| In recent years, China's ceramic tile industry was developed rapidly, mechanization and automation of production lines was improved, but automation level of tile quality inspection is relatively low. Along with the computer vision technology widely used in industry, agriculture and other fields, many research institutions at home and abroad do lots of researches on quality inspection for ceramic tile production detection, while researches on shaped ceramic tiles classification are rare.According to the situation and R & D need of tiles manufacturers to enhance the competitiveness, this thesis based on digital image processing technology, researched practical issues of region detection,feature extraction and image matching of shaped tiles depth, combined with related hardware technology of microcontroller and other, developed the classifying system online for shaped tiles based on image.This research mainly includes the following aspects:First, paper gives the review of tile visual detection system home and abroad and the lack in shaped tiles, proposing necessity of classifying system online for shaped ceramic tiles.Secondly, paper introduces the classifying system's overall design, specifically including the selection of cameras, selection of processor and interactive equipment, pattern signal transferring circuit board design and light source selection.Thirdly, paper introduces the operation characteristics and specific features of system software interface and the key technology used in software design process.Fourthly, we focus on the image segmentation algorithm based on boundary feature and the continuous detection algorithm based on gray intensity change on the falling edge, which are used for region detection and pattern acquiring, and explained the actual analysis of tile image to shape the design ideas of the algorithms related.Fifthly, we focus on necessary image pre-processing and feature extraction algorithm based on edge density to realize our shaped tile image pattern match. And then describe contrast of recognition effects applying algorithm based on co-occurrence matrix and algorithm based on edge density.The last part gives the summary of full text, analyzes the insufficient of the system and proposes the future improvements. |