| Egg is an important animality nutritious food in daily life, its quality associate with egg production grading, marketable competition and economical benefit, egg inspection and grading are also significant tache in augmenting its export commerce. Manual egg quality inspection lacks of objectivity and is time-consuming. An automatic egg quality inspection method is proposed based on image processing technology and fuzzy C-mean clustering algorithm.Egg images are acquired and pre-processed,egg feature parameters are extracted such as pixels summation, vertical axe, longest, upper and lower horizontal axes by moment theory, and weighted up with electronic balance, then the stepwise regression model of egg weight detection is found: Notable pertinence in egg weight and image pixels number is obtained; its correlation coefficient is 0.892. The linear polynomial and exponential fitting equation of egg weight's actual value versus examination value are found, the correlation coefficient are up to 0.9736, 0.9738 respectively.Aimed at the default of fuzzy C-mean which is highly sensitive to noise data, outliers and is confined to the distribution of class, possibilistic clustering method is adopted which can preferably deal with it, the improved FCM algorithm based on concepts of weighted feature samples is proposed. The algorithm is applied to classification experimentation of pixels summation and longest horizontal axe in egg image and is proved to be a better clustering efficiency. The classification accuracy is up to 88 .89%or bigger eggs, 100 %for normal eggs and 90 .00%for smaller eggs.Improved weighted FCM is utilized to cluster the pixels distribution of 2D gray histogram by incorporating the spatial neighborhood into the conventional FCM clustering algorithm, and defection images of egg sample are classified according to subjection degrees. The classification accuracy is up to 91 .18%for crack eggs, 92 .31%for normal eggs and 100 %for dirty eggs.
|
PDF Full Text Request