Visual Code Marker Detection

An image processing algorithm to detect visual code markers and extract the embedded data is presented. The algorithm uses an adaptive thresholding technique to produce a binary image with an accurate representation of each visual code marker from an input RGB image. Visual code marker detection is performed on the binary image using a series of tests that leverage prior knowledge of the distinctive features of a visual code marker, namely, the long and short guide bars and three cornerstones. Embedded data bits are subsequently extracted using a projective transformation matrix that maps the code plane to the image plane and prior knowledge of the data area in the code plane. The algorithm is able to detect visual code markers in an image, correctly identify the origin coordinates, and accurately extract the data in each visual code marker under challenging imaging conditions.