基于机器学习的激光遥感图像特征分割

Feature segmentation of laser remote sensing image based on machine learning

激光遥感图像的特征分割中受到相似纹理和亮度不均匀性的影响,导致图像的冗余信息较多,为了提高激光遥感图像分割精度,提出基于机器学习的激光遥感图像特征分割方法。构建激光遥感图像分块拓扑结构模型,根据图像的拓扑结构信息进行图像像素分组,根据激光遥感图像像素分组之间的属性差异性进行冗余信息滤波,提取激光遥感图像的超像素特征量,构建像素晶格边界,采用机器学习算法进行图像特征分割的最优边界和网格寻优,实现激光遥感图像特征分割优化。仿真结果表明,采用该方法进行激光遥感图像特征分割的像素分布均匀性较好,对冗余信息的滤除能力较强,提高了图像特征分割的精度和识别能力。

The feature segmentation of laser remote sensing image is affected by similar texture and non-uniformity of brightness, which leads to more redundant information. In order to improve the segmentation accuracy of laser remote sensing image, A method for feature segmentation of laser remote sensing image based on machine learning is proposed .The block topology model of laser remote sensing image is constructed, and the pixel grouping is carried out according to the topological structure information of the image, and the redundant information filtering is carried out according to the attribute difference between the pixel groups of the laser remote sensing image. The super-pixel feature quantity of laser remote sensing image is extracted and the lattice boundary of pixel is constructed. The machine learning algorithm is used to carry out the optimal boundary and mesh optimization of image feature segmentation to realize the optimization of laser remote sensing image feature segmentation. The simulation results show that the proposed method is more uniform in pixel distribution and better in filtering redundant information , and improves the accuracy and recognition ability of image feature segmentation.