The work aims to propose an improved algorithm to quickly complete the extraction of blister area, and complete the defect identification of aluminum plastic plate, such as defect, missing, etc. in view of the shortcomings of traditional NCC algorithm, such as large calculation and failure to adapt to image rotation. Firstly, the original image was preprocessed and then transformed to the designated position by affine transformation. Secondly, a single blister area was extracted as a template and an integral map was constructed. Finally, the image to be measured was obtained, and then transformed to the same position for table matching. Compared with the traditional method, the speed of the improved algorithm was greatly improved. The matching time of 1920×1200 images was only 21 ms, the detection error rate of the experimental samples was 0, and the missed detection rate was 3.5%. The improved NCC matching algorithm not only meets the accuracy requirements, but also has the advantage of faster speed, which can be well applied to the rapid extraction of blister areas in the defect detection of aluminum plastic blister packaging, and the identification of defects with more than 10% of drug particle and missing loading defects.