基于线性谱聚类的林地图像中枯死树监测

将基于线性谱聚类超像素的方法应用在森林病虫害防治领域，可智能监测无人机森林虫害图像中的枯死树，为森林有害生物的监测工作提供技术支撑。分别以湖北省受松材线虫与辽宁省受红脂大小蠹侵害的松林无人机图像为试验数据，首先使用线性谱聚类超像素分割算法将图像划分为多个超像素；然后基于枯死树木的颜色特征，初步提取可能为枯死树的超像素区域；最后基于枯死树木与其他干扰地物具有不同的纹理特征，计算超像素的区域密度和缝隙量，利用支持向量机对初步提取的超像素进行分类，从而检测出图像中的枯死树。基于线性谱聚类超像素和支持向量机的枯死树监测方法可有效排除与枯死树木颜色相近的其他干扰地物，较准确地提取出枯死树木。基于线性谱聚类超像素的枯死树监测方法能实现松林中枯死树的快速、准确检测及定位。

Applying the linear spectral clustering-based superpixel method to forest pest control enables intelligent monitoring of dead trees in UAV-captured forest pest images, providing technical support for forest pest monitoring operations. The experimental datasets used are UAV images of pine forests infested by pine wood nematode in Hubei Province and those infested by red turpentine beetle in Liaoning Province, respectively. First, the linear spectral clustering superpixel segmentation algorithm is employed to partition the images into multiple superpixel regions; then, based on the color characteristics of dead trees, superpixel regions that may correspond to dead trees are preliminarily extracted; finally, since dead trees exhibit distinct texture features from other interfering ground objects, the regional density and gap measure of each superpixel are calculated, and Support Vector Machine (SVM) is used to classify the preliminarily extracted superpixel regions, thereby detecting dead trees within the images. The dead tree monitoring method based on linear spectral clustering superpixels and SVM can effectively eliminate other interfering ground objects with colors similar to dead trees, and accurately extract dead trees. The dead tree monitoring method based on linear spectral clustering superpixels can achieve rapid and accurate detection and positioning of dead trees in pine forests.

搜集汇总

背景与挑战

背景概述

该数据集专注于利用线性谱聚类超像素和支持向量机技术，从无人机拍摄的林地图像中智能监测枯死树，特别针对湖北省和辽宁省受虫害侵害的松林。数据集提供了2.4 MB的文档数据，支持森林病虫害防治领域的研究，具有快速、准确的枯死树检测及定位能力。