Study Of Ampoule Solution Impurity Detection And Recognition Algorithm Base On SVM

Pharmaceutical industry is recognized as the world most one of the international high-tech industry development prospects, and small volume injections of pharmaceutical industry in an ampoule is widely used in clinical medical. In the medical packaging industry, because the glass has superior protection performance and good chemical stability, the vast majority of injection drug are use ampoule for packaging. Due to the limitation of production equipment, ampoule preparation produced in China product quality is uneven, ampoule injection impurity particles (glass chip, fiber, rubber particles) into the blood vessels will cause great damage to the human health, drug quality is good or bad is directly related to the medication safety of the broad masses of the people, the efficient detection of drugs has become the urgent needs of the modernization of industry. In view of this, this paper on the ampule visible particle detection process and core algorithm deeply and systematically.1, This paper expounds the ampoule liquid impurity detection research status at home and abroad and its development trend; The machine vision technology, automatic identification technology and statistical learning theory which are closely related to the system are introduced.2, Analyze the ampoule image in the acquisition, transmission, and storage of liquids may be in the process of all kinds of noise. According to the characteristics of the noise, the use of gray value normalization algorithm, Gauss filtering and other image preprocessing methods were proposed to enhance the contrast of the image, filter out the noise in the image.3, The theoretical basis, classification and process of image registration are briefly introduced. A transformation model estimation method based on least square support vector machine (LS-SVM) is proposed. The advantages and disadvantages of three kinds of image interpolation methods (nearest neighbor interpolation, bilinear interpolation, and the three time convolution interpolation method) are compared and analyzed.4, An image difference algorithm combining background difference and frame difference is proposed. Through the differential operation of the target area image, the motion and geometric features of the impurities are extracted. Finally, the support vector machine (SVM) is used as the classifier to identify the characteristics of the extracted impurities. The experiments show that the algorithm can accurately identify the ampule impurities, the security problems existing in the production process of ampoule preparations provide clear direction for improvement.Treatment based on support vector machine of figure accurate registration and identification method is used in this paper due to mechanical vibrations of the image acquisition device, aiming offset and other factors lead to image relative offset, and can better type of impurity identification, the algorithm of ampule impurity detection and recognition has a certain practical reference value.