Coccidian Oocysts Recognition And Roughly Counting Using The Invariant Moment And BP Net

The method of counting egg number is still one of the most immediate and pervasive way in parasitic disease diagnose for livestock nowadays. Usually the specimen smear is observed by manual way with the microscope in diagnosis. It leads heavy workload, complex operating, and counting error, which changing by the experiences of the person who did it. So it’s lack of objectivity, accuracy, and inconvenient for saving and searching the database of the specimen.The seven kinds of Eimeria tenella Coccidian Oocysts from Standard database images are used in this experiment. First, an efficient preprocessing algorithm, which adapt to Pseudo Jacobi Fourier moments, is presented. Secondly, We have analysed the invariant feature of improved Pseudo Jacobi Fourier(p=4,q=3) moments which computed in the Descartes coordinate system; Then Pseudo Jacobi Fourier(p=4,q=3) moments based BP net are trained by120images out of the seven kinds of Coccidian Oocysts from Standard database. Finally, recognize and count the coccidian oocysts which selected randomly from database.The accuracy of recognition and counting reached about90%.Experimental result shows that using the special preprocessing method and BP net can significantly enhance the performance of improved Pseudo Jacobi Fourier(p=4,q=3) moments and also can perfect the application of invariant moment in image recognition.