基于核距离加权的 k-最近邻红外小目标检测

摘要

The obvious nonlinear and non-stable distribution which come from the edge of urban complex background have a great impact on infrared small target detection.By using the k-nearest neighbor discriminant classified deci-sion,an infrared small target detection algorithm based on kernel distance weighted k-nearest neighbor is proposed.The kernel method classifies the raw data of every predicted window by mapping into a high dimensional space,and the distance is weighted for nearest neighbor data.After cropping image,the predicted results can be calculated accu-rately.The experimental results show that the new method has a better performance in suppressing background and enhancing target.%城市复杂背景边缘给空中红外小目标检测带来的非线性、非平稳热辐射信号影响严重。在采用 k-最近邻分类判别决策的基础上，提出了一种基于核距离加权的 k-最近邻红外小目标检测算法。该方法将每个预测窗口内的原始数据核映射到高维空间中进行分类，再对各近邻进行距离加权，遍历图像后得到预测结果。实验结果证明了该方法在抑制背景、增强目标方面都有较好的效果。