全面提高农林院校硕士生计算机应用水平

随着我国改革开放步伐的加快、加入世贸组织日期的临近,高等农林院校硕士研究生的计算机教育将面临更大的挑战和机遇。如何抓住机遇,迎接挑战,全面提高农林院校硕士研究生计算机应用水平,是目前农林院校硕士生教育急需解决的问题之一。本在调查研究的基础上分析探讨了目前国内农林院校硕士研究生计算机教育存在的问题,并给出了几点改革建议。     

基于显著性增强和迁移学习的鱼类识别研究

针对水下鱼类无法快速准确识别的难点,提出一种具有图像主体自动增强功能的鱼类迁移学习方法。该方法将鱼类RGB图像转换至Lab颜色空间后,利用中央周边算子计算得到整个输入图像的显著性值,进而提供鱼类目标的潜在区域,并结合GrabCut算法获取鱼类分割图像,最终将融合分割图的原始图像送入优化后的残差网络中进行训练。通过对23种鱼类进行识别试验,结果显示,固定ImageNet数据集上ResNet-50预训练模型的conv1层和conv2层参数,微调高层参数的方法能够取得最好的识别效果,且在公开的Fish4Knowledge数据集上,该模型取得了最高的识别准确率,平均识别精度达到99.63%。与其他卷积神经网络方法的对比结果显示,本方法在Fish4Knowledge和Fish30Image数据集上的识别精度和时间性能均具有较大优势,其中识别准确率至少提升4.98%。多个数据集上的试验验证了模型的有效性。     

Springtime abundance of chaetognaths in the shelf region of the northern Gulf of Alaska, with observations on the vertical distribution and feeding of Sagitta elegans

Chaetognatha were among the most dominant macrozooplankton taxa collected in 6 years of springtime collections, both throughout the water column and near-bottom on the continental shelf in the northern Gulf of Alaska. Three species ( Sagitta elegans, S. scrippsae, and Eukrohnia hamata ) were collected in 248 collections, although S. elegans was the numerically dominant species overall during most cruises. Collections taken in epibenthic sleds generally contained about two orders of magnitude more chaetognaths than those that sampled throughout the water column. Two size modes were apparent in the S. elegans size-frequency distributions which were believed to belong to different cohorts. Growth rate was on the order of 2–3 mm per month. The two cohorts showed asynchronous diel vertical distribution patterns, with the smaller individuals found near the surface during the day whereas the larger individuals were near the surface at night. About 7.5% of the S. elegans examined contained food. Copepods made up the majority (∼89% by number) of the diet of both large and small individuals. Euphausiid juveniles and cirripede larvae were also observed, as well as several incidences of cannibalism by large chaetognaths on smaller individuals. Based on the results of two diel series, this species was observed to feed mainly during the night-time.     

闽南-台湾浅滩渔场狗母鱼类食性的研究

本文对闽南一台湾浅滩渔场狗母鱼类(多齿蛇鲻、花斑蛇鲻和大头狗母鱼)的食料生物组成、摄食强度的季节变化以及消化器官与食性等方面进行了探讨。结果表明，这三种狗母鱼类都属于游泳动物食性的底层鱼类，而且是同种残食的中级肉食性鱼类，其食料生物组成比较简单，主要摄食鱼类和头足类，还兼食长尾类、短尾类和口足类。被捕食对象的大小，随着捕食对象(多齿蛇鲻或花斑蛇鲻)叉长的增加而增大。狗母鱼类周年均有摄食，多齿蛇鲻和花斑蛇鲻在产卵盛期(3—4月)摄食强度较高。它们的消化器官的形态结构与其肉食性相适应，多齿蛇鲻和花斑蛇鲻的消化器官和食性相近似，比大头狗母鱼的食性较凶猛。     

瓣结鱼食性的研究

以采自贵州境内清水江、都柳江的318尾瓣结鱼(Tor(Foliter)brevifilis brevifilis)为研究对象,从食物组成、摄食强度和摄食形态学等方面研究了瓣结鱼食性的特点。结果表明:瓣结鱼的食物由双翅目昆虫幼虫等17类饵料生物组成,并以双翅目昆虫幼虫的出现率最高。瓣结鱼的充塞度按春、夏、秋、冬季逐渐增高,季节性变化明显。充塞指数随体长生长显著减少,而摄食率则随体长生长逐渐增加。充塞指数和摄食率均以V期性腺个体为最低。瓣结鱼为底层杂食性鱼类。     

基于稳定同位素的口虾蛄食性分析

为了探讨口虾蛄的食物组成,利用稳定同位素方法对2015年5月在汕尾红海湾海域采集的口虾蛄及其饵料生物的碳、氮稳定同位素比值(δ~(13)C和δ~(15)N值)进行分析,定量研究不同饵料生物在口虾蛄食物中的贡献比率。结果表明,口虾蛄的δ~(13)C值为–18.1‰~–16.3‰,δ~(15)N值为10.9‰~13.5‰,平均值分别为–17.1‰±0.5‰和12.7‰±0.7‰。δ~(13)C和δ~(15)N值的变化范围均较大,表明口虾蛄的食物来源较多。口虾蛄的食物主要由鱼类、虾类、贝类、蟹类和桡足类组成。其中,贝类为口虾蛄的主要食物,平均贡献率为38.6%;其次为蟹类和桡足类,平均贡献率分别为22.9%和16.0%;虾类的平均贡献率为13.6%;鱼类的平均贡献率最低,仅为8.9%。根据δ~(15)N值及营养级的计算公式得出,口虾蛄的营养级为3.01±0.22,在其5类食物中,桡足类的营养级最低,仅为1.77±0.12;其次为贝类;蟹类和虾类的营养级分别为2.78±0.21和2.89±0.16;鱼类的最高,为2.98±0.15;它们的营养级均低于口虾蛄。此外相关分析显示,口虾蛄的δ~(15)N值与其个体体质量间存在极显著的正相关关系,说明不同大小的口虾蛄营养级有所差异。     

海蜇—对虾—缢蛏—牙鲆综合养殖池塘的食物网分析

在面积6.7 hm^2,水深1.5～2.0 m的池塘中进行海蜇—中国明对虾—缢蛏—褐牙鲆综合养殖,在4—9月养殖周期内,采用碳氮稳定同位素法和IsoSource线性混合模型计算各种饵料对海蜇、中国明对虾、缢蛏和褐牙鲆的平均饵料贡献率,以探讨综合养殖池溏的营养级和池塘食物网结构。试验结果表明,海蜇的主要食物来源为投喂的轮虫;中国明对虾的食物来源为鳀鱼、虾夷扇贝下脚料和蜢虾;缢蛏的主要食物来源为底栖硅藻,其次为中国明对虾与褐牙鲆粪便;褐牙鲆的主要食物来源为中国明对虾,其次为虾夷扇贝下脚料和鳀鱼。海蜇的营养级为2.09～3.43,平均值为2.81;中国明对虾的营养级为3.11～4.04,平均值为3.72;缢蛏的营养级为2.37～2.94,平均值为2.62;褐牙鲆的营养级为2.63～3.83,平均值为3.42。综合分析发现,海蜇摄食浮游动物,缢蛏滤食底栖硅藻,中国明对虾和褐牙鲆摄食投喂的虾夷扇贝下脚料和鳀鱼,褐牙鲆也摄食部分中国明对虾,两者的粪便不仅为浮游单胞藻和底栖硅藻生长提供营养,还为缢蛏提供了饵料。上述研究结果为多品种海水综合养殖池塘的生态健康养殖提供科学依据。     

Feeding habits of albacore Thunnus alalunga in the transition region of the central North Pacific

ABSTRACT:   The feeding habits of albacore Thunnus alalunga (fork length: 48.9–76.2 cm, n  = 132) were examined from late spring to early autumn in relation to its northward migration in the transition region between the subtropical and subarctic fronts in the central North Pacific. Samples were collected at night using surface gill nets or during daytime pole-and-line surveys in 2001 and 2002. During May and June, albacore fed mainly on Japanese anchovy Engraulis japonicus , which accounted for 27.2%, 67.0%, and 45.5% of the total stomach contents by number ( Cn ), wet weight ( WW ), and frequency of occurrence ( F ), respectively, and secondarily on the subarctic gonatid squid Gonatopsis borealis ( Cn , 15.8%; WW , 10.8%; F , 28.8%). From July to September, albacore continued to depend on Japanese anchovy ( Cn , 48.2–52.8%; WW , 79.9–95.2%; F , 27.8–85.4%). These results corresponded well with the remarkable rebound of the Japanese anchovy stock since the 1990s. Gonatopsis borealis , the main squid prey from May to June, almost disappeared from the stomachs of albacore from July to September, probably due to the northward migration of this squid to subarctic waters in summer. The feeding impact of albacore on the Japanese anchovy stock in the transition region was conservatively estimated to be from 1400 to 2100 tons per day from late spring to early autumn.     

基于变量优选与机器学习的干旱区湿地土壤盐渍化数字制图

土壤盐渍化是导致土壤退化和生态系统恶化的主要原因之一,对干旱区的可持续发展构成主要威胁。为了尽可能精确地监测土壤盐渍化的空间变异性,该研究收集新疆艾比湖湿地78个典型样点,其中选取54个样本作为训练集,24个样本作为独立验证集。基于Sientinel-2 多光谱传感器（Multi-Spectral Instrument,MSI）、数字高程模型（Digital Elevation Model,DEM）数据提取3类指数（红边光谱指数、植被指数和地形指数）,经过极端梯度提升（Extreme Gradient Boosting,XGBoost）算法筛选有效特征变量,构建了关于土壤电导率（Electrical Conductivity,EC）的随机森林（Random Forest,RF）、极限学习机（Extra Learning Machine,ELM）和偏最小二乘回归（Partial Least Squares Regression,PLSR）预测模型,并选择最优模型绘制了艾比湖湿地盐渍化分布图。结果表明：优选的红边光谱指数基本能够预测EC的空间变化;红边光谱指数与植被指数组合建模效果总体上优于其与地形指数的组合,3类指数组合的建模取得了较为理想的预测精度,其中RF模型表现最优（验证集R2=0.83,RMSE=4.81 dS/m,RPD=3.11）;在整个研究区内,中部和东部地区土壤盐渍化程度尤为严重。因此,XGBoost所筛选出的环境因子结合机器学习算法可以实现干旱区土壤盐渍化的监测。