提出了一种新的近红外多光谱检测法对碳纤维复合材料结构机翼表面的残冰进行检测。建立了该方法的理论模型并进行了实验验证。首先，根据冰与水和除冰液在不同红外波段下具有不同反射光谱特性，提出了用于残冰检测的理论算法。然后对参考通道，低通道及高通道所测图像的灰度值进行比较分析， 总结得到了判断机翼蒙皮上残冰存在的对比度阈值C。最后，分析了误差可能产生的区间及消除方法。实验结果表明，对于白色涂层的复合材料结构蒙皮，当C>0.03时，就可以断定结构表面上存在残冰；当C<0时，可以断定表面上肯定没有冰；当C处于0~0.03时，为安全起见需要对蒙皮进行二次检查以消除不确定因素产生的误差。而对于其他颜色涂层蒙皮的残冰检测，要根据不同的颜色设定不同的对比度阈值C。本研究证明了可以将冰与水的近红外反射光谱差异特性应用到以复合材料结构为机翼蒙皮的残冰检测中。

In order to realize the ice detection for aircraft wings made of carbon fiber composite materials, a near infrared multispectral solution to ice detection and measurement is presented. Firstly, a new algorithm for ice detection is proposed based on the different reflectance spectra of water and ice in a certain wavelength band. Then, by analyzing the gray value of reflectance spectra between water and ice detected in the reference sub-band, the lower sub-band and the upper sub-band, the threshold of contrast C to ice detection and measurement on aircraft wings is determined. Finally, this paper analyses the position region of contrast C that error may happen in the measurement, and describes the method to eliminate the error. Experimental results show that for the white painted wing, if the contrast is greater than 0.03, it is considered that the ice is present on the surface;but if the contrast is less than zero, no ice is present. While the contrast is between zero and 0.03, a second detection should be done to reduce or eliminate the effects of uncertain factors.Furthermore, while the aluminum is painted by other colors, different thresholds of contrast should be set to remove this effect. It is proved that the different reflectance spectra of water and ice in a certain wavelength band can be used to detect and measure the ice on aircraft wings made of carbon fiber composite materials.