基于复用分割技术的中继蜂窝网的频率规划

研究了两跳固定中继蜂窝网的频率规划问题.与其他固定中继蜂窝网的研究不同,假设了固定中继节点与其所属基站链路上没有任何性能增强技术,并在此假设条件下,根据复用分割原则,提出了两种新的中继频率规划方案.同时将提出的频率规划方案与基于信道借用的频率规划方案及传统无中继的频率规划方案进行了比较.理论分析和仿真结果表明, 所提出的频率规划方案不但能提高小区边缘用户的服务质量,且与基于信道借用的频率规划方案相比,能够获得更大的系统性能提高.研究结果还表明,为了充分发挥固定中继网络的潜力,有必要在固定中继节点与其所属基站链路上引入性能增强技术.     

基于实体级遮蔽BERT与BiLSTM-CRF的农业命名实体识别

字符的位置信息和语义信息对命名方式繁杂且名称长度较长的中文农业实体的识别至关重要。为解决命名实体识别过程中由于捕获字符位置信息、上下文语义特征和长距离依赖信息不充足导致识别效果不理想的问题,该研究提出一种基于EmBERT-BiLSTM-CRF模型的中文农业命名实体识别方法。该方法采用基于Transformer的深度双向预训练语言模型（Bidirectional Encoder Representation from Transformers,BERT）作为嵌入层提取字向量的深度双向表示,并使用实体级遮蔽策略使模型更好地表征中文语义;然后使用双向长短时记忆网络（Bidirectional Long Short-Term Memory,BiLSTM）学习文本的长序列语义特征;最后使用条件随机场（Conditional Random Field,CRF）在训练数据中学习标注约束规则,并利用相邻标签之间的信息输出全局最优的标注序列。训练过程中使用了焦点损失函数来缓解样本分布不均衡的问题。试验在构建的语料库上对农作物品种、病害、虫害和农药4类农业实体进行识别。结果表明,该研究的EmBERT-BiLSTM-CRF模型对4类农业实体的识别性能相较于其他模型有明显提升,准确率为94.97%,F1值为95.93%。     

多源感知高效循环智能控制设施大棚架构与实现

以提高作物经济效益、减少资源消耗、增加农业循环利用为目标,提出一种多源感知高效循环设施大棚智能控制及农业专家系统。利用太阳能加热相变储热材料增加棚温;通过特殊通风窗设计增大空气对流面积降低棚温,减少高能耗设备的启用;阴阳棚混合种植实现农业循环利用;采用多源感知提高控制精准,采用动态权值融合算法提高数据融合精度,可通过计算机、手机实时监控。该农业专家系统控制策略结合作物生长模型、温室控制模型与能耗模型3种模型,提出根据大棚作物光合作用速率的多因素节能优化控制方法。根据计算控制增加的经济效益与付出能耗费用的差值最大,得出最优控制策略输出,以促进作物光合作用产生高收益,并减少温室控制成本。整套监控系统已试运行于徐州现代农业实验示范基地。     

Influence of Energy and Protein Content of the Concentrate and Restricting Concentrate Intake on Growth and Development of Weanling Horses

Forty-four Thoroughbred and Quarter Horse weanlings were used in three experiments to evaluate the effects of energy content, protein content, and feeding levels of the concentrate on growth and development when fed with Coastal bermudagrass hay. The animals were confined to drylot paddocks and fed the assigned concentrate in individual feeding stalls. The hay was group-fed in the paddocks. In the first experiment, the concentrate was fed at 1.5 kg/100 kg BW daily or ad libitum for two 1.5-h feeding periods daily. The ad libitum-fed weanlings had greater withers height gain (P<0.01), and hip height gain (P<0.05) than the limit-fed weanlings. Bone mineral gains were numerically greater for the ad libitum-fed weanlings (P>0.10). None of the other body measurements were different (P>0.05). In Exp. 2, weanlings were fed a basal concentrate ad libitum for two1.5-h feeding periods daily or a fat-added concentrate at equal amounts to the hay intake. Nutrients were balanced to meet NRC recommendations. The fat-added concentrate resulted in numerically greater withers height (P>0.10), heart girth (P>0.10), body length (P>0.10), hip height (P>0.10), and bone mineral (P>0.10) gains. Weight gain and feed efficiency favored the basal diet. In Exp. 3, weanlings were fed concentrates providing either 14 or 18% CP as described previously. The BW gains favored the high protein diets for three of the first four periods (P<0.05), and some numerical advantage persisted to the end of the trial (P>0.10). None of the other growth measurements were different for the two diets. Calculations on nutrient intake and BW gains of the animals on the three experiments revealed that lysine intake appears to be a greater factor than energy or protein in influencing BW gain of weanlings.     

基于旋跳自漂移修正的WSN网络信号源定位算法

为解决移动无线传感网在部署过程中存在信号源定位困难、数据链路防抖动性能差及信号上传带宽受限等难题,提出了一种基于旋跳自漂移修正机制的超高速移动WSN网络信号源定位算法.首先,考虑到区域内锚节点具有的信号源稳定特性,依据信号强度和跳数的逆向关系,构建信号强度-节点跳数强度搜寻方法,实现了信号源的定位,并从中选取性能最佳的区域节点;随后,基于移动无线传感网节点在高速移动过程中存在的频率漂移,计算锚节点跳数漂移误差,通过修正锚节点间物理差的方式来提高节点的覆盖能力,降低区域节点对区域外信号源的频率干涉覆盖现象,从而提升信号上传带宽;最后,考虑到sink节点具有的中央控制特性,通过sink节点控制来建立锚节点粒子群移动模型,使用周期递归的方式消除移动过程中锚节点移动性过高带来的频率及坐标抖动难题,进一步降低了信号源定位过程中的误差.仿真实验表明,在复杂网络条件下,与当前移动WSN网络定位中常用的超混沌偏移修正定位算法、螺旋递归自适应定位算法和逐次跌跳收敛定位算法相比,本文算法具有更高的定位精度与信号上传带宽性能,且相应的定位误差更低.     

基于GNSS姿态与电机编码器的农机转向角度测量系统研制

典型的农业机械自动驾驶系统需要在车辆转向轮上安装角度传感器测量转向角度,存在安装不便与可靠性差问题。该研究提出一种基于GNSS模块和电机编码器组合的转向轮转向角度测量系统。该系统通过组合利用GNSS姿态测量值与运动模型得到转向轮期望角度,利用电机转向速度和全液压转向阀的传递模型推算转向角度变化值,经卡尔曼滤波融合解算得到车辆转向轮的实时转向角度。与霍尔式绝对角度传感器对比的动态测试结果表明,该系统在直线行驶时的测量标准方差小于0.91°,在转向轮-10°～+10°区间,测量标准方差小于1.0°;直线自动驾驶作业时的导航误差小于2.5 cm,曲线作业自动驾驶的导航误差小于9.0 cm,满足农机自动调头等场景应用。     

Rpest—An indicator linked to a crop model to assess the dynamics of the risk of pesticide water pollution: Application to banana-based cropping systems

Like many intensive monocultures, some features of banana-based cropping systems may be detrimental to the environment. Pesticide use is a major cause of surface water and groundwater pollution. The risk of water pollution due to pesticides is very high in the insular areas of the French West Indies (FWI). In order to assess these risks and help design more sustainable cropping systems, we propose using an indicator to assess the risk of pesticide pollution over time (Rpest). Rpest provides dynamic assessments through a linkage with a cropping system model that simulates environmental factors and agricultural practices. An expert validation check was conducted and demonstrated that Rpest can rank cropping systems by risk as well as experts. A sensitivity analysis highlighted that the indicator can take the properties of active ingredients and pedoclimatic data into account in assessments. Rpest helps to pinpoint high pollution risk periods. It can also be used to test alternatives and compare systems. This tool can be integrated into a model-based, global, prototyping methodology used for more sustainable cropping systems.     

Design of Linux password high speed crack model on FPGA

In order to improve the crack speed of linux password, the crack core design idea based on dataflow is proposed. After analyzing core crack module MD5 of linux, three crack models are designed, and their ALMs consumption and run speed are analyzed. Under the structure of full pipeline, based on the idea of dataflow, the Linux crack core is designed to crack linux password fast. Experimental results show that while cracking linux password on FPGA of EP4CE40F29C8, the speed can reach up to 24.95 million/s. In the framework of fully pipeline, the idea based on dataflow makes all the pipeline data blocks in efficient working condition. Linux password crack speed increases greatly.     

Monitoring and modeling temperature variations inside silage stacks using novel wireless sensor networks

By monitoring silage temperature at different locations inside silage stacks, it is possible to detect any significant increases in temperature occurring during silage decomposition. The objectives of this study were: (1) to develop novel noninvasive wireless sensor nodes for measuring the temperature inside silage stacks; (2) to design a suitable sensor protection housing that prevents physical and chemical damage to the sensor; and (3) to mathematically model temperature variations inside a silage stack, using system identification techniques. The designed wireless nodes were used to monitor temperatures in a full-sized silage stack over 53 days. Results showed that the wireless sensor nodes accurately monitored the temperature inside the silage stack at depths of 25 and 50 cm and reliably transmitted the measured data through the network; between 98.9% and 99.4% of the packets disseminated from the sensor nodes were successfully delivered to the gateway. The reliable performance of the network confirmed the correct choice of network characteristics (i.e., frequency range of 433 MHz, a handshaking communication protocol, and 10 mW transmission power). The designed sensor housings were capable of withstanding the high loads that occurred during ensiling, storage, and feed-out. Mathematical models estimating the relations between the silage temperatures (at depths of 25 and 50 cm) and air and soil temperatures were obtained. Black-box modeling using the prediction error method (PEM) was selected as the identification method. Among different black-box models such as ARX, ARMAX, output-error (OE), and Box–Jenkins (BJ), with different model orders, a third-order Box–Jenkins model structure gave the best performance in terms of prediction accuracy. The success rate of the models proposed for silage temperature prediction ranged between 90.0% and 94.3%. Furthermore, there was no significant autocorrelation remaining in the residuals. The results of this study indicate that the designed wireless sensor nodes could potentially be used for detecting silage decomposition processes and improving the efficacy of silage conservation systems.