A Comparative Study Using Artificial Neural Networks for Military Imagery Classification

• Main Authors: Yu-Shang Lun, 藍于絢
• Other Authors: 左杰官
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/07828659123205905404

碩士 === 國防大學管理學院 === 資訊管理學系 === 97 === In recent year, multimedia information has been playing an important role to the human communication, and the mostly be used are imagery. It becomes an important research subject to find the right image from huge image database and to automatically capture image data feature. The picture searching system is based on CBIR. The advantage of such method is the searching result totally determined by the picture content. The subjective decision by the human will become less when picture is no more searched by text form. It is a trend that CBIR and digital image content serve an important role, the information capture has become an important study. Our research sets using military image information as a goal, we try to develop the method for military-related image information capturing and categorying. The method we had developed providing the construction of the follow-up military image information. We can extract the original image features by color, texture and shape. Besides we can compare separately by feature values captured from FFB, PNN, LVQ, ELM. This research mainly contributes to realize the effect of using color, texture and shape features. We are going to discover under which neural network training will get the highest success rate. The results will be one part of the identification of military image information and the military image information digital museum in the future.