林业化学除草技术的研究

从1974～1990年,在苗圃、幼林、造林地及防火线上筛选适用的除草剂,结果表明:松、杉苗圃以除草醚、果尔作土壤处理,防效达90％以上,高于1m的松、杉、阔叶树、果树以草甘膦作叶面喷洒,防效达85％以上;造林地以草甘膦、2,4-D丁酯、草甘膦+调节膦及威尔柏+草甘膦作叶面喷洒,防杂草、灌木效果达95％,防火线上以草甘膦与调节膦或威尔柏混用为好。在室内生测表明:草甘膦+调节膦有增效作用,草甘膦具有解偶联剂作用,抑制膜上ATP酶活性。     

外来入侵种及其对森林生态系统的威胁:概念和对策

外来入侵种问题越来越受到关注，其对经济、人类生活、环境和生物多样性的危害也被广泛认识。结合介绍外来入侵种的定义、危害性和入侵过程，对入侵种的生物学特征、易受入侵种危害的生态系统的特征和控制措施进行了归纳；结合国内外研究和管理现状，对我国的外来入侵种研究和管理对策提出了具体建议。     

Alley cropping and mulching withErythrina poeppigiana (Walp.) O. F. Cook andGliricidia sepium (Jacq.) Walp.: effects on maize/weed competition

The potential of allye cropping systems to sustain a high productivity with low external inputs and the reduction of maize/weed competition through weed suppression in different alley cropping and sole-cropped mulched systems was studied in Costa Rica at CATIE. Data were recorded eight years after establishment of the experiment. Plant residues ofErythrina poeppigiana trees (10 t/ha dry matter) planted at 6 by 3 m reduced weed biomass by 52%, whileGliricidia sepium trees (12 t/ha dry matter) planted at 6 by 0.5 m reduced weed biomass by 28%, in comparison to controls.Erythrina had a considerable impact on grass weeds, whileGliricidia reduced the incidence of some dicot weeds. Weed competition significantly reduced maize yield in all systems. Nevertheless weed suppression contributed to the higher maize grain yield underErythrina andGliricidia alley cropping of 3.8 t per hectare as opposed to the unmulched control yield of 2.0 t per hectare.     

The performance of the leaf mulches ofLeucaena leucocephala,Flemingia macrophylla andGliricidia sepium in weed control

The performance of the leaf mulches ofLeucaena leucocephala, Gliricidia sepium andFlemingia macrophylla in weed control has been tested in two trials. The length of the period during which a mulch layer yields significantly less weedbiomass compared to the control plots is called the ‘effective life-span’ of the mulch. Of the three mulch materials only that ofF. macrophylla shows promise in retarding weed development. In the second trialF. macrophylla leaf mulch was applied at rates of 3, 6 and 9 tons dry matter per ha. The effective lifespan of a mulch layer of 3 tons is between 12 and 13 weeks. The treatments 6 and 9 tons have effective life-spans of over 14 weeks. For moderate quantities (up to 5 tons of dry leaf mulch per ha) the effective life-span is estimated at about a 100 days. The value of mulching in weed control is limited to the control of weed species that multiply by seed. Regrowth originating from roots or stumps from former vegetation is unlikely to be checked by a mulch layer. The subject of this paper is part of a broader study, presently conducted in Southern Ivory Coast, on the production, agronomical value and use in alley-cropping systems of the species mentioned in the title of the paper. A final report is scheduled for publication in 1988.     

24%氨氯吡啶酸水剂防治林地阔叶杂草试验

选用24％氨氯吡啶酸水利，对林地一年生及多年生阔叶杂草、灌木进行防除试验。结果表明，试验剂量在1200～3600g ai／hm^2范围内时，对杂草的防效可达90％以上，对灌木防效达70％以上。该药剂对针叶用材林幼林慎用。     

森泰除草剂防除香椿幼林地杂草的试验研究

使用25%森泰水剂和5%森泰颗粒剂防除香椿幼林地的杂草研究试验结果表明:25%森泰水剂和5%森泰颗粒剂均能够有效防除香椿幼林地的五节芒、苦竹、刚竹、芒萁骨等多年生恶性杂草,防除效果均可达93%以上。使用25%森泰水剂除草比人工除草可减少用工56 1%～92 2%,降低除草成本36 1%～61%。同时,可根据幼林地杂草的种类、高度、盖度和水源等因素选择合适的森泰除草剂的剂型和剂量组合,以达到理想的防除杂草效果。     

Woody and herbaceous competition effects on the growth of naturally regenerated loblolly and shortleaf pines through 11 years

Four levels of vegetative competition were used to quantify the growth of loblolly and shortleaf pines (Pinus taeda L. and P. echinata Mill.) in naturally regenerated, even-aged stands on the Upper Coastal Plain of southeastern Arkansas, USA. Treatments included: (1) no competition control, (2) woody competition control, (3) herbaceous competition control, and (4) total control of nonpine vegetation. After pines became established from natural seeding, herbicides were used to control herbaceous plants for 4 consecutive years and woody plants for 5 consecutive years. Even though 89% of crop pines on untreated check plots were free-to-grow 11 years after establishment, crop pines on vegetation control plots were larger (P<0.001) in mean dbh, total height, and volume per tree. From age 5 through 11 years, crop pine diameter growth increased on woody control plots and decreased on herbaceous control plots because of hardwood competition in the latter treatment. At age 11, crop pine volume production averaged 207 m³/ha on total control plots, 158 m³/ha on herbaceous control plots, 130 m³/ha on woody control plots, and 102 m³/ha on untreated check plots.     

Effects of Gliricidia sepium (Jacq.) Steud and Leucaena leucocephala (Lam.) de Wit on growth and development of Imperata cylindrica (L.) Raeuschel.

Field studies carried out in a forest transition site on a tropical Alfisol in southwestern Nigeria, show that shading by gliricidia (Gliricidia sepium (Jacq.) Steud) and leucaena (Leucaena leucocephala (Lam.) de Wit) hedgerow species reduced density of speargrass (Imperata cylindrica (L.) Raeuschel) by 67% and 51%, respectively. Shoot biomass of speargrass decreased by 81% and 78% in gliricidia and leucaena hedgerows, respectively. Reduction in speargrass rhizome biomass in gliricidia plots was 96% while rhizome reduction in the leucaena plots was nearly 90%. Rhizome mortality was significantly higher in gliricidia plots than in leucaena and control plots. The bulk of rhizomes was found between 10 cm and 20 cm of soil depth but rhizome did not penetrate further than 30 cm depth in the Alfisol in which this study was carried out. Gliricidia was better than leucaena hedgerow species in suppressing speargrass.     

基于改进DeepLabv3+的水稻田间杂草识别方法

针对实际稻田环境中水稻与杂草相互遮挡、难以准确区分的问题,提出一种基于改进DeepLabv3+的水稻杂草识别方法。以无人机航拍的复杂背景下稻田杂草图像为研究对象,在DeepLabv3+模型的基础上,选择轻量级网络MobileNetv2作为主干特征提取网络,以减少模型参数量和降低计算复杂度;融合通道和空间双域注意力机制模块,加强模型对重要特征的关注;提出一种基于密集采样的多分支感受野级联融合结构对空洞空间金字塔池化模块(ASPP)进行改进,扩大对全局和局部元素特征的采样范围;对模型解码器部分进行改进。设置消融试验验证改进方法的有效性,并与改进前DeepLabv3+、UNet、PSPNet、HrNet模型进行对比试验。试验结果表明,改进后模型对水稻田间杂草的识别效果最佳,其平均交并比(MIoU)、平均像素准确率(mPA)、F1值分别为90.72%、95.67%、94.29%,较改进前模型分别提高3.22、1.25、2.65个百分点;改进后模型内存占用量为11.15 MB,约为原模型的1/19,网络推算速度为103.91 f/s。结果表明改进后模型能够实现复杂背景下水稻与杂草分割,研究结果可...     

基于改进YOLOv3网络模型的茶草位置检测算法

精准高效的茶草识别是智能茶园植保机械进行除草工作的关键。针对目前茶园除草智能化程度较低等问题,提出改进YOLOv3网络模型的茶草检测算法。首先,分季节和时间段,在多个茶叶品种的种植园中以自适应的距离和角度采集茶草混合图像并建立试验数据集。接着,使用K均值聚类算法重新设计先验锚框尺度。然后,以YOLOv3网络模型为基础,选取17×17的网格划分图像区域;采用残差网络（ResNet）作为主干网;加入过程提取层,增强草株检测性能。最后在原损失函数中引入广义交并比损失。通过消融试验和不同目标检测算法对比试验验证此改进算法对茶树与杂草的检测效果。试验结果表明,改进 YOLOv3网络模型对杂草的检测精确率和召回率分别为85.34%和91.38%,对茶树的检测精确率和召回率最高达到82.56%和90.12%;与原YOLOv3网络模型相比,检测精确率提高8.05%,并且每秒传输帧数达到52.83 Hz,是Faster R-CNN网络模型的16倍。这些数据说明所提算法在茶园复杂环境下,不仅对于茶树和杂草具有更好的识别效果,而且满足实时检测的要求,可以为智能茶园植保机械提供技术支持。