地面熱像目標之辨識研究

• Main Author: 黃敏昱
• Other Authors: 黃炳森
• Format: Others
• Language: zh-TW
• Published: 2004
• Online Access: http://ndltd.ncl.edu.tw/handle/41157040869646401648

碩士 === 國防大學中正理工學院 === 電子工程研究所 === 91 === Since the imaging process of an infrared image is to use the collection of measured spectral values from infrared sensors, the produced image quality is easily affected by the factors of weather, temperature, and lighting. Therefore, the recognition accuracy is decreased. Hence, the selected features shouldn't be affected by weather for the recognition of the ground thermal target. PCA (Principal Component Analysis) is used to statistically analyze images and remove the correlation and noise. Then, those images are mapped into the eigenspace where the recognition is achieved. On the other hand, the feature of fractal dimension is not affected by the translation, scaling, rotation, and illumination of the object. It is suitable for the recognition of thermal targets. Therefore, the aim of this research is to combine those two features to recognize thermal targets. The first step is to search the most possible candidate targets from the scene by using PCA and the feature of fractal dimension is further used for verification. Apart from those two features, the characteristics of thermal targets are used in the experiments to eliminate impossible targets. Since the number of possible targets is greatly reduced, the computation time of target detection is also decreased. Therefore, the real-time requirement is fulfilled.