局部加权拟合的无人机遥感影像多尺度检测  

Multi-scale detection of unmanned aerial vehicle tremote sensing images based on local weighted fitting

作　　者：雷磊 张智 王良 王少军 王辰曦 Lei Lei;Zhang Zhi;Wang Liang;Wang Shaojun;Wang Chenxi(State Grid Shaanxi Electric Power Research Institute,Shaanxi Xi'an,710100,China;State Grid(Xi'an)Environmental Protection Technology Center Co.,Ltd.,Shaanxi Xi'an,710100,China;State Grid UHV Engineering Construction Company,Beijing,100052,China;State Grid Shaanxi Electric Power Company,Shaanxi Xi'an,710048,China)

机构地区：[1]国网陕西省电力公司电力科学研究院,陕西西安710100 [2]国网(西安)环保技术中心有限公司,陕西西安710100 [3]国家电网有限公司特高压建设分公司,北京100052 [4]国网陕西省电力公司,陕西西安710048

出　　处：《机械设计与制造工程》2023年第11期77-80,共4页Machine Design and Manufacturing Engineering

摘　　要：传统影像检测方法的几何校正准确率低,边缘细节提取效果差,为此提出基于局部加权拟合的无人机遥感影像多尺度检测方法。通过几何校正方法提取遥感影像的完整边缘,采用改进局部加权拟合算法确定局部校正区域;以多尺度、多方向结构元素获取影像边缘特征,结合多尺度自适应均衡边缘检测技术实现形态学重构和梯度运算,计算影像灰度值方差权重比;通过自适应均衡求和获取遥感影像边缘轮廓,实现无人机遥感影像多尺度检测。实验结果表明,该方法能够增强影像几何校正准确率,提升边缘细节提取效果,影像内噪声较少。Aiming at the low accuracy of geometric correction and poor edge detail extraction by traditional image detection methods,a multi-scale unmanned aerial vehicle remote sensing image detection method based on local weighted fitting is proposed.The complete edge of remote sensing image is extracted by geometric correction method,and the local correction area is determined by improved local weighted fitting algorithm.Multi-scale and multi-directional structure elements are used to obtain the image edge features.Combined with the multi-scale adaptive equalization edge detection technology,morphological reconstruction and gradient operation are realized,and the variance weight ratio of the image gray value is calculated.The edge contour of remote sensing image is obtained by adaptive equalization sum,and multi-scale detection of UAV remote sensing image is realized.Experimental results show that the technique can enhance the accuracy of geometric correction,improve the edge detail extraction,and reduce the noise in the image.

关 键 词：局部加权 拟合算法 遥感影像 几何校正 多尺度检测 

分 类 号：P231[天文地球—摄影测量与遥感]