早熟梨新梨7号胚培养直接成苗的影响因素探讨

以早熟梨新梨7号为试验材料,在离体条件下研究了基本培养基种类、激素配比、胚龄大小及低温预处理等因素对早熟梨幼胚离体培养的影响。结果表明:MS培养基诱导新梨7号胚萌发和子叶转绿方面与White培养基相似,而在培养后期促进生根和成苗率方面优于White培养基;适宜培养基激素浓度比例为2.0 mg/L 6-BA+0.1 mg/L NAA;以胚龄85 d以后的幼胚培养较易成功。低温预处理对促进65 d和75 d胚龄的幼胚萌发和成苗效果优于85 d以后梨胚,且低温处理对胚萌发、子叶转绿的促进作用大于对生根率和成苗率的促进作用。     

基于无人机视频影像的油菜苗检测与计数

针对油菜生长早期传统人工苗情调查方法效率低、主观意识强,不能满足大面积或经常性高精度苗期调查作业需求的问题,该研究基于无人机影像及机器学习技术,提出一种油菜苗视频流检测模型及计数方法。通过对YOLO系列基础模型添加多头自注意力,用BasicRFB（basic receptive field block）模块替换原有的空间池化结构（spatial pyramid pooling-fast,SPPF）模块,并对Neck部分添加一维卷积及更换下采样方式等,进一步结合DeepSORT（deep simple online and real-time tracking）算法和越线计数技术实现对油菜苗的数量统计。算例测试结果表明,改进后YOLOv5s的交并比阈值0.50的平均精度均值达到93.1%,交并比阈值0.50～0.95的平均精度均值达到了67.5%,明显优于Faster R-CNN、SSD和YOLOX等其他经典目标检测算法,交并比阈值0.50的平均精度均值分别高出14.82、26.37和3.3个百分点,交并比阈值为0.50～0.95范围下的平均精度均值分别高出25.7、33.9和6.7个百分点。油菜苗计数试验结果表明,离线视频计数时,在合理的种植密度区间内,所提算法的油菜苗计数精度平均达到93.75%,计数效率为人工计数的9.54倍;在线实时计数时,在不同天气情况下,计数平台的油菜苗计数精度最大相差1.87个百分点,具有良好的泛化性,满足油菜苗计数实时性要求。     

水下底栖生物轻量化目标检测算法YOLOv7-RFPCW

针对水下底栖生物检测模型参数量过多,计算量过大,同时水下环境复杂容易造成错检和漏检,影响模型在水下底栖生物检测任务中的准确性的问题。提出了一种水下底栖生物轻量化检测算法YOLOv7-RFPCW。对YOLOv7网络重新设计轻量级网络结构,降低了特征提取网络的参数量和计算量,减少模型体积。设计了P-ELAN和P-ELAN-W模块,进一步轻量化特征提取网络;针对水下图像颜色失真,目标的空间位置不准确的问题,采用CBAM注意力模块加强特征融合,减少信息丢失,以更好地适应特殊的水下环境;针对水下目标容易出现形状变形的问题,使用WIOU-V3损失函数替换默认的CIOU损失函数,提高水下底栖生物检测的鲁棒性。试验结果显示,改进后的模型YOLOv7-RFPCW的参数量和计算量分别减少了75.9%和30.7%,模型体积减小了75.3%,精度提升了1.9个百分点。这一综合性的提升兼顾了轻量化和精度,为在水下环境中部署提供了可行的解决方案,验证了所提出的改进算法能胜任水下底栖生物检测任务。     

基于改进YOLOv7的杨梅树单木检测

为了快速检测和统计杨梅树的数量,该研究提出了一种基于改进YOLOv7的杨梅树单木检测模型：YOLOv7-ACGDmix。首先,对YOLOv7的可扩展高效长程注意力网络（extended-efficient long-range attention networks, E-ELAN）进行改进,通过融合兼具卷积和注意力机制优势的ACmix（a mixed model that enjoys the benefit of both self-attention and convolution）结构得到AC-E-ELAN模块,提升模型的学习和推理能力,引入可变形卷积（deformable convolutional networks version 2, DCNv2）结构得到DCNv2-E-ELAN模块,增强模型对不同尺寸目标的提取能力;其次,采用内容感知特征重组（content-aware reassembly of features, CARAFE）上采样模块,提高模型对重要特征的提取能力;然后,在主干和头部网络部分添加全局注意力机制（global-attention mechanism, GAM）,强化特征中的语义信息和位置信息,提高模型特征融合能力;最后,采用WIoU（wise intersection over union）损失函数减少因正负样本数据不平衡造成的干扰,增强模型的泛化性。在公开数据集上的试验结果表明,YOLOv7-ACGDmix模型的精确率达到89.1%,召回率达到89.0%,平均精度均值（mean average precision, mAP）达到95.1%,F₁-score达到89.0%,相比于原YOLOv7模型分别提高1.8、4.0、2.3和3.0个百分点。与Faster R-CNN、SSD、YOLOv8模型相比,改进模型的平均精度均值（mAP_0.5）分别提高了9.8、2.2、0.7个百分点。实地采集杨梅树样本数据的检测精确率87.3%、召回率85.7%。试验表明,改进模型为基于无人机影像的杨梅树单木检测提供了一种有效的解决方案,对果园精准管理的发展具有重要意义。     

Prokaryotic Expression and Polyclonal Antibody Preparation of African Swine Fever Virus Georgia 2007/1 Strain CD2v Protein

The study was aimed to express the EP402R gene of African swine fever virus (ASFV) Georgia 2007/1 strain via prokaryotic expression system,obtain the recombinant CD2v protein,and prepare polyclonal antibodies against the purified recombinant CD2v protein.After codon optimization,ASFV EP402R full-length gene was linked into pET-28a(+) expression vector to construct prokaryotic recombinant expression plasmid.After induction by 1 mmol/L IPTG at 16 ℃ for 12 h,the recombinant protein was identified by SDS-PAGE and Western blotting.The purified recombinant CD2v protein was used as immunogen to prepare mouse anti-CD2v polyclonal antibodies.The antibody titer was measured by indirect ELISA and the specificity was further analyzed by indirect immunofluorescence assay (IFA) and Western blotting.The results showed that ASFV EP402R gene was successfully cloned into pET-28a(+),and pET-28a-EP402R was obtained.The recombinant plasmid was transformed into E.coli BL21(DE3) for expression,the recombinant protein was expressed mainly in the form of inclusion bodies,with molecular mass at about 47 ku,while some of the recombinant protein could also exist in a soluble form.Western blotting results showed that the purified protein had good immunoreactivity.The indirect ELISA result showed that the polyclonal antibodies had a high titer of 1:512 000,IFA and Western blotting results indicated that it could specifically recognize recombinant CD2v protein.These results confirmed the recombinant CD2v protein expressed via prokaryotic system had good immunogenicity,and the prepared polyclonal antibodies had high titer and specificity.This research provided technical support for further study of ASFV EP402R biological function,as well as its gene-deletion based vaccine development.     

Expression and functional analysis of chemokine receptor 7 in canine lymphoma cell lines

C-C chemokine receptor 7 (CCR7) contributes to cell homing to lymph nodes (LNs). Recent studies reported that CCR7 is also expressed in tumor cells, which correlates with LN metastasis in various cancers. However, the expression of CCR7 in tumor cells is unknown in dogs due to the lack of appropriate antibodies. In the present study, a fusion protein of C-C chemokine ligand 19 (CCL19) was employed as an alternative method to CCR7 antibodies. The fusion CCL19 protein specifically detected CCR7 expressed in canine lymphoma cell lines, which showed active chemotaxis to both canine and mouse ligands. The present study will help further research on the involvement of canine CCR7 in LN metastasis.     

猪轮状病毒重组VP7蛋白抗原间接ELISA诊断方法的建立

本研究以纯化的原核表达的猪轮状病毒VP7抗原表位区域为抗原,建立了检测猪轮状病毒抗体的间接ELISA诊断方法。特异性试验表明,该抗原与其他7种常见猪病病毒（TGEV、PEDV、CSFV、PCV2、PRRSV、PPV、PrV）的阳性血清不发生交叉反应,批内和批间重复性试验的变异系数均小于10％;对来自不同猪场的血清的检测结果表明,该ELISA方法与中和试验检测结果符合率达94．8％。本试验建立的ELISA诊断方法具有良好的重复性、敏感性和特异性,为PRV的快速诊断、免疫猪群抗体监测和轮状病毒流行病学调查提供了一种快速、简便的血清学诊断方法。     

苯并噻二唑甲酸三氟乙酯诱导仙客来抗枯萎病的研究

为了明确新型化合物苯并噻二唑甲酸三氟乙酯对土传病害的诱导抗病作用,本文研究了该化合物诱导仙客来对枯萎病的抗病效果。通过叶面喷施苯并噻二唑甲酸三氟乙酯可诱导仙客来对枯萎病的抗病效果,在仙客来移栽缓苗后7d叶面喷施该化合物,不同浓度对枯萎病的抗病效果不同,其中100mg.L-1的处理,在诱导后25、30和35d的抗病效果分别为82.90%、58.59%和64.17%;在仙客来成株枯萎病发病初期用该化合物诱导,抗病效果仍可达63.91%,显著高于噁霉灵处理（47.63%）。该研究将为寻求通过叶面喷雾施药防治园艺作物土传病害的新途径起到参考作用。     

水动冷却塔直驱混流式水轮机转轮数值模拟优化

针对比转数仅有54的水动冷却塔直驱混流式水轮机,选取影响水轮机转轮效率的6个重要水力参数β1,β2,b0/D1,Φ/D1,Z,a/D1,每个因素取2个水平,设计L8（2^7）正交表,生成8个设计方案.通过CFD软件分别对8个方案的水轮机进行全流道数值模拟,采用理论分析和CFD技术分析转轮水力损失特性,对比其结果、优化转轮参数,再将优化后方案进行数值模拟和外特性试验,对比得到的结果,表明使用正交化方法结合CFD技术可在较少的试验方案下有效地改进水轮机叶轮水力性能,达到水轮机优化设计的目的.在设计流量点Q为5 000 t/h处,优化后的模型转轮输出功率为156.57 kW,水轮机效率达到86.51%;数值模拟效率比试验效率低0.11%,实测结果的基本趋势与数值模拟结果相一致.     

基于改进YOLO v4的单环刺螠洞口识别方法

针对养殖池塘内单环刺螠自动采捕和产量预测应用需求,提出一种基于深度学习的单环刺螠洞口识别方法,以适用于自动采捕船的嵌入式设备。该方法通过将YOLO v4的主干网络CSPDarkNet53替换为轻量型网络Mobilenet v2,降低网络参数量,提升检测速度,并在此基础上使用深度可分离卷积块代替原网络中Neck和Detection Head部分的普通卷积块,进一步降低模型参数量;选取带色彩恢复的多尺度视网膜(Multi-scale retinex with color restoration,MSRCR)增强算法进行图像增强;利用K-means++算法对数据集进行重新聚类,对获得的新锚点框尺寸进行线性缩放优化,以提高目标检测效果。在嵌入式设备Jetson AGX Xavier上部署训练好的模型,对水下单环刺螠洞口检测的平均精度均值（Mean average precision,mAP）可达92.26%,检测速度为36f/s,模型内存占用量仅为22.2MB。实验结果表明,该方法实现了检测速度和精度的平衡,可满足实际应用场景下模型部署在单环刺螠采捕船嵌入式设备的需求。