Machine Vision Image Segmentationg Based On Rough Sets

Today with the speedy development of digital information technology, image analysis has been widely used in practice. Generally speaking, in any field of image application, image segmentation is indispensable for extraction and measurement of target image. In recent years, image segmentation has become more and more important in image coding and a lot of other image applications. In this paper, we mainly discussed design and realization of image segmentation algorithm in machine vision based on rough sets.In machine vision, vision sensors are used to get planar image from three-dimensional scenery. By using vision processing algorithm, one image or multi images are processed, analyzed andexplained, and symbolic description on relational scenery is obtained. The algorithm provides useful information for robots to finish given tasks and leads the robots' actions.In many practical instances, robots must have the ability of distinguishing features of each object in order to differentiate different objects. To solve the problem, it is necessary to select a suitable method to process the images so as to get a good result. Additionally, in complicated image processing system, data processing and computation are complex and time consuming. In order to set up an image processing system easy to realize, this paper firstly put forward to a preprocessing method, which before making up a knowledge representation system, we fistly use multi-thresholds segmentation method based on tricolor components to enhance and filter the image. Then based on RS knowledge, an image is looked on as a knowledge representation system. In this stablished system, tricolor components of color image are as condition properties, and their corresponding colors are as decision properties. Here, since all the pixels in the preprocessed image are given special threshold, these datas can bedirectly processed as property values, which have been discretized and normalized. Because these numerical values are obtained after removing noises and enhancing image, the gotten initial decision table is comparatively accurate. This processing method provides convenience for posterior assorting decision establishment, assorting decision table reduction, minimum decision rule extraction and final image segmentation. It simplifies the algorithm and decreases computational complexity.In this paper, we analyzed vision subsystem of the soccer robot's matching system in detail and provide the flow chart of the algorithm based on rough sets. According to the requirement of good distinguishment of color and fast processing speed, We analyzed the feasibility of digital image segmentation method based on RS. Finally the results and differences of the images, which are processed using traditional method and RS method are given. From the results we concluded the method based on rough sets used in digital image processing is better than traditional ones.We design the simulation program in MATLAB 6.1 Image Processing Toolbox. In this paper how to ascertain the structure ofthe arithmetic and ensure the speedy processing are difficulties and emphases.Finally, the algorithm was estimated combining the simulation program. The result show the arithmetic is good for digital image processing.