一种可迭代基于多向自相关的航拍电力线图像增强方法

An Iterable Multidirectional Autocorrelation Approach for Aerial Power Line Image Enhancement

针对无人机航拍电力线图像环境背景复杂、电力线目标细弱导致目标识别率低的问题,提出了一种可迭代运行的多向自相关(iterable multidirectional autocorrelation,IMA)增强方法.该方法根据航拍图像中电力线目标的局部纵向及横向灰度分布特征设计有效的滤波模板,用方向滤波的结果进行自相关增强.同时,这种自我增强可以多次迭代运行以达到满意的图像增强效果.通过一系列实验将Canny、Hessian与IMA方法的增强结果进行对比,实验结果显示,所提出的IMA方法比Canny和Hessian方法更适于无人机航拍电力线图像的增强操作.IMA方法不但运算速度快,而且能在大幅减弱航拍图像中复杂环境背景的同时增强电力线目标,从而有效提高图像的电力线目标检测识别率.

A power line image photographed by UAV （unmanned aerial vehicle） has usually a complex background, wherein the thin power lines are so weak that the target lines detection rate is low. To solve this problem, an iterable multidirectional autocorrelation （IMA） approach is proposed to enhance image. Firstly, an effective filtering template is designed according to the local grey level distribution along longitudinal and lateral directions of a power line in a UAV aerial image, and the results of the directional filtering are used to perform an autocorrelational enhancement. The autocorrelational enhancement can be performed iteratively to get a satisfactory image enhancement result. Image enhancement results of IMA are compared with those of Canny, Hessian approaches in a series of experiments. Experiments results show that the proposed IMA approach is more suitable for UAV aerial image enhancement than Canny and Hessian approaches. The IMA approach is fast, and it can weaken complex background in aerial image dramatically while enhancing power line targets, which effectively improves recognition rate of power line targets in images.