A Robust Palmprint Zone Detection Method

• Main Authors: WU, TUNG-MAO, 吳東懋
• Other Authors: HUANG, YEA-SHUAN
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/9sd37p

碩士 === 中華大學 === 資訊工程學系 === 105 === In this paper, we propose a variety of methods for verifying the reliability of finger webs and knuckles and combined with finger web location calibration, to create a set of Palmprint Area detection system. First, by adaboost trained finger web and finger detectors for finger web and knuckle detection. Then as far as possible without affecting the correct detection results of the case, the use of template confirmation, four kinds of finger web detectors and three kinds of finger detectors to delete the wrong finger and finger detector results. After use the results of the detection to filter out the wrong web and knuckles, use the results of finger detection to calculate the location of the web that is not detected by the webbing detector and calibrates the results. Finally, according to the position of the web to determine the palm width of the width of the region, palm area center and palm area rotation angle, And the palm print area of the center point of the horizontal direction of the fine-tuning, In the final output palmprint area. This method obtained 95% recognition rate in PolyU hand image database, And the excess error finger web detection results reduced by 76.8%. The method mentioned in this paper have a stable effect in the case of rotation angle is not too big indeed.