Detection of the small target in clutter, usually regarded as singular points in the infrared image, is an important issue
in infrared searching and tracking (IRST) system. Because of the far range of the target to the sensor, the stealth
technology, the effects of inherent sensor noise and the phenomena of nature, the target is more difficult to be detected.
Multispectral sensor system has been proved it could greatly improve detection of the small, hard-to-find targets by
multispectral processing techniques (such as sensor or image fusion). Aiming at the problem of multispectral IR Target
Detection, a kind method of the multispectral IR target detection is proposed, based on the existed detection systems. In
this method, the image registration is done firstly to make the different sensors have a same scene. Then, a fusion rule,
named as adaptive weighted voting theory, is developed to combine the target detection results from the different spectral
sensors. The adaptive weighted voting theory can give the different weights, based on the different spectral IR
characteristics, and these weights decide the detected target is identified as real target or background. The experimental
results show that the proposed method can reduce the detection uncertainty and improve the detection performance.
Compared with the single spectral detection results and the others fusion detection methods, it can decrease the lost
alarm rate and the false alarm rate effectively. The proposed method has been employed in our IR surveillance system,
and it is easy to be used in the various circumstances.