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。     

40%砜吡草唑悬浮剂的除草活性及对花生的安全性评价

为明确砜吡草唑在花生田应用的可行性, 采用整株生物测定法测定40%砜吡草唑悬浮剂对马唐Digitaria sanguinalis、狼尾草Pennisetum alopecuroides、稗Echinochloa crus-galli、藜Chenopodium album、反枝苋Amaranthus retroflexus及马齿苋Portulaca oleracea的除草活性, 评价药剂对花生的安全性, 并利用田间小区药效试验进行验证。结果表明:40%砜吡草唑悬浮剂对6种供试杂草均表现出极高活性, GR 50为0.76~2.90 g/hm 2, 在13.32 g/hm 2剂量下, 对供试杂草防效均达100%。40%砜吡草唑悬浮剂对3种禾本科杂草、3种阔叶杂草的毒力分别是960 g/L精异丙甲草胺乳油的5.60~9.30倍和22.16~25.24倍。在盆钵底部渗灌、喷淋浇水2种灌溉条件下, 40%砜吡草唑悬浮剂对花生的GR 10分别为212.03、87.68 g/hm 2, 剂量越高, 对花生生长的抑制愈明显, 对花生安全性欠佳。在盆钵底部渗灌条件下, 砜吡草唑在花生和供试杂草间的选择性指数为29.45~142.30, 具有较高选择性。田间试验表明, 40%砜吡草唑悬浮剂施用剂量120~240 g/hm 2时, 对杂草的总体防效在92.6%以上, 花生增产率21.09%~22.71%, 杂草防效及花生产量均显著高于960 g/L精异丙甲草胺乳油720 g/hm 2处理, 所有药剂处理未见药害症状。综上, 40%砜吡草唑悬浮剂可作为花生田低量、高效化学除草的备选药剂, 但需关注砜吡草唑过量施用及降雨积水等因素导致的花生安全性问题。     

辽宁省大豆田杂草发生与危害

采用倒置"W"九点取样法调查辽宁省大豆田杂草,明确了辽宁省大豆田杂草有19科35种,其中阔叶杂草占74.29%,禾本科杂草占14.29%,其他杂草占25.71%;一年生杂草占54.29%,多年生杂草占31.43%。相对多度达10%以上的杂草依次为藜、鸭跖草、稗草、苘麻、反枝苋、红蓼、马唐、铁苋菜、马齿苋和山苦菜。其中藜和鸭跖草相对多度在30%以上,为辽宁省大豆田杂草优势种群。     

Allelopathy in crop/weed interactions--an update

Since varietal differences in allelopathy of crops against weeds were discovered in the 1970s, much research has documented the potential that allelopathic crops offer for integrated weed management with substantially reduced herbicide rates. Research groups worldwide have identified several crop species possessing potent allelopathic interference mediated by root exudation of allelochemicals. Rice, wheat, barley and sorghum have attracted most attention. Past research focused on germplasm screening for elite allelopathic cultivars and the identification of the allelochemicals involved. Based on this, traditional breeding efforts were initiated in rice and wheat to breed agronomically acceptable, weed-suppressive cultivars with improved allelopathic interference. Promising suppressive crosses are under investigation. Molecular approaches have elucidated the genetics of allelopathy by QTL mapping which associated the trait in rice and wheat with several chromosomes and suggested the involvement of several allelochemicals. Potentially important compounds that are constitutively secreted from roots have been identified in all crop species under investigation. Biosynthesis and exudation of these metabolites follow a distinct temporal pattern and can be induced by biotic and abiotic factors. The current state of knowledge suggests that allelopathy involves fluctuating mixtures of allelochemicals and their metabolites as regulated by genotype and developmental stage of the producing plant, environment, cultivation and signalling effects, as well as the chemical or microbial turnover of compounds in the rhizosphere. Functional genomics is being applied to identify genes involved in biosynthesis of several identified allelochemicals, providing the potential to improve allelopathy by molecular breeding. The dynamics of crop allelopathy, inducible processes and plant signalling is gaining growing attention; however, future research should also consider allelochemical release mechanisms, persistence, selectivity and modes of action, as well as consequences of improved crop allelopathy on plant physiology, the environment and management strategies. Creation of weed-suppressive cultivars with improved allelopathic interference is still a challenge, but traditional breeding or biotechnology should pave the way.     

无人机播前喷施除草剂防除棉田杂草效果评价

【目的】 分析不同喷药处理喷施33%二甲戊灵乳油对棉田杂草防效及棉花生长发育的影响,为棉田无人机播前喷施除草剂提供理论依据。【方法】 在新疆机采棉种植模式下,选择土壤封闭剂二甲戊灵为供试除草剂,设计播前无人机3个不同浓度喷药,车载喷药机喷药和不喷药5个处理,分析不同处理对棉田杂草防效及棉花生长发育的影响。【结果】 无人机喷施二甲戊灵进行棉田播前土壤处理,处理2(3 450 mL/hm²时)控草效果与车载喷药机处理相当,防治效果均达到94.00%,且显著好于不施药效果,出苗率均达59.71%,与不施药处理差异不显著。增加药剂浓度1 200 mL/hm²对棉花出苗有影响,出苗率降低7.54%,但差异不显著。无人机喷药处理株高、叶龄、地上部干物质重、叶面积指数与车载喷药机喷药处理间差异不显著,显著高于不喷药处理。【结论】 无人机喷施播前二甲戊灵浓度3 450 mL/hm²时与车载喷药机喷施浓度3 450 mL/hm²防草效果和棉花生长发育相当,可在大田推广使用。     

21%氟磺·烯草酮油悬浮剂对绿豆田一年生杂草防效研究

为寻找绿豆田适宜除草剂,于绿豆1~2片复叶期,一年生禾本科杂草3~5叶期进行21%氟磺·烯草酮油悬浮剂药效评价试验。结果表明:21%氟磺·烯草酮油悬浮剂对绿豆田稗草、狗尾草、藜和反枝苋等一年生杂草具有较高防除效果,对绿豆生长安全。在生产中推荐应用剂量为1 800g·hm-2,喷药量为450L·hm-2。     

Effects of mechanical and chemical methods on weed control,weed seed rain and crop yield in maize,sunflower and soyabean

Eight field experiments with maize (Zea mays L.), sunflower (Helianthus annuus L.) and soyabean (Glycine max (L.) Merr.) were carried out in central Italy in order to evaluate the effects of mechanical and chemical methods (spring-tine harrowing, hoeing, hoeing-ridging, split-hoeing, finger-weeding, herbicides in the row + inter-row hoeing, herbicides broadcast) on weed control, weed seed rain and crop yield. The choice of chemical and mechanical treatments in maize and soyabean compared to sunflower, required to be managed more carefully in order to maximize the weed control reducing yield losses. A global rating of weed control methods, based on their weed control efficacy, was obtained as useful means to assist farmers and technicians to choose the more appropriate weed control method. The combination of herbicides intra-row and hoeing inter-row gave best efficacy (on average 99% of weed control), with a 50% reduction in the chemical load in the environment. Hoeing-ridging gave good results, both inter- and intra-row (on average 93% of weed control); this method was also effective in reducing competitive ability and seed production of uncontrolled weeds. Split-hoeing or finger-weeding showed some limitations giving satisfactory results only when combined. Harrowing gave lowest weed control, although when combined to other mechanical methods, can help achieve a better efficacy.     

控草和施肥对桉树生长影响的试验研究

为探索永安地区桉树速生丰产的营林抚育技术,通过采用随机组合,对2010年造林的巨桉桉树林设计了控草和施肥量不同的抚育组合：试验结果表明,对桉树林实行控草和施放肥料有利于促进林木生长,提高林分生物量和生产力;同时通过经济效益分析表明,在永安地区桉树林的最佳施肥量为底肥施放0.5kg桉树专用肥＋0.5kg钙镁磷（永安地区土壤缺磷元素）,当年追施一次0.25kg桉树专用肥,次年追施一次0.25kg桉树专用肥;控草方式以人工控草为主,有条件的林地实行化学除草,特别是五节芒等恶性杂草较多的山场优先考虑进行化学除草。     

威尔伯在幼林抚育中的应用研究

以红松、红皮云杉、长白落叶松造林地为对象,利用化学制剂威尔伯进行幼林抚育试验,结果表明:450 ml·hm-2剂量除草效果最差,仅有12%,在生产上无应用价值;450 ml·hm-2剂量有一定的除草效果,但不明显,持效期也短;1 125 ml·hm-2剂量无论除草效果还是除草速率都优于450 ml·hm-2和750 ml·hm-2。长白落叶松对威尔伯敏感,不能应用。化学除草抚育较人工抚育更能促进幼树生长,而且费用较低。