基于图像分割的无人机遥感影像目标提取技术

Target Extraction Technology of Uav Remote Sensing Image Based on Image Segmentation

为全面增强遥感影像上地物波谱的反射特性能力,实现对无人机目标的无误提取,提出基于图像分割的无人机遥感影像目标提取技术;在类哈尔滤波器结构的支持下,按照区域环境中地物目标的颜色特征,完成低层影像特征的量化处理,实现基于图像分割技术的无人机遥感深度图获取;定义与无人机遥感影像相关的基本名词,通过原始特征选择的方式,判定地物波谱的平均反射特性水平,得到准确的特征元素相关性测度数值,完成无人机遥感影像的目标融合处理;在此基础上,分割多分辨率条件下的影像目标,在定义图像纹理与尺度条件的同时,得到最终的地物提取结果,实现基于图像分割无人机遥感影像目标提取技术的应用;对比实验结果表明,在初始采集相位条件及亚像素条件下,目标地物的波谱宽度均超过7.0μm,遥感影像的反射特性能力大幅提升,满足对无人机目标无误提取的实用需求。

In order to comprehensively enhance the ability of the reflection characteristics of the ground object spectrum on the remote sensing image and realize the unmistakable extraction of the UAV target,a target extraction technology of UAV remote sensing image based on image segmentation is proposed.With the support of the Haar-like filter structure,the quantization of low-level image features is completed according to the color features of the ground objects in the regional environment,and the acquisition of remote sensing depth maps for UAVs based on image segmentation technology is realized.Define the basic terms related to UAV remote sensing image,determine the average reflection characteristic level of the ground object spectrum through the method of original feature selection,obtain accurate feature element correlation measurement value,and complete the target fusion processing of UAV remote sensing image.On this basis,segment the image target under multi-resolution conditions,while defining the image texture and scale conditions,get the final feature extraction results,and realize the application of UAV remote sensing image target extraction based on image segmentation.The comparison experiment results show that under the initial acquisition phase conditions and sub-pixel conditions,the spectral width of the target feature is more than 7.0μm,and the reflection characteristic capability of the remote sensing image is greatly improved,which meets the practical need for unmistakable extraction of UAV targets.