Research On Ceramic Tile Dimension Detection Algorithm Based On Machine Vision

Ceramic tile, as a kind of extremely important building decoration materials,plays an important role in the national life. Currently, although China is the largestproducer of ceramic tile, but compared with developed countries, product quality islow and has no competitiveness in the international market. One of the importantfactors is missing high-precision automated quality detection equipment. In order toimprove the quality of Chinese products, it’s urgent to develop a high-precisionquality detection equipment, to realize the automation quality detection of ceramictile.The studied ceramic tile images are collected by the way of using upper andinclined parallel light, compared with the under light, the images have droplets, beltwear, chamfering, defiled and patterns such as interferences, how to exclude thesedisturbances and accurately locate tile edges is a focus. The tile images obtainedhave a lot of redundant pixels, but system requires a higher detection rate, how toquickly achieve detection algorithm is another focus.In anti-interference, proposes the two gradient templates and candidate edgepoints screened formula. First, search out the edge reference area with coarselocation gradient template, most of interferences can be directly eliminated. Thenin the edge reference area, adopting the strong anti-interference and accuratelocation gradient template to eliminate interference further, accurately locate tileedges. Last use Hough Transform and Least Squares Method to fit tile edges,effectively avoid the interferences of edge bending, missing angle and fake edge,etc. which will have influence on point location.In quickness, first only search edge points in reference area, rather than searchin the whole image blindly, data quantity is greatly reduced; Second when use theaccurate location gradient template to calculate the gradient, the fast algorithm isadopted; Finally, adopt the idea of incremental refinement fitting edge in the smallrange near the edge reference, can effectively reduce the calculation amount of theHough Transform.Apply the algorithms in the subject to ceramic tile dimension detection system,the system has good anti-interference ability, can detect different tile in roughenvironment, and duplicate detection accuracy is+/-0.02mm, detection accuracyis+/-0.1mm, detection rate is22~30piece/min, meet the requirements ofenterprise for on-line quality detection of ceramic tile.