| Summary: | 碩士 === 長庚大學 === 電子工程學研究所 === 96 === As hi-tech becoming more and more advanced, the computer has ushered the automatic or AI system in a new era. “Counterfeiting” is one of the problems in the world. Nowadays, there are several anti-counterfeiting methods in the world. They are anti-counterfeiting label, printing, nano-meter sculpture or dispenser and so on. Therefore, we implement a recognizing image system to check that the labels are correct or not after the process of dispenser machine.
There are two main parts in the recognition system: “template training” and “target image recognition”. After saving the feature information of template training system into the word type of data-based, we progress the first recognition on the target image. When it is fail to match at this time, we compare it with the list of template images which have the low recognition rate at the first matching, and start to execute the second recognition (Local Normalized Correlation; LNC) to find the highest coefficient of LNC(being close to 1). Finally, the matching rate is displayed on the screen. This system includes several sub-systems, like as “image pre-process”, “auto threshold value”, “template rotation”, “region separation”, “feature extraction”, “the first of features matching”, “the second of LNC matching”, and so on. We implement the system into PC environment and embedded system (PXA255 Experiment Board). Both results of the recognition rate are above 80%. At PC, the implement time is between 192 ms and 25207 ms. In embedded system, the shortest and longest time of the process is 8.83 and 231.67 s, respectively.
|
|---|
