基于阈值分割的松材线虫病树计数与定位

松材病虫病是最紧迫的威胁之一，最近对中国针叶林造成了严重破坏，为了阻止松材病虫病的快速传播，必须在早期准确检测和计数受感染的树木。提出了一种基于密度图的方法，用于从高分辨率航拍图像中估计具有松材病虫病的树木数量，方法引入了阈值分割、形态处理和超像素技术，以最大限度地减少由于建筑物和岩石等相似背景物体引起的误差。实验表明，所提出的方法优于密度学习 算法，平均绝对误差、均方根误差和绝对误差方差分别从21.3、22.2和127.2减少到8.0、11.7和59.6，计数准确率从54.7%提高到81.3%。

Counting and Location of Trees with Pine Wilt Disease Based on Threshold Segmentation

Pine wilt disease （PWD） is one of the most urgent threats, which has caused severe damage to conifer forests in China recently. To stop a rapid PWD spread， it is imperative to accurately detect and count infected trees at its early stage. This work presents a density-map based approach to estimate the number of the trees with PWD from a high-resolution aerial imagery. Threshold segmentation， morphology processing， and superpixels technology are introduced to minimize the errors due to similar background objects such as buildings and rocks. Experiments show that the proposed approach performs better than the density learning algorithm， and mean absolute error， root mean square error and variance of absolute error decrease from 21.3， 22.2， and 127.2 to 8.0， 11.7， and 59.6 respectively， and counting accuracy increases from 54.7% to 81.3%.