基于位姿状态的林区道路视觉导航

在森林抚育中,为实现林区车辆的精确自主导航,提出一种基于位姿状态的林区道路视觉导航方法。根据林区道路的特点,采用统计学原理选取合适的色差备选通道,利用二维Otsu原理初步分离道路和背景区域,依据真阳率和假阳率定量评价方法选取2B?G为最佳色差通道,因此采用基于2B?G通道的二维Otsu图像分割方法。分析了道路的左右边缘特征,建立基于边缘不完整的位姿状态获取示意图,将选定的道路区域分块求取中心点,计算出试验平台的位姿信息。同时,引入误差校正模型对位姿计算值进行修正,最后选取纯追踪导航方法进行实车试验。试验结果表明,在机器视觉系统引导下,试验平台在不同状态下均能较好地收敛至期望导航线。当速度为0.5 m/s时,在3种初始状态下进行直线导航试验,其全程平均横向偏差为4.6~5.9 cm,平均航向偏差为3.5°~4.0°,稳态航向偏差为1.9°~2.4°,稳态横向偏差为1.6~2.7 cm,试验平台沿林区期望导航线自主行驶的平均相对误差均值为5.2%,说明导航质量具有较好的鲁棒性,满足林区自主导航的作业要求。     

基于UWB传感器的树木定位研究

树木位置是森林样地调查中重要的调查因子,由于林下环境复杂,树木定位的技术方法面临很大的挑战。针对当前树木位置测量技术方法存在的效率低、携带不便、成本高、不适用于非视距测量等问题,提出了一种基于UWB传感器并结合海拔计来获取树木位置信息的方法,并自行设计了树木位置采集装置。选取8个树种多样、坡度不一的样地进行试验。首先,在每个样地中布置四个UWB基站;然后,携带内置UWB定位标签的装置采集每棵树木与各个基站之间的距离;最后,将所采集的数据进行MATLAB仿真,研究不同大小权重值对定位精度的影响。结果表明:存在一个最优权重值,可使平均定位误差下降3.31cm;相较其它四边形定位算法复杂度降低,更适用于枝叶茂密、地形复杂等非视距情况下的树木定位。     

基于机器视觉的森林防火监测技术初步研究

简要介绍了机器视觉图像处理技术,以实例说明了机器视觉技术在森林防火中的应用.     

基于立体视觉测量的移动机器人导航技术研究

立体视觉测量是机器视觉的重要分支,已经在视觉检测、医学辅助治疗和月球车导航等领域获得了应用.本文对于移动机器人立体视觉导航的全过程进行研究,分析立体视觉测量信息到真实三维景物信息的计算、人工势场法进行路径规划这两项关键技术.实验结果表明,本文提出的方法可以为移动机器人准确导航.     

基于机器视觉的树木图像实时采集与识别系统

研究了树木图像的实时采集与识别技术 ,以及树木图像处理及分析软件系统的开发 ,并自行设计制作了室内模拟实验系统。树木图像处理技术包括图像的获取、图像的增强、图像边缘检测、图像分割和图像的特征提取与识别。本文的研究能够为分离施药目标与非施药目标提供依据 ,有助于精确农药喷雾的实现。     

基于Halcon的相机标定方法研究与精度验证全文替换

机器视觉已经在现代工业测量领域中得到了广泛应用,而相机标定直接影响机器视觉的测量精度。通过搭建相机视觉实验平台,在分析了单目相机标定原理基础上,以GB050-2-7×7型圆点标定板为实验对象,利用Halcon软件对圆点标定板进行相机标定实验并验证测量精度。当采集的标定板图像数量为16时相机标定精度最优,此时相机的平均偏差为0.0707mm。在进一步测量实验中测量出标定板上7个特征圆点的直径平均值为1.9491mm,6个特征圆点之间的间距的平均值为2.0632mm,测量精度都高于0.1mm,验证了Halcon软件标定结果的测量准确性。     

基于机器视觉的木窗双端铣削加工尺寸测量方法

木窗是一种以木材或木质复合材料为主要构件的门窗产品,具有良好的生态性能和美观效果,适用于多种建筑形式和风格,其中木窗尺寸是衡量木窗加工是否合格的重要指标。对于传统木窗双端铣削加工中人工测量尺寸方式存在的精度低、效率低等问题,提出一种基于机器视觉的木窗双端铣削加工尺寸测量方法,以期提高尺寸测量精度及加工效率。该方法针对木窗厚度引起的透视效应,提出一种物平面提升法,以消除透视投影带来的误差。首先对木窗图像采取灰度化、平滑去噪、图像增强及轮廓分割等举措,完成图像预处理,提取出木窗内外轮廓区域。对轮廓区域应用Canny算子获取木窗像素级轮廓。通过优化的Zernike矩亚像素边缘提取算法对木窗像素级边缘进行更精确的定位,得到亚像素级轮廓坐标。通过最小二乘法联合RANSAC算法对亚像素轮廓坐标进行拟合,得到拟合轮廓及角点坐标,并使用透视矫正模型计算出木窗尺寸。实验利用3种厚度规格相同但尺寸不同的松木材质矩形木窗,分别测量其内框和外框的边框尺寸及对角线尺寸,并与对应的实际物理尺寸对比,验证了所提木窗尺寸测量方法的检测精度。研究结果表明,所提方法与实际物理尺寸值相比,其绝对误差范围在±0.12 mm之...     

基于视觉变换网络的自监督森林火灾识别

【目的】森林火灾识别是避免森林火灾大面积蔓延的一项重要研究。随着深度学习的快速发展,基于卷积神经网络的模型因其在图像识别领域的优异表现,被广泛应用到森林火灾识别任务当中。然而,基于卷积神经网络的方法通常在标签数据不充分时,难以取得令人满意的森林火灾识别结果。【方法】本研究提出了一种基于视觉变换网络的自监督森林火灾识别模型（Self supervised forest fire identification model based on visual transformation network）,来提高模型在标签稀缺情况下的森林火灾识别精度。具体来说,该模型采用视觉变换网络作为主干网络,通过视觉变换网络中的多头自注意力机制来捕获森林火灾图像的全局信息特征。并且引入自监督学习中的图像重建任务来辅助模型训练,从而减少模型对标签数据的依赖。模型通过对掩盖图像的特征恢复和重建学习相关语义信息。同时,本研究还提出了一种基于傅里叶低频混合变换的数据增强方法来提高模型的泛化性和鲁棒性。【结果】通过开展详细的试验来验证模型的有效性,结果表明,与其他常见的网络模型相比,FFDM模型在森林火灾识别任务中取...     

Assessing carbon sequestration in plantation forests of important conifers based on the system of permanent sample plots across Taiwan

ABSTRACT

Plantation forests play a critical role in forest management due to their high productivity and large contribution to carbon sequestration (CSE). The purpose of this study was to assess the CSE of plantations containing four important conifer species distributed across Taiwan, namely, the China fir (Cunninghamia lanceolata), Japanese cedar (Cryptomeria japonica), Taiwania (Taiwania cryptomerioides) and Taiwan red cypress (Chamaecyparis formosensis). Data regarding the plantations were obtained from a survey of permanent sample plots (PSPs). We used these data to calculate the CSE in each PSP and adopted CSE_mean and CSE_period as indicators to assess the CSE of the four conifers. According to the CSE_mean obtained from analysis of variance and the least significant difference method, two groups were identified among these four conifers: the Japanese cedar (4.03 Mg ha^?1 yr^?1) and Taiwania (3.52 Mg ha^?1 yr^?1) yielded higher CSE_mean values and the China fir (1.79 Mg ha^?1 yr^?1) and Taiwan red cypress (2.36 Mg ha^?1 yr^?1) yielded lower CSE_mean values. The same patterns were observed in the CSE_period values; however, no significant difference in CSE_period was observed between Taiwan red cypress and either of the two groups. Therefore, Japanese cedar and Taiwania have high CSE potential among conifers.     

基于GIS的桃山林场森林景观格局研究

以河北省木兰围场国有林场管理局桃山林场施业区内的森林资源为对象,利用GIS技术对其森林景观及分布规律进行分析,结果表明:桃山林场森林景观以混交林为主体,其面积占景观总面积的60.42%,斑块数量占景观斑块总数量的29.01%,优势度达到3.474 5。在斑块规模中,以混交林斑块最大,纯林和荒山荒地也有较大规模,疏林地也有相当规模分布,未成林造林地、宜林沙荒地、灌木林、耕地和其它用地景观规模最小;各类斑块的极差值均较大,斑块大小的分化较强烈。混交林的形状最复杂;其次是未成林造林地、纯林、荒山荒地、灌木林、宜林沙荒地、耕地和疏林地;其它用地的形状规整,边缘形状简单,说明这类景观斑块受人为干较严重。     

Effects of density and structure on production in the communal forests of the Mexican Sierra Madre Occidental

The Sierra Madre Occidental in the Mexican State of Durango, is home to about 5 million ha of species-rich forest ecosystems. Local communities depend on these forests for their livelihood. The preservation of the species richness and productivity of this unique resource requires improved understanding of the effects of forest structure and density on production. The community structure was analysed for three species groups (all species; 12 species of Pinus; and 16 species of Quercus). Based on data gathered in 580 remeasured observational field plots, the main results of this study were as follows: (1) two distinct clusters were identified in each species group, (a) the dominant species with a smaller range of growth rates, and (b) the remaining species which have a smaller maximum size and a greater variation of growth rates. (2) Density (expressed as basal area per hectare) and the percentage contribution to the total basal area of all pine species have a significant effect on forest production (expressed as volume increment per hectare per year). (3) Potential production is surprisingly high (in excess of 20 m³ ha^?1 y^?1 for some stands) in these forests, but the high potential cannot be realised in many areas because of reduced basal areas caused by overharvesting. (4) The available evidence does not confirm the findings of a recent study that suggests a continuous increase in production with increasing richness. Our observations show that production attains a potential maximum at intermediate levels of richness (between 7 and 9 species per 0.25 ha) and declines with further increases in species richness. This study represents a comprehensive analysis of the effects of density and structure on the community forests of the Sierra Madre Occidental based on a large set of permanent field plots where all trees have been mapped.     

Comparison of temporal and spatial changes in three major tropical forests based on MODIS data

Numerous studies have shown that intact tropical forests account for half of the total terrestrial sink for anthropogenic carbon dioxide.Here,we analyzed and compared changes in three main tropical forest regions from 2000 to 2014,based on time-series analysis and landscape metrics derived from moderate-resolution imaging spectroradiometer data.We examined spatialpattern changes in percentage of tree cover and net primary production(NPP)for three tropical forest regions—Amazon basin,Congo basin,and Southeast Asia.The results show that:the Amazon basin region had the largest tropical forest area and total NPP and a better continuity of TC distribution;the Southeast Asia region exhibited a sharp decrease in NPP and had comparatively separate spatial patterns of both TC and NPP;and the Congo basin region exhibited a dramatic increase in NPP and had better aggregation of forest NPP distribution.Results also show that aggregative patterns likely correlate with high NPP values.     

The lateral spread of tree root systems in boreal forests: Estimates based on N uptake and distribution of sporocarps of ectomycorrhizal fungi

In nutrient poor environments, such as boreal forests, many of the most important interactions between plants take place belowground. Here, we report the results of two approaches to obtain estimates of the lateral spread of tree roots.