Project Details

Organisation:	Data Storage Institute (DSI), A*STAR
Mentor:	Dr Tan Eileen
Members:	Tong Haowen Joel
Project Title:	STUDY OF LUMINOUS INTENSITY AND ITS APPLICATIONS IN LASER-AIDED VIRTUAL ON-SCREEN IMAGE INTERACTION SYSTEMS

A. Brief Description of Project
This investigation focuses mainly on isolating the red channel, and hopes to explore the possibility of using a red laser pointer as the physical pointer in this system, on the basis that since a red laser pointer has a much higher luminous intensity than that from a projector projecting 0xFF0000 (full red), the output image can be thresholded and thus the point isolated irrespective of the background colour.

B. Results
The results collected imply that only full red (0xff0000) interferes with the specified threshold, B4. For every test conducted on 0xff0000, there is an extremely high peak or two, suggesting that multiple placards being displayed simultaneously could have caused the interference. This could be due to the threshold not being high enough, and thus a second test was done determining the exact threshold where background interference would not be visible, yet the pointer would be easily spotted.

For this test, the same parameters specified above apply, except that the threshold values of 190, 200 and 210 were tested first without any laser firing. It can be implied from the chart that a threshold value of 190 (0xBE) was not sufficient as interference was still detected. However a raise to 200 solved this issue.

Since the webcam driver auto-balances colors in various conditions such as contrast and brightness, I hypothesized that the high luminous intensity emitted by a laser beam might interfere with the color values of the screen; therefore similar tests were conducted with the laser being fired. The optimal result from these curves should be parabolic, unconnected mountain peaks with zero interference in between.

However, as can be seen from the test on a 200 threshold level, there was interference. This was especially so on the 3rd attempt, where there was a huge influx of interference at 9:4:47:15 to 9:4:48:156 and 9:4:31:393 to 9:4:32:796 (refer to graph), as hypothesized. Hence a higher value of 210 was used. The graphs depicting the threshold values of 210 were better. Generally, optimal peaks were achieved and the peaks clearly defined. Most of the values peaked at around the same point. However, during the experiment, the laser had to be pressed for a longer period of time in order for the dot to be registered. This is reflected in the graphs, where the peaks are more leveled up than the previous experiment. In lieu of this strain on the system, the threshold value was brought down to 205. While there was a slightly worse pickup of interference as illustrated in the graph (connected peaks), it was negligible. It was also hypothesized that the bullet holes left in the game could have enhanced the contrast of 0xFF0000 or created destructive interference and thus the webcam may have picked these up. The trigger worked fine at threshold value 205.

A second color value tested, 0xE60000, worked fine with relatively fewer interruptions than 0xFF0000. It was chosen as it is a perfect substitute for 0xFF0000.

In conclusion, with a white background, the system works very well with most color values. However, while values above 0xE60000 for fully red color values can work well, it is not advisable to use such colors.

C. Reflection

Three content knowledge/skills that we have learnt

1. Image recognition algorithms (although many were not used later), e.g. Sobel filter, Hough transform, etc.
2. Basis of artificial intelligence (how to think like a computer)
3. Types of image manipulation for digital images (how convolution kernels work, how to distort images)


Two interesting aspects about our learning

1. How to code image recognition algorithms into Flash
2. How to make idea sellable

One take-away for life

1. Never say die

Gallery

Insert photos (in jpg, gif format), videos (using Youtube) or animation here.For files such as animations or online games that are larger than 2 MB, please pass them to Mr Tan Seng Kwang directly in the form of a CD and he will put up the link for you.

e.g. to insert photo please save them in jpg or gif formats that are at most 400 pixels in width and height, go to "Image" in the EasyEdit Toolbar.

e.g. to insert video, go to "Widget (video, rss)" in the EasyEdit Toolbar. Make sure you have uploaded the video in either Youtube or GoogleVideo.

e.g. to insert links to existing websites/blogs, click on "Link" in the EasyEdit Toolbar.