Research Of Defect Detection About Aluminum-plastic Blister Package Based On Smart Camera

Aluminum-plastic blister packaging is the main form in blister packaging for tablets andcapsules, and it has been widely used in the pharmaceutical industry. In the packagingprocess, defective products may be produced. Traditional methods for packaging defectdetection are time-consuming, inefficient and inaccurate. They may lead to fault detectionand leak detection. In this thesis, on-line solutions about defect detection of thealuminum-plastic blister package based on a smart camera are introduced.Firstly, the design of the hardware system and the software algorithm is demonstrated.TMS320DM6437is chosen as the core processor in the smart camera, the parallel regionalalgorithm is chosen as the main segmentation algorithm, and BLOB analysis is chosen as thedetection classification algorithm. Secondly, the algorithm design of both the self-learningprocess and detection process is described step by step, combined with experimental pillimages. Thirdly, the optimization of the entire software system based on hardware features isintroduced, and a significant increase of speed is achieved. Finally, a series of tests fordifferent kinds of drugs have been conducted, and the results are given.The overall system uses self-learning approach. It can automatically learn geometricparameters of different drugs and detect lack or incomplete defect. Experimental results ofthe pharmaceutical blister package defect detection system show that in natural light, theaccuracy of pill breakage detection is10%at the speed of12frames/sec for752×480images. The detection accuracy rate can reach more than99%when the board is located inthe center with the angle less than5. The detection accuracy rate for different kinds of drugscan be95%when the board is located in the edge of the view field.