The Research On Defective Capsule Recognition Based On Gray Image And Color Image

Our country is now the major capsule production place in the world, and the annualoutput is about200billion capsules which accounted for about30%of the world.However, the quality detection of capsule at present is manual inspection in mostfactories. Because of a wide variety and huge number of capsules, and certainrequirement for inspection speed, the manual inspection is not only slow, low accuracyand high cost, but also has great damage to the human eye. Therefore, to use automaticinspection instead can avoid many disadvantages of manual inspection, which is of greatsignificance for the capsule production.This paper introduces an automatic inspection platform for capsule defect based onimage processing, and focuses on defect recognition algorithm aim at capsule image.Capsule defect can be divided into the physical defect and color defect. Inspection ofphysical defect is based on gray image. On the basis of the segmentation and smoothingof capsule image, this paper has studied the inspection of capsule on the conditions thatthe edge of capsule defect is clear and fuzzy. For those defect with clear edge, useadaptive Canny edge detection to get the position of the capsule defect, and use adaptivesegmentation based on Otsu and edge information to acquire defect area and thencalculate its characteristic value. For those defect with fuzzy edge, this paper extractsfeatures of image histogram and apply them to a BP neural network to detect the defect.Because the distribution of the input data of network is uneven, this paper proposes annormalization method based on clustering. Inspection of color defect is based on colorimage. The format of the image acquired first is RGB, this paper uses a method forcalculating the average gray level based on normalized histogram, which is used tocalculate the R,G and B component value of the same color region area, and convertthem to the H,S and V component value and then calculate color difference. This methodhas take the advantages of that the RGB color space can easily express an image and theHSV color space is in accordance with human vision sensory, thus improve the efficiencyof image processing.The automatic capsule defect inspection platform can realize fast and accuratedetection, avoid the shortcomings of manual inspection method, improve the efficiency of capsule production, there for it has certain market prospect.