基于蚁群算法的低照光谱图像场景分类仿真

Simulation of Low-Light Spectral Image Scene Classification Based on Ant Colony Algorithm

目前的低照光谱图像场景分类方法未对原始光谱图像进行预处理,获取的光谱图像分类结果存在准确率低、召回率低、F1值低等问题。基于蚁群算法的低照光谱图像场景分类方法利用K-AP算法对原始光谱图像去噪,通过元灰度斜率公式和L1范数优化模型修复光谱图像环线并消除图像垂直条纹。优化后的光谱图像输入蚁群算法中提取特征,输出光谱图像特征分类矩阵,实现低照光谱图像场景的分类。仿真实验结果表明,所提方法的图像分类结果具有较高的准确率,且召回率和F1指标也较高。

The current method does not pre-process the original spectral image,and the obtained spectral image classification results have low accuracy,low recall rate and low F1 value.The ant colony algorithm based low-light spectral image scene classification method uses K-AP algorithm to de-noise the original spectral image,and uses the element gray slope formula and L1 norm optimization model to repair the spectral image ring line and eliminate the vertical fringe.The optimized spectral image is input into ant colony algorithm to extract features and output spectral image feature classification matrix to realize the classification of low-light spectral image scenes.Simulation results show that the image classification results of the proposed method have high accuracy,high recall rate and F1 index.