基于红外小目标图像库的特征识别探测研究

The Analysis of Recognition and Detection Based on the Character of Small Infrared Target Image Library

为了进一步提高红外小目标的识别概率,利用主分量分析法提取了小目标红外图像的主分量特征用于识别探测。在具体计算时,分析了红外小目标图像的一般特征,并建立了图像库作为训练样本。最后对训练样本进行了灰度峰值分类,并选取不同数量的小目标图像进行实验。结果表明,目标在红外图像上呈现的灰度峰值大小几乎不影响小目标的识别,而随着训练样本数的增加,识别效果越好。

In order to increase the recognition rate of small infrared target tracking, the principal component character of the infrared small target was extracted based on PCA for recognition detection. In exact calculation, the infrared image characters of small target were analyzed and formed an image library as test sample. Finally, the small infrared target images were classified based on the gray peak value, and experiment was conducted with different amount of training sample images. The results show that the gray peak value of small infrared target image makes no difference in small target recognition. The more sample image is, the better recognition performance has.