基于改进自适应阈值局部三值模式的遥感图像分类

遥感图像背景复杂,存在光照变化和噪声干扰,导致图像分类准确率不高。针对该问题,在计算邻域像素离散度的基础上,通过对其施加不同权重以细化阈值范围,提出一种改进的自适应阈值局部三值模式(ATLTP)纹理特征提取算法,以提高遥感图像分类精度。首先,对原始遥感图像进行灰度拉伸预处理以增强图像对比度;然后,采用改进自适应阈值局部三值模式提取遥感图像的纹理特征;最后,利用支持向量机对遥感图像进行分类。在标准遥感图像数据集中稀疏建筑物和密集建筑物分类的实验结果表明:采用改进后的局部三值模式纹理特征对遥感图像进行分类的性能要优于传统的局部三值模式,验证了改进算法的有效性。

Remote sensing image classification based on adaptive threshold local ternary pattern

The classification accuracy of remote sensing images is usually affected by the complex background, the interference of varied illumination and abundant noise. To improve the classification performance, a novel texture feature extraction algorithm is proposed based on an improved Adaptive Threshold Local Ternary Pattern (ATLTP). On the basis of calculating the standard deviation of neighboring pixels, different weights are applied to the standard deviation to accurately limit the range of threshold. As for the proposed image classification algorithm, the original remote sensing image is firstly pre-processed by grayscale stretching method to enhance the image contrast. Secondly, the ATLTP is used to extract the texture features of remote sensing images. Finally, the remote sensing image is classified by using Support Vector Machine (SVM). The experimental results of classifying sparse buildings and dense buildings in the standard remote sensing image dataset demonstrate that the improved ATLTP is better than the traditional Local Ternary Pattern (LTP) in classifying the remote sensing image, which proves the effectiveness of the improved algorithm.