Warp and weft density is commonly considered as one of the most important index for estimating the quality of fabrics.
However, the detecting method adopted is basically base on manually counting. It is rather inefficient. This article
demonstrates how to construct efficient system to achieve automatically detection. Fabric is first coupled into the image
sensor by a specific optical system, whose resolution matching is carefully designed. And other component parameters
are also well considered. What is more, the system contains a light source to provide uniformed illumination. A circuit
board with camera module is constructed to perform the role of image processing platform. With amazingly agile
performance provide by an ARM920T processor, and featuring an incredible breadth of peripheral interfaces, the
platform is well suit for fabric density detection. The periodicity within the captured image furnishes convenience for
analysis in frequency domain. After FFT the intensity peaks ranked orderly around the original point. However, the peaks
in the spectrum are always blurred by severe interference whose form is usually an image or a pattern on the fabric. For
this reason, a specific algorithm should be worked out to erode the peaks from the blurred edge to the center. Sometimes,
the central peak is so severely blurred and the central peak is actually useless for us, a unique algorithm is employed to
kick off the central peak. By this way, the position of the surround peaks can be easily located. And the density can also
be easily worked out.