Productivity and costs of slash bundling in Nordic conditions

The number of slash bundlers and the volume of slash bundling have been rapidly increasing during the last few years in Finland. However, no comprehensive time or follow-up studies have been carried out on slash bundling technology in Finland or in any other country. Metsäteho Oy carried out studies on the productivity and costs of slash bundling in different Nordic recovering conditions. The study methods included both time and follow-up studies. Data were collected during the summer and winter period primarily in Norway spruce (Picea abies L. Karst.) dominated clear cutting sites. The bundling techniques performed by different types of bundler (Fiberpac 370, Timberjack 1490D, Pika RS 2000, Valmet WoodPac) were studied.

The average productivity of slash bundling was 18.1 bundles per operating (E₁₅, including delays shorter than 15 min) hour with the Timberjack 1490D and Fiberpac 370 bundlers in the follow-up study. The operator of the slash bundler had the greatest effect on the productivity of bundling. The prerequisite for increased bundling volumes is a reduction in the costs of the most expensive sub-stage of the bundling supply chain, i.e. bundling itself. This requires improved recovery conditions at bundling sites, increased bundling productivity, larger sized bundles, and the execution of bundling operations in two work shifts using an efficient bundler and effective operator working methods. Implementation of these development measures will bring the bundling supply chain up to a speed that makes it the most competitive supply chain for forest chips in terms of total supply costs for long-distance transportation distances of more than 60 km.     

基于森林火灾平台的红外图像识别技术研究

森林火灾探测一直是森林资源保护工作中的老大难问题。早期的火灾探测技术多是基于火灾的烟雾和温度特征的,但其判别标准过于单一,误报率较高。随着计算机技术的发展和红外探测器的发明,出现了基于红外图像处理的森林火灾防护技术,该技术主要对红外探测器获得的森林现场的红外图像进行一定的处理后识别图像中是否有火灾出现。本文的主要工作是进行红外图像增强算法的研究,通过对各种算法的描述和仿真实验结果分析,提出一种相对较适合森林背景红外图像的处理算法一将基于频域处理的提升小波变换与直方图修正结合起来的红外图像增强算法。     

Assessing terrestrial wildlife populations in the Toronto and Region Area of Concern

Beneficial use impairments (BUIs) under the Great Lakes Water Quality Agreement identify environmental issues requiring remedial action within the Great Lakes Areas of Concern (AOCs). We conducted this study to support the assessment of the wildlife component of BUI 3: degradation of fish and wildlife populations. We compared bird and amphibian (frogs and toads) data from the Toronto and Region Conservation Authority’s Terrestrial Long-term Monitoring Program in the Toronto and Region AOC to an adjacent, but otherwise similar, reference watershed, Duffins Creek. Twelve of 13 targets were met within the AOC for forest bird, wetland bird, meadow bird and amphibian populations based on averages of mean annual values at sites within the AOC that were within two standard deviations of averages at sites in the Duffins Creek reference watershed between 2008 and 2017. Even though wildlife populations within the AOC were within the normal range of variability expected from a reference watershed, they were often at lower levels than within the Duffins Creek reference watershed. In addition, forest bird and amphibian populations were negatively affected by urbanization within the AOC and meadow bird indices declined. We conclude that while wildlife populations within the AOC currently meet targets for BUI 3, they continue to be negatively impacted by numerous stressors that are primarily related to past and ongoing urbanization. Thus, continued restoration of wildlife habitat and protection of existing habitat within the AOC is highly recommended.     

Forest harvesting in rural properties: Risks and worsening to the worker’s health under the ergonomics approach

Through an ergonomic approach, this study aimed to evaluate whether workers involved in forest harvesting activities in rural properties are subject to the development of work-related diseases, as well as their consequences. A cross-sectional study with 267 forest workers in rural properties in the Brazilian Central region was used. The following evaluations were performed for each of the activities (felling, delimbing, tracing, manual extraction and manual loading): physical workload evaluation; biomechanical evaluation; risk of repetitive strain injuries/work-related musculoskeletal disorders; environmental noise and vibration factors, where relevant, and thermal overload through the Wetbulb Globe Temperature index. The results showed that in general the physical workload was very high, exceeding the cardiovascular load limits and adding to the need for work reorganization. All activities exposed workers to serious and imminent risks of developing injuries to the spine and lower limbs. Likewise, in general the risk of the appearance of RSI/WMSDs was very high for all activities evaluated. Such results are due to the combination of organizational work factors and factors of the work environment such as exposure to bad weather, uneven terrain, lifting and handling loads above tolerable limits, excessive noise and thermal overload. All of this allows to conclude that ergonomic risks and workers’ health hazards in forest harvesting in rural properties are latent and very worrying, and that these workers are exposed to a form of labor exploitation that invariably leads to physical and emotional exhaustion and therefore to their decreased labor capacity and useful working life.     

基于改进遗传算法和SVM的森林火灾视频目标鉴别

计算机视觉是目前计算机技术研究的热门课题之一,目标对象特征的选择和提取是计算机视觉的核心问题。在分析和研究火焰图像中火焰预判方法的基础上,提出了从纹理、动态、几何3个方面筛取火焰图像的组合特征,并且重点介绍了动态特征的提取方法。在火焰识别的特征选择方面做了较深入的研究,综合利用了火焰的内容特征,基于遗传算法提出了新的适应度函数,使得特征选择更加科学,而基于结构风险最小化理论的SVM在识别过程中可充分发挥优势,因此在实验中取得了理想的效果。     

基于面向对象的平潭岛大比例尺森林资源监测方法

利用高分辨率遥感影像对森林资源进行大比例尺动态监测可以有效提高林业部门对森林资源管理的时效性。以福州市平潭岛为研究区域,利用高分辨率的WorldView\|2遥感影像,结合大比例尺森林小班矢量图层,基于面向对象的分类思想,采用分层监督分类的方法,提取森林资源变化图斑,实现大比例尺森林资源的动态监测。此种方法有效利用了原始小班边界,快速提取了变化区域,总体分类精度达到了90.85%,表明利用该方法进行大比例尺森林资源变化区域提取是有效可行的。     

车载自组织网络分布式入侵检测算法研究

随着汽车产业的发展,汽车给我们带来便利的同时也给我们带来一系列的问题,例如交通拥堵、交通安全等。为了解决这些问题车载自组织网络应运而生。但是车载自组织网络中信息安全问题一直存在,采用分布式入侵检测模型和基于此模型下的改进随机森林分类算法可以很好地解决这个问题。     

Interactions of Impact Shock Waves in a Thin-Walled Explosive Container. II. Impact by a Cone-Nosed Projectile

Interaction of impact shock waves that could detonate an explosive (Composition B) confined in a thin-walled container impacted by a cone-nosed projectile is numerically studied, based on the Forest Fire explosive reaction rate model. The normal impact on the container by a small projectile with a conical nose is considered. Depending on the cone angle of the nose part of the projectile, the zone of interaction of initiating shock waves can be remote from the central axis of the impact. The off-the-central-axis detonation is interpreted from the viewpoint of different interaction modes in the explosive container, which are reflected from the cassette wall, change their directions, and superimpose onto each other, leading to explosive detonation.     

基于局部概率抽样的标签噪声过滤方法

分类学习任务中,在获取数据的过程中会不可避免地产生噪声,特别是标签噪声的存在不仅使得学习模型更复杂,而且容易造成过拟合并导致分类器泛化能力的下降。标签噪声过滤算法虽然在一定程度上可以解决上述问题,但是仍然存在噪声识别能力较差、分类效果不够理想以及过滤效率低等问题。针对这些问题,提出一种基于标签置信度分布的局部概率抽样方法来进行标签噪声过滤。首先利用随机森林分类器对样本的标签进行投票,从而获取每个样本的标签置信度;然后根据标签置信度的大小,将样本划分为易识别样本和难识别样本;最后分别采用不同的过滤策略对样本进行过滤。实验结果表明,在标签噪声存在的情况下,所提方法在大多数案例上能够保持较高的噪声识别能力,并且在分类泛化性能上也具有明显优势。     

基于表示学习和深度森林的长链非编码RNA编码短肽预测模型

长链非编码RNA（lncRNA）中的小开放阅读框（sORFs）能够编码长度不超过100个氨基酸的短肽。针对短肽预测研究中lncRNA中的sORFs特征不鲜明且高可信度数据尚不充分的问题,提出一种基于表示学习的深度森林（DF）模型。首先,使用常规lncRNA特征提取方法对sORFs进行编码;其次,通过自编码器（AE）进行表示学习来获得输入数据的高效表示;最后,训练DF模型实现对lncRNA编码短肽的预测。实验结果表明,该模型在拟南芥数据集上能够达到92.08%的准确率,高于传统机器学习模型、深度学习模型以及组合模型,且具有较好的稳定性;此外,在大豆与玉米数据集上进行的模型测试中,该模型的准确率分别能达到78.16%和74.92%,验证了所提模型良好的泛化能力。