基于改进ResNeXt50残差网络的锦鲤选美方法

锦鲤选美的不同等级之间具有高相似度的特点,目前都是人工进行选美分级。为解决人工选美所存在的效率低、主观性强、成本高的问题,提出了一种基于迁移学习和改进ResNeXt50残差网络的锦鲤选美方法。本文首先构建了红白、大正、昭和3种锦鲤的选美等级数据集。其次,采用迁移学习策略提高训练速度,并从SE注意力模块、Hardswish激活函数和Ranger优化器3方面对ResNeXt50模型进行了改进,构建了SH-ResNeXt50锦鲤选美分级模型。试验结果表明：SH-ResNeXt50模型有效提升了锦鲤选美的等级分选能力,模型准确率达95.6%,损失值仅0.074,优于常用的AlexNet、GoogLeNet、ResNet50和ResNeXt50网络模型。最后,采用Grad-CAM分析SH-ResNeXt50模型的可解释性,结果表明SH-ResNeXt50模型和人工识别的感兴趣区域基本一致。本文所提出的方法实现了具有高相似度的锦鲤不同等级的智能分选,对其它具有高相似度的生物等级分选具有借鉴意义。     

Hysteretic Behavior of Composite Deep Beam Infilled Steel Frame

The Quasi static tests of one steel frame and two the composite deep beams infilled steel frames were carried out. The effects of the deep beams on the load capacity, ductility, hysteretic property and energy dissipation of pure steel structure were analyzed. It is found that the hysteresis curve is a straight line with the constant of the stiffness at the beginning and without residual deformation. And the hysteresis curve of specimen was full after yielding, and the skeleton curves had a clear plastic flow phase with triple linear. The lateral drifts of the beam specimens at failure were 1/25 and 1/22. The composite deep beams enhance the initial stiffness, yield load and maximum load bearing capacity of steel frame. Therefore, seismic performance of the composite deep beams is better.     

番茄营养液深液流无限生长型栽培（Ⅰ）

以营养液深液流为栽培方式,以无限生长型番茄为研究对象,应用植物的超越补偿原理,根据番茄生长发育不同时期对光照、温度、湿度等的不同要求,通过对温室综合环境的调节与控制,以及对番茄营养生长与生殖生长的控制与转化,经过一年半的栽培试验,首次在国内实现了单株番茄结果超万个;对番茄果实的化验结果表明,试验栽培出的果实品质优良,营养丰富,硝酸盐、亚硝酸盐和有机磷的含量均低于国家绿色食品规定的标准。     

我国龙眼产业化进程中的问题与对策

我国龙眼在产业化进程中,存在着种植品种结构发展不平衡,生产经营规模化、标准化程度低;品牌未形成、市场建设滞后、缺乏行业自律和深加工比例少;科研投入不足,技术力量薄弱,缺乏整体布局等问题。提出采用调整优化龙眼品种结构,培育“龙头企业”,组建龙眼产业协会与合作社,严格企业自律,建设集散市场;实行标准化生产,树立品牌;实施冷链流通,建设完整的物流系统;加大龙眼采后深加工研发投入,促进产业结构优化;按ISO、HACCP等国际认证体系规范我国龙眼产业,建立高效联动的信息发布和预警预报机制等对策。     

Application of modified Deep Belief Network in Forecasting Cotton Diseases and Insect Pests

[Objective] The occurrence and development of cotton diseases and insect pests are mainly related to environmental information. Because this environmental information is various, complex and unstable, the study on the prediction methods of cotton diseases and insect pests is a certain challenge. This study aims to establish a forecasting model for the timely and accurate prediction of cotton diseases and insect pests. [Method] A forecasting model of cotton diseases and insect pests is proposed based on environmental information and a modified Deep Belief Network (DBN) that is constructed by a three-layer restricted Boltzmann machine (RBM) and a supervised back-propagation (BP) network. In the method, the RBM is used to transform the original environmental information vectors into a new feature space related to the diseases and pests; the BP network is trained to classify and forecast the features generated by the last RBM layer and two rules of dynamic learning and comparison and dispersion are adopted to accelerate the training process of RBM. The proposed model was validated on a dataset of cotton bollworm, aphids, spider, cotton Verticillium wilt, and Fusarium wilt in a recent six-year period. [Result] Compared with the traditional prediction models of cotton diseases and insect pests, the proposed model can deeply explore the extensive correlation between the occurrence of cotton diseases and pests and environmental information. The results show that the proposed model has a higher accuracy compared with the classical predictive models, and the average forecasting accuracy is above 83%. [Conclusion] The proposed method is an effective crop disease and pest forecasting method that can provide a technical support for preventing cotton disease and insect pests.     

Yield Performance of Wheat Isolines with Different Dosages of the Short Arm of Rye Chromosome 1

Translocations of the short arm of rye (Secale cereale L.) chromosome 1 (1RS) in wheat (Triticum aestivum L. cv. Pavon 76) are known to increase root biomass. Such an increase enhances water and nutrient uptake and may improve grain yield. Two greenhouse experiments and a field experiment were carried out at the University of California, Riverside, in 2012 and 2013 under well‐watered and terminal drought treatments to evaluate phenotypic characters associated with varying dosages of 1RS, including grain yield. The genotypes used were cultivar Pavon 76 (R₀), Pavon 76/Pavon1RS.1AL (F₁ hybrid) with a single dosage of 1RS (R₁A), Pavon 1RS.1AL with two dosages of 1RS (R₂A), Pavon 1RS.1DL (R₂D) also with two dosages of 1RS and Pavon 1RS.1AL‐1RS.1DL (R₄AD) with four dosages of 1RS. There was a significant positive correlation between number of dosages of 1RS and root biomass. However, no correlation was found between root biomass and grain yield per plant. Drought in the field experiment reduced grain yield significantly. Under well‐watered field conditions, grain yield of R₂A (215.9 g plant^?1) was significantly greater than those of R₂D (191.8 g plant^?1) and R₄AD (161.7 g plant^?1). Also, grain yield of R₄AD was significantly less than those of F₁, Pavon 76 and R₂D under well‐watered conditions. Under drought field conditions, no significant differences were found among the genotypes for grain yield was found between F₁ (14.7 g plant^?1) and R₄AD (12.4 g plant^?1). Harvest index was significantly greater in well‐watered (44.2 %) than in drought (34.6 %) field conditions. On average, genotypes F₁ (42.3 %) and R₂A (40.6 %) had higher harvest index than R₂D (38.3 %) and R₄AD (35.5 %) in the field. Also, Pavon 76 (40.2) and R₂D (38.3) had higher harvest index than R₄AD. Drought tolerance was lowest for R₄AD due to its relatively lower grain yield potential. In general, Pavon 1RS.1AL carrying two dosages of 1RS showed higher grain yield under wet treatments. Pavon 1RS.1AL‐1RS.1DL carrying four dosages of 1RS produced the largest shoot and root biomasses, but the least grain yield.     

多邻域保持嵌入的人脸识别方法

现有流形学习算法在学习人脸数据时,假设所有数据点位于单一低维嵌入流形之上,当数据点实际分布在不同的流形上时,单流形假设就会影响数据真实空间结构。为此提出一种基于多邻域保持嵌入（multiple neighborhood preserving embedding, M-NPE）的学习算法来发现不同类别数据在不同维度的低维嵌入空间中分布的多流形结构。首先,单独学习不同类别数据的流形,得到反映其本质特征的流形;再通过遗传算法搜索每个流形的最优维数;最后依据最小重构误差分类器对样本分类。在Extended Yale B和CMU PIE这2个大型人脸库上实验结果验证了该算法的有效性。     

Gene cloning of ZmMYB59 transcription factor in maize and its expression during seed germination in response to deep‐sowing and exogenous hormones

Discovering the key genes responsible for deep‐sowing germination is very meaningful to cultivate maize varieties with strong germination ability in soil. In this study, ZmMYB59 gene whose functions remain unknown was successfully cloned. Subcellular localization showed that ZmMYB59 was localized in the nucleus. By analysing cis‐acting elements of its promoter sequence, many elements of MYB binding domain responsive to hormones (particularly gibberellin, GA) were found. On this basis, ZmMYB59 expression in different sowing depths, germination stages, tissues and treatments was analysed by real‐time PCR. ZmMYB59 expression levels in embryos or seedlings were significantly decreased with the increase in sowing depth at 2 days or 4 days after sowing (DAS). Further analysis showed that its expression in mesocotyls was also significantly downregulated with the increase in sowing depth at both 6 DAS and 8 DAS. Similarly, seed soaking with 10^‐5 M GA₃ inhibited ZmMYB59 expression in mesocotyls at 3 DAS. These results suggest that ZmMYB59 may play a negative regulatory role in seed germination in deep soil and the regulation is involved in GA signalling pathway.     

高职食品生产加工技术学习领域课程构建的研究

食品生产加工技术学习领域课程体系构建可通过"块型"与"条型"的课程组合来完成。"块型"课程是从食品生产过程中抽提出通用的单元操作及典型生产设备操作规程作为"单元操作"类课程设置,"条型"课程是选择在食品产业中占有较大比重或重要影响的产品/原料来构建"生产技术"类课程。