes, and dropped the Blueness plane.
Now what you need to do is to effectively generate a scatter plot of the average Redness against average Greeness of your Fruit (or whatever planes you have selected). What you should find is that just ripe fruit should form a recognizable cluster on your plot. Now you need to automate recognizing any fruit with those characteristics.
The way that I accomplished this was to form what I call a Tutor mode system. Probably a Neural Net is one way of doing this. I had a Human expert viewing the product, At first the system generated more or less random results regarding the suitability of the product, but the Tutor corrected the result, which was fed back (e.g. back propagation). In this fashion the discriminator 'learned' where the acceptable product cluster was, and required less and less input from the tutor, who eventually just walked away from the machine, leaving it to work autonomously.
I am afraid I don't have any Matlab code for the system, it was written in C++ and is (was?) the basis of a commercial product.
This still leaves you the task of designing you system, and leaves enough research for you to enjoy the project. But hope it might give you a direction to go in.
Regards
Dave Robinson
.
- References:
- detection RGB color on palm oil fruit picture to detect the maturity of that fruit
- From: noridayuothman
- detection RGB color on palm oil fruit picture to detect the maturity of that fruit
- Prev by Date: brightness calculation
- Next by Date: bit-plane operation, for loop
- Previous by thread: detection RGB color on palm oil fruit picture to detect the maturity of that fruit
- Next by thread: Re: detection RGB color on palm oil fruit picture to detect the maturity of that fruit
- Index(es):
Relevant Pages
|
