Identification vs. Color Measurement

This image represents a single-colored block. The image of the block is multi-colored with the distribution of colors in the 2-dimensional image space providing information on the shape and orientation of the object in 3-dimensional space.

Color measurement is usually concerned with determining the color of a known object. Color measurement might be used by the manufacturer of the block to insure that the right color paint is being applied. In a process control application, the measured color may often be used directly in a feedback loop to control the paint color.

Color-based identification is concerned with using the colors in an image to determine which object or class occupies a particular region. Here the information required is the identity of the object or class. In the case of blocks this might be to assure that blocks intended for different age groups are not inadvertently packaged together. The colors act much like a bar code or RFID tag; the more colors appear in a region the more information can be carried. Systems like WAY-2C, designed for identification should expect the presence of multiple colors in each region of interest.

If color-based identification is your objective you should consider testing your system on images such as the one to the right. Each of the nine panels has a simple but distinct two-color pattern. Using a system designed for color-based identification, it should take less than ten minutes for you to permanently train the system to correctly identify all panels, no matter how complex their color distribution. For this test image the identification should be reliable whether an entire panel is shown, or just a randomly chosen region as small as one percent of the panel area. To download a full-sized copy of this image to test your system, select the appropriate file format from the column on the right:

BMP GIF JPG PNG TGA TIF

Definitions