基于LightGBM和处方数据的番茄病害诊断方法

为高效地挖掘植物病害处方数据并辅助精准诊断,以番茄病毒病、番茄晚疫病、番茄灰霉病3种病害为研究对象,构建基于贝叶斯优化LightGBM的番茄病害智能诊断模型,探索作物病害处方数据挖掘及其精准诊断。重点对处方原数据（文本数据标签和One-hot编码等）进行预处理,以基于Wrapper的递归特征消除法进一步提取作物病害处方数据的特征;利用基于LightGBM算法构建番茄病害诊断模型,并与K近邻（KNN）、决策树（DT）、支持向量机（SVM）、随机森林（RF）、梯度提升决策树（GDBT）、AdaBoost和XGBoost常见机器学习模型运行结果进行比较分析并进行优化;设计基于LightGBM模型的Android手机端植物医生病害诊断APP。实验结果表明,基于贝叶斯优化的LightGBM模型综合诊断准确率可达到89.11%,比其他7种机器学习模型的诊断准确率平均高3.65个百分点;同时特征选择后的LightGBM模型在保证模型准确率的基础上降低了前期数据收集难度,模型综合准确率提高至89.34%,其中番茄病毒病的诊断精确度和F1值均达到96%以上,运行时间减少了47.73%;最后通过番茄叶霉病和番茄早疫病两种病害对本文模型进行了泛化能力测试,实验结果表明该模型具有较强的泛化能力和实用性。基于LightGBM模型设计的APP可以实现用户人群友好的交互式可视化且满足实际诊断需求。     

覆膜与水分控制对宁夏设施滴灌番茄产量与品质的影响

【目的】探索宁夏设施滴灌番茄覆膜的效果和适宜的灌溉制度,为宁夏地区设施滴灌番茄节水高产种植提供理论依据。【方法】通过2 a试验,在覆膜(M)与不覆膜(NM)条件下,设定4种水分控制水平,灌水频率为7~10 d,W1、W2和W3处理的灌水上限分别为100%FC(田间持水率)、80%FC和70%FC,以当地灌水量为对照(CK,灌水上限为123%FC),研究了覆膜和水分控制对设施滴灌番茄生长、产量、品质与水分生产效率的影响。【结果】番茄株高和茎粗在苗期―开花坐果期生长迅速,受覆膜处理影响显著(P<0.01),受水分控制影响不显著(P>0.05)。随着灌水量的增加,番茄产量先增加后降低,W2M处理的产量和水分生产效率均为最大值,分别为89844.88 kg/hm2和502.5 kg/(hm²·mm),相比覆膜CK分别增加21.4%和63.7%。相比于不覆膜处理,覆膜番茄产量平均增加18.1%,差异达到显著水平(P<0.05),其还原性维生素C和可溶性固形物量分别提高28.9%和22.8%(P<0.05)。覆膜和水分控制均对番茄还原性维生素C量、可溶性固形物和可溶性总糖量的影响达到显著水平(P<0.01),覆膜处理还对可滴定酸量和糖酸比的影响达到显著水平(P<0.05)。【结论】对于宁夏设施滴灌番茄,采用覆膜栽培与80%FC的灌水上限可以获得较高产量和较好品质。     

基于WGAN和MCA-MobileNet的番茄叶片病害识别

针对番茄病害识别模型参数量大、计算成本高、准确率低等问题,本文提出一种基于多尺度特征融合和坐标注意力机制的轻量级网络（Multi scale feature fusion and coordinate attention MobileNet, MCA-MobileNet）模型。采集10类番茄叶片图像,采用基于Wasserstein距离的生成对抗网络（Wasserstein generative adversarial networks, WGAN）进行数据增强,解决了样本数据不足和不均衡的问题,提高模型的泛化能力。在原始模型MobileNet-V2的基础上,引入改进后的多尺度特征融合模块对不同尺度的特征图进行特征提取,提高模型对不同尺度的适应性;将轻量型的坐标注意力机制模块（Coordinate attention, CA）嵌入倒置残差结构中,使模型更加关注叶片中的病害特征,提高对病害种类的识别准确率。试验结果表明,MCA-MobileNet对番茄叶片病害的识别准确率达到94.11%,较原始模型提高2.84个百分点,且参数量仅为原始模型的1/6。该方法较好地平衡了模型的识别准确率和计算成本,为番茄叶片病害的现场部署和实时检测提供了思路和技术支撑。     

Processing tomato quality as affected by irrigation scheduling

In Southern Italy the paucity of water is a very important problem and due to the fundamental importance of the water for the processing tomato, in this study the influence of irrigation on processing tomato quality was investigated in the years 2002 and 2003. A hybrid processing tomato (Lycopersicon esculentum L. Mill cv. Ability) was treated with different water regimes obtained combining amount of water and irrigation interval. The effects of the trials were evaluated taking into account the physical and chemical characteristics of the fruits, as well as the content in antioxidants moieties. Furthermore, the relationship among all the quality parameters and the seasonal amount of irrigation water was evaluated.The results, comparing different water regimes obtained combining amount of water and irrigation interval, gave useful indications on the possibility to improve nutritional tomato quality by reducing irrigation water during tomato cropping. In particular, relatively to the Mediterranean areas of tomato cultivation and to the texture of the soil considered in this study, the best compromise between quality and quantity of the processing tomato fruit was achieved both with the extension of the irrigation interval (L40 or L60) and with the limitation of irrigation volume for the second part of the tomato crop cycle (thesis 100-50). Extending the irrigation interval and limiting irrigation volume for the second part of the tomato crop cycle, appeared to be the best management strategy to optimize the yield and nutritional quality of processing tomato.     

Resistance of 57 greenhouse-grown accessions of Lycopersicon esculentum and three cultivars to Tuta absoluta (Meyrick) (Lepidoptera: Gelechiidae)

The aim of this study was to evaluate resistance to Tuta absoluta (Meyrick) (Lepidoptera: Gelechiidae) by antixenosis on 57 Lycopersicon esculentum Mill. accessions from the Horticultural Germplasm Bank (HGB) of Universidade Federal de Viçosa and by the three commercial cultivars (Santa Clara, Moneymaker and TOM-601) under greenhouse conditions. A randomized complete block design was used with three replications. Infestations with T. absoluta adults were performed weekly and the following characteristics were evaluated: number of small, large and total mines/leaf and % of leaves mined at days 60, 75 and 90 after planting. Low infestation occurred at days 60 and 75 but at day 90, infestation was sufficient to evaluate insect damage. Based on these data it was concluded that only accessions HGB-674 and HGB-1497 appeared to be the most promising. In addition, to determine possible chemical causes of resistance, hexane extracts were analyzed at day 90 by gas chromatography/mass spectrometry and the major peaks identified by a mass spectral database using similarity index. Nine hydrocarbons, viz., hexadecane, heptadecane, eicosane, tricosane, 2-methyltricosane, tetracosane, hexacosane, octacosane and triacontane were identified in the hexane extracts in many samples. Tricosane, tetracosane and hexacosane presented significant correlations with the leaves mined. Only tricosane presented a negative correlation with the number of small mines (r = −0.28), total number of mines (r = −0.27) and % of leaves mined (r = −0.22). However, tetracosane and hexacosane presented significant positive correlations (r = 0.25 and 0.24, respectively) with the % of leaves mined.     

A SCAR marker linked to the N gene for resistance to root knot nematodes (Meloidogyne spp.) in pepper (Capsicum annuum L.)

The root-knot nematodes (Meloidogyne spp.) are important nematode pests and cause serious diseases in pepper in the world. No molecular markers linked to the nematodes resistance N gene have been reported. In this paper, ‘Carolina Wonder’ (Capsicum annuum L.), a sweet pepper line resistant to root-knot nematode with N gene, ‘20080-5-29’ (C. annuum L.), an inbred line susceptible to root-knot nematode with good horticultural characteristics, and their F₂ progeny with 320 individuals were used as materials. Evaluation of resistance and susceptibility of parental lines, F₁ and F₂ progeny inoculated with root-knot nematodes (Meloidogyne incognita) were carried out. ‘Bulked segregant analysis’ method was used to search for polymorphic markers from 512 pairs of AFLP primers. Based on the assessment of resistance and susceptibility and polymorphism of the AFLP marker in F₂ population, the genetic linkage distance between the AFLP marker and the N gene was estimated. One AFLP marker E39/M41-339 was obtained and transferred to a SCAR marker amplifying a 315 bp DNA fragment linked to the N resistant allele and a 331 bp fragment linked to the N⁺ susceptible allele. The distance between the molecular marker and the nematodes resistance N gene is 6.3 cM. This research delivered a valuable tool for the marker assisted selection of nematodes resistance in pepper.     

基于计算机视觉技术的番茄叶片叶绿素含量的检测

 研究利用计算机视觉技术快速测定叶绿素含量的方法,建立了根据番茄叶片颜色特征确定其叶绿素含量的一元二次拟合模型。在计算机视觉图像采集系统中采集番茄叶片图像,利用MATLAB图像处理工具提取图像的颜色特征参数,对颜色特征参数和番茄功能叶叶绿素含量做相关分析,建立回归模型。结果表明：RGB颜色系统的R/G、(G-R)/(G+R)、G-R、色度坐标r、r-g及HIS颜色系统的H值均与叶绿素含量呈极显著非线性相关性,可用于测定番茄叶片叶绿素含量。从建立的6组模型中筛选出拟合度较高的3组模型进行检验,预测误差在0～22.22%之间。用预测精度最高的G-R颜色特征预测叶绿素含量的模型为Chl.a = 0.0926 + 0.1208 (G-R) - 0.0009 (G-R)²,Chl b = - 0.0252 + 0.0397 (G-R) - 0.0003 (G-R)²和Chl.(a+b) = 0.1271 + 0.1600 (G-R) - 0.0011 (G-R)²。     

氮磷不同配比施肥对加工番茄生长及产量的影响

配方施肥明显改善了加工番茄的生长,增加了植株分枝数、单株坐果数和单果重,对提高加工番茄的产量有显著效果;过多的氮肥投入反而影响番茄的产量.结果表明:N:P2O5=1:1,投肥总量为施纯N 1.073～1.38 kg/hm2,P2O50.981～1.38 kg/hm2,产量和效益最佳.     

抗坏血酸钛对樱桃番茄产量和品质的影响

抗坏血酸钛在樱桃番茄上的试验于2006～2008年进行,发现对樱桃番茄叶面喷施抗坏血酸钛具有显著的提高产量和改善品质的作用。叶面喷施10～20 mg/kg的钛2次,对促进保护地、露地樱桃番茄早熟增产、优质及提高经济效益效果显著。抗坏血酸钛能促进樱桃番茄植株生理活性增强,提高过氧化氢酶活性,有利于樱桃番茄叶绿素含量增加和光合作用强度提高,果实中可溶性固形物和维生素C含量增加等。     

樱桃番茄嫁接砧木筛选研究

以格格樱桃番茄作接穗,农优野茄、托鲁巴姆、金钻砧木和砧木1号作砧木,进行嫁接和嫁接苗栽培试验.试验结果表明.4种砧木与樱桃番茄嫁接亲和性由强到弱依次为:砧木1号、金钻砧木、农优野茄和托鲁巴姆;嫁接苗青枯病发病率与自根苗差异均达到极显著水平,嫁接苗产量比自根苗增产15.9%～48.7%,在抗青枯病和产量方面表现都比较好的砧木为农优野茄和砧木1号.