Customized Discriminant Analysis Optical Character Recognition for Diverse-Font License Plate Recognition and Its Android Embedded System Implementation

• Main Authors: Lin Li chieh, 林立杰
• Other Authors: Chian C. Ho
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/bx6dw4

碩士 === 國立雲林科技大學 === 電機工程系 === 103 === Since 2012, there are 2 font types of license plates coexisting in vehicles of Taiwan. Because some identical characters between 2 font types of license plates have low feature similarity, and moreover some different characters between 2 font types of license plates have even high feature similarity. The conventional license plate recognition system based on old-font type of character training dictionary is inevitable to get a seriously poor recognition rate. This thesis proposes Customized Discriminant Analysis Optical Character Recognition (CDA-OCR) for diverse-font license plate recognition to fuse the feature diversity of some identical characters between 2 font types of license plates for maximization of the within-class similarity, and to separate the feature diversity of some different characters between 2 font types of license plates for minimization of the between-class similarity, during the stage of character feature training. Experimental results show the developed license plate recognition system based on proposed CDA-OCR gets a nearly perfect recognition rate for 2 font types of license plates coexisting in vehicles of Taiwan, without extra cost of execution time. On the other hand, on Android portable platform, this thesis smoothly implements “Billing Device Input by Automatic License Plate Recognition” constructing of 2 modules of “automatic license plate recognition” and “portable printer”. It can solve faults of error-prone key-in and time-consuming key-in to improve the operational efficiency of roadside parking lot fee collection service and drive-through service.