Research On Recognition Methods Of The Storage Media In The Mail Packages

Storage media have been used widely and brought about great convenience to people’s daily life. But there are huge hidden dangers behind the convenience. For example, using the storage media to commit crimes like corruption and stealing the safety information often occurs. Therefore, identification the storage media in the mail packages has great significance to combat the phenomenon of corruption and to reduce the national financial loss.This paper uses the way of X-ray transillumination to recognize the storage media in the mail packages. A new pattern X-ray identification system to recognize the storage media in the mail packages is designed in this paper. This thesis has carried on various experiments. The various experiments choose the IC card, ID card, magnetic stripe of the bank card, SD card, U disk, CD, and the net shield of bank as the experimental storage media, and choose book, EMS, packing box, the keyboard and mouse as the mail packages. This paper makes preprocessing and comes up with a method of glittering and translucent to enhance and segment to the x-ray images collected in the various experiments, feature extraction and K-means clusering algorithm to recognize and classify.In the image processing part, this paper focuses on researching the image enhancement and segmentation. This paper uses the histogram equalization, image sharpening and gray level transformation to enhance images, and uses threshold segmentation and region growing to segment images.The thesis comes up with using the algorithm of glittering and translucent to process multi-level gray scale operation and segmental threshold used to obtain binary images. Depending on the differences of the storage mediums and packages, this paper chooses different thresholds and different levels to processing these X-ray images. After being processed, the gray and binary image details are shown.In the image feature extraction part, this paper has mainly researched the edge detection algorithms, such as Roberts operator, Sobel operator, Prewitt operator, Krisch operator, Laplacian operator, LoG operator, Canny operator, and mathematical morphology algorithm, and made comparisons. After that, the thesis extracts the contour extraction of binary images after segmentation. Feature extraction is based on Moment invariants algorithm. Classification and recognition is based on K-means clustering algorithm.