how to acquire and process images in uneven lighting situations?

[பரத்_குமார்(bharathkumar)]

I want  to  acquire and process same images in different lighting situtions, like in morning till evening or night  how to do it using vision assistant in labview. For example I want to use pattern matching of  multimeter pin lead in all lighting conditions. please help

[Klemen]

Hello,

pattern matching should work for uniform lighting changes, since it is based on normalized cross correlation (normalization with the image standard deviation). Basically every image processing function has some parameters that can be altered for different types of setup.

It is really difficult to say what would work for you without the provided images in different conditions. Perhaps pattern matching for rough detection and than edge based filter for fine tunning. You could also apply some sort of linear contrast correction based on the time of the day (or something similar such as histogram equalization etc...).

It is always easier to help if there is some material too work with provided (your actual images).

Regarding acquisition you should take a look at auto-contrast and auto-brightness. Maybe test this a bit, but in my opinion the acquisition should have all values manually selected.

Best regards,

K

[பரத்_குமார்(bharathkumar)]

Thanks for  ur reply K. I am trying to track my face using vision assistant  due to uneven lighting condition I am unable to do it I tried pattern matching bys minimizing the score however it will not track other person face. I have had little sucess with pattern matching. So I converted the images to grayscale and tried thresholding due to changing brighness on the face threshold is not perfect. But I am able to track my face to medium extent but after thresholding I used circle detection which is detecting other dark circle regions i.e if you bring any dark circular object it detects that too. Is there a way to combine pattern and threshold together to perfectly detect the face?? Thanks for posting the link I am going to see the algorithms and works done by you it is intresting.

[Klemen]

Hello,

for face tracking, take a look at this:

http://mymobilerobots.com/myblog/academic/tutorial-opencv-2-4-3-face-tracking-detection-using-vs-201...

If i have some time in the future, i will try to build the code as a dll (that could be called in labview).

Best regards,

K

[ZcuBa]

For the multimeter pin problem:

Assuming that the needle is of a different color than the scale..

Try seperating the image into color channels HSL.

simplified:

Hue is the angle of the color on a color wheel for a pixel

Saturation is the ratio of color and gray for a pixel

Luminance is the gray image, where white is alot of light, and black is nearly no light.

Typically the Hue channel will be less effected by lighting conditions than the other channels, (if there is light enough to see the colors)

so using the image with only the hue channel, can make it easier for you to track it consistently