Research On The Method Of Ceramic Tiles With The Random Texture Different Aera Real Time Detecting

In the current ten years, sensor technique level and computer processing speed have been improved continuously with the development of some realms such as ar- tificial intelligence, image processing, machine sense of vision and so on., the pro- duction of the ceramics had realized automated. However, the ceramic surface col- or automated classification is still examined by persons mainly. The problem of the ceramic surface color automated classification is still a developing problem in the world, and there are still the following problems need to face: how to design a suit of image collecting system on the producting line to obtain steady and credible cer- amic sample image; how to pocess extracting surface color features of the ceramic according to small color differences among the ceramics and classification, design detection and classification methods to apply to new categorys(such as antique brick, rain flower brick),how to realize to run steadily on the realized curcumstance of dusty and vibration seriously. Otherwise, the productions are detected by machine should match with vision feeling of people's eyes.This paper discuss mainly the problem of color deifferences detect of the random texture ceramic on the circumstance of pratical production (on the circumstance of dusty, vibration, lighteness influence), whether the exzamined methods of steady and the results after examined could match with people's vision feeling, and does some design about it's hardware realization. The work in this paper as follows:(1)This paper does some analysions and discussions about the research of the present condition, and analyz key factors in the surface color classification, extractsthe image color features, chooses the color space and color difference, and common extract arithmetics of the image feature, introduce the use conditions and merits and shortcomings.(2)This paper analyzes the features of the images in this paper, color is main image feature, and we classify to the image's color, choose color space is premise, extract the image's feature is key factors, chooses color measure formula is assurance. This paper uses CIELab color space, uses impoved fuzzy cluster arithmetic to extract the image's color feature, the image's color difference calculation uses CIEL*a*b* color difference formula in this paper.(3)As collected image is the format of RGB in this paper, but extract color features in the CIEL*a*b* color space, it's need to transform the color space, this paper transforms the color space according to the transform furmula offered by the CIE, then apply the method of extracting color features, extract image's color features, last apply CIEL*a*b* color deifference formula to calculate image's color difference.(4)Due to the system of the title surface color classification need to be installed in the workshop, so it is easy to be interfered by some noises as to many measure and classify arithmetics cannot run steadily in this circumstance such as dusty and vibrated. For the purpose of testing and vertifying the steady of the paper's arithmetic, we imitate three common interferences—dusty, vibration, lightness interference in the pratical product environment. The three interfered images are measured by this paper's arithmetic, test show, this paper'arithmetic has better steady performance. In this paper's interfered images, the arithmetic cannot recognize those interfered extremely images and people can distinguish clearly. This paper's arithmetic has well steady performance to those that are interfered slightly.(5)As this paper's images have random textures, the features of random texture images are complex images those are composed of many kinds of colors and shapes, and the probility of appearance is random. So to testify steady performance of arithmetic further, we move the camera, in this way what the images include will change, there are not color difference in people's vision. We do color difference calculation such images by using this paper's arithmetic, and the result is that there are not color differences. Test show, this paper's arithmetic can measure the color difference of the ceramic tiles with random texture.(6)This paper discusses the influences of threshold to the result of clusting, and chooses proper threshold can make the complex degree of arithmetic and run time reach the best, and for the arithmetic approving assurance that can reach real time performance. This paper analyzes deifferect thresholds to influence curves of clusting results. (7)This paper does some research about the hardware realizing of the color differences classification system. Images are collected by image sensor 7313, memory chip uses IC61LV25616, it's memory capability is 256K16bit. FPGA collects images'data by the collection signal that is sent out by main controller DSP, and memory data, communition with ARM. The chip is the company of Alter ACEK1K50. ARM realize the color differnence arithmetic, the chip is LPC2132. ARM receives the ceramic tiles images form FPGA, executes color classification, and make the result transform to the main controller DSP. DSP controls the mechanic structure's revelant actions, done on-line classification to the ceramic tile images. DSP adopts TMS320F2812 that is the company of TI the series of dsp2000, contols the time of image collecting, and the action of ceramic tiles classification. We compile the experiment programmes by C++Builder.