基于视觉传达技术的激光雷达遥感图像缺陷分割模型  

Defect segmentation model for LiDAR remote sensing images based on visual communication technology

作　　者：农琳琳 NONG Linlin(Nanning University,Nanning 530200,China)

机构地区：[1]南宁学院,南宁530200

出　　处：《激光杂志》2024年第9期177-182,共6页Laser Journal

基　　金：广西自然科学基金项目(No.2020GXNSFAA148093)。

摘　　要：随着遥感技术的发展,激光雷达遥感图像在各个领域的应用越来越广泛。然而,这些图像中常常存在各种缺陷,如噪声、失真、遮挡等,这些缺陷的存在会对图像的分析和应用造成不良影响。因此,对激光雷达遥感图像实施缺陷分割是必要的。本研究旨在开发基于视觉传达技术的激光雷达遥感图像缺陷分割模型,以提高遥感图像处理的准确性和可靠性。应用双边滤波函数实施激光雷达遥感图像的离散点云去噪,在去噪的同时很好地保留点云的曲面几何特征。基于视觉传达技术中的图像增强技术实施激光雷达遥感图像增强处理,以提高遥感图像中视觉信息的理解性与可读性,选用的图像增强技术为局部对比度增强变分模型。设计融合注意力机制的语义分割网络作为遥感图像缺陷分割模型,实现激光雷达遥感图像缺陷分割。实验测试结果表明,设计模型的mPA较高,整体高于95%,随着测试数据量的增长,设计模型的mPA没有出现显著下降。设计模型对于不同地形场景的MIoU均较高,对于复杂的地形场景也能保持较高的MIoU,鲁棒性较强。With the development of remote sensing technology,the application of LiDAR remote sensing images in various fields is becoming increasingly widespread.However,there are often various defects in these images,such as noise,distortion,occlusion etc.which can have adverse effects on the analysis and application of the images.Therefore,it is necessary to implement defect segmentation on LiDAR remote sensing images.The aim of this study is to develop a defect segmentation model for LiDAR remote sensing images based on visual communication technology,in order to improve the accuracy and reliability of remote sensing image processing.The application of bilateral filtering functions for discrete point cloud denoising of LiDAR remote sensing images preserves the surface geometric features of the point cloud well while denoising.Based on the image enhancement technology in visual communication technology,LiDAR remote sensing image enhancement processing is implemented to improve the understanding and readability of visual information in remote sensing images.The selected image enhancement technology is the local contrast enhancement variational model.Design a semantic segmentation network that integrates attention mechanism as a remote sensing image defect segmentation model to achieve defect segmentation in LiDAR remote sensing images.The experimental test results show that the mPA of the designed model is relatively high,overall higher than 95%.As the amount of test data increases,there is no significant decrease in the mPA of the designed model.The design model has a high MIoU for different terrain scenes and can maintain a high MIoU for complex terrain scenes,with strong robustness.

关 键 词：视觉传达技术 点云去噪 激光雷达遥感图像 注意力机制 缺陷分割模型 

分 类 号：TN23[电子电信—物理电子学]