In this easter, we finished a "non-electronic" project and find no where to post. Hope it is not so off topic here:)
There are few web cameras installed to monitor the weather or traffic conditions online, so what else information can we get from these images? One important thing to us is the color of the sky. It is a sign of weather and also the pollution condition. So this small project comes up.
we developed a program to analyze the color of the images and look for the best blue.
The process includes four steps:
1) Fetching image from the web camera (an automatic web image fetcher is used)
2) Color Segmentation (Masking out the non-sky component)
3) Analyze the blueness (Finding out the best blue)
4) Output a small square image colored in best blue found in step 3
The color segmentation is the most difficult part, the blueness is hard to define. Luckily, we can easily find many digital images nowadays in flickr or other online gallery system. We used the images to train our color segmentation engine and also with some human fine tunings.
We got 365 images from the Hong Kong Observatory (Yes, we live in HK) in the year 2006.
The sky with no blue-like color is ignored in step 4, it happened in cloudy or rainy or dusty day and so you will see a gray or white cloudy sky, i.e. no color captured.
A processed image from step 1 - 3
(Buildings are masked)
A sky color palette is generated for the year 2006 in HK:
- the pink box is the start of the month (e.g. JAN)
In a statistical view, we make a sky color bar chart for each month.
What we can get in this chart?
1) Second half of the year, we get a clearer sky.
2) In August, we have more blue sky days.