Pattern recognition based on machine vision checking up request to ensure accurately the seed pattern, therefore ,a pattern recognition means based machine vision checking up was built, it was made up of the composing of system equipment the gathering data, the image processing, the distilling character, the seed pattern recognition, the coming true of Matlab programme, the precision analysis of pattern recognition, the advantage of technic and bug etc. It was proved, the process of Pattern recognition based on machine vision checking up is feasible and effective, and non-contacting, non-damnification, non-polluting, high precision, convenient, fast, automatization was realized on this process.For CCD vidicon, special request of surroundings, illumination, technic capability did not needed on the gathering data process, it's shooting frequency was 30 Hz.In order to research the means of recognition and review the error of checking up, the examination of two-factors and hypo-positive join and revolving . all possible rotating speed and air pressure were included in the examination.Gathered samples were processed(included adding the contrast, removing the background of image using difference, threshold value segmentation, two value processing, median filtering etc),which seeds were remained in the image.In the process of distilling character, seed number was judged and discerned with characteristic section area as parameter , which eliminate leak count seed because of overlapping of seed and distill accurately frame seed number of sample, according to frame seed number of sample,the frequency distributing of frame seed number can be gained.In the process of solving models, for built the model, the best probability of frame seed number can be solved using of least two multiply non-linear programming.In the process of analyzing error, each error in each tache affecting for general error was analysed.In the process of image procession and analyzing data, all the programme were coming true using Matlab language. The programme of image procession and analyzing data realized processing data fast. Efficiency and precision of algorithmic were better and the algorithmic was of pertinence, expansibility and easy to ameliorate. |