Image and Feature Space Based Domain Adaptation for Vehicle Detection

The application of deep learning in the field of object detection has experienced much progress. However, due to the domain shift problem, applying an off-the-shelf detector to another domain leads to a significant performance drop. A large number of ground truth labels are required when using another domain to train models, demanding a large amount of human and financial resources. In order to avoid excessive resource requirements and performance drop caused by domain shift, this paper proposes a new domain adaptive approach to cross-domain vehicle detection. Our approach improves the cross-domain vehicle detection model from image space and feature space. We employ objectives of the generative adversarial network and cycle consistency loss for image style transfer in image space. For feature space, we align feature distributions between the source domain and the target domain to improve the detection accuracy. Experiments are carried out using the method with two different datasets, proving that this technique effectively improves the accuracy of vehicle detection in the target domain.     

机床再生颤振及颤振抑制方法

再生颤振是加工和结构变形之间不稳定相互作用的结果,它极大地制约了生产加工。主要描述车削和铣削加工的建模、分析、检测以及对再生颤振的控制。     

Insights into biological delignification of rice straw by Trametes hirsuta and Myrothecium roridum and comparison of saccharification yields with dilute acid pretreatment

Rice straw is the most abundant agricultural residue on a global scale and is widely available as feedstock for cellulosic fuel production. However, it is highly recalcitrant to biochemical deconstruction and also generates inhibitors that affect enzymatic saccharification. Rice straw from eastern Arkansas was subjected to dilute acid pretreatment (160 °C, 48 min and 1.0% sulfuric acid) and solid-state fermentation with two lignocellulolytic fungi, Trametes hirsuta and Myrothecium roridum, and their saccharification efficacies were compared. T. hirsuta and M. roridum were tested separately; pretreatment of rice straw with either strain for seven days resulted in 19 and 70% enrichment of its holocellulose content, respectively. However, liquid chromatography analysis of the alkali extracts showed significant differences in cell wall degradation by T. hirsuta and M. roridum. T. hirsuta removed 15% more phenolic compounds and 38% more glucan than M. roridum, while M. roridum removed 77% more xylan than T. hirsuta. Fungal and dilute acid pretreated biomass was then hydrolyzed using Accellerase^® 1500, a saccharification cocktail. Saccharification efficiency of M. roridum was 37% higher than that of dilute acid pretreatment of rice straw, requiring 8% lower enzyme loading and 50% shorter enzymatic hydrolysis duration. Alkali extraction of fungal pretreated biomass also yielded 10–15 g kg⁻¹ of acid precipitable polymeric lignin (APPL), which is a valuable co-product for biorefineries. In comparison to dilute acid pretreatment, fungal pretreatment could be a cost-effective alternative for the degradation of recalcitrant biomass, such as rice straw.     

Searching and learning english translation long text information based on heterogeneous multiprocessors and data mining

Cooperation with the entire language information retrieval aims to develop intelligent models that use non-compliant and English. It has been constructed to construct the translation model extraction of similar Source Language Sentences Extraction (SLSE). The SLSE, by creating a new pipeline of SLSE through the establishment of the network described above, the first bilingual dictionary of the model of training data, including the Deep Neural Network (DNN) that has been proposed, because it is often in such, it is building because it plays a very important role in the model, SLSE is evaluated by people to ensure the richness of the content that has been used as training data. If the English applicable, not built from 60 gigabytes of text. Since the processing time for it when not highly structured data mining is large, it is troublesome to collect data on this scale. To shorten the processing time, data mining is implemented as a whole.     

Alkali catalyzed liquefaction of corncob in supercritical ethanol–water

Corncob liquefaction in supercritical ethanol–water was performed with and without the addition of an alkali catalyst by direct addition or biomass impregnation in a 250-cm³ batch reactor. The effects of temperature, solvent and alkali addition on the biomass conversion level and oil yield were investigated to find the optimum condition. For non-catalytic liquefaction using a 1:1 (v/v) ethanol: water ratio, a maximum oil yield and conversion level of 49.0% and 93.4%, respectively, were obtained at 340 °C. For alkali catalytic liquefaction, the oil yield with KOH addition (57.5%) was higher than that from KOH-impregnated corncob liquefaction (43.3%). The oil from liquefaction with KOH addition had higher heating value (26.7–35.3 MJ kg⁻¹) than the corncob (19.1 MJ kg⁻¹). The dominant components of the obtained oil were found by GC/MS analysis to be aldehyde, ester, phenol derivatives and aromatic compounds.     

因素空间理论下基点分类算法研究

目前，基于因素空间理论的背景基提取算法计算过程复杂，初始化必须依赖各因素极值，基点数量提取冗余等原因，未能在应用中取得很好效果。为此，结合内点判别法和知识可继承、可扩展的思想，提出一种计算简单、初始化独立、基点数量小的改进的背景基提取算法。然后，利用改进的背景基提取算法构造出一种全新的数据分类算法-基点分类算法，基点分类算法以提取每一类样本的背景基为预测模型，再通过新定义的λ-背景基，优化预测模型。数值实验表明:基点分类算法原理简单、构造难度小、分类模型泛化能力强，预测能力准确率高，同时严格的模型限定区域又能为识别新类别提供新方法。     

Corrosion behavior of Cu–Sn bronze alloys in simulated archeological soil media

The corrosion behavior of synthetic Cu–Sn bronze alloys with six different Sn contents was examined through an electrochemical test and a synthetic test in a simulated corrosive medium. The mechanism of corrosion and the morphology of the corroded surfaces were characterized through field emission scanning electron microscopy equipped with energy-dispersive spectroscopy. At the corrosion potential, the corrosion behavior appears to be determined by the charge transfer step and the diffusion process. It was found that the bronze-IV (Cu–26.8Sn) specimen exhibited the best corrosion resistance, as evidenced by a low corrosion current density and a high impedance. This improvement resulted from an increase in the content of the Cu–Sn solid solution in the alloy, which was conducive to forming a relatively more protective passive film on the surface of the bronze alloy. This finding would be valuable in the anticorrosion protection of archeological artefacts after their excavation.     

氯甲烷在镁修饰的ZSM-5分子筛催化剂上催化转化研究

在镁修饰的ZSM-5分子筛催化剂上,氯甲烷可以被催化转化为烃类产品。分别采用了离子交换和浸渍的方法来修饰ZSM-5分子筛。离子交换的Mg-ZSM-5催化剂和浸渍的Mg/ZSM-5催化剂都可以提高反应产物中的低碳烯烃的选择性,降低烷烃的选择性,并且保持较高的反应活性,但高的浸渍量对反应活性有一定的影响。NH3-TPD的结果表明,浸渍镁的ZSM-5催化剂的强酸中心数目明显减少。镁的修饰对催化剂酸性的影响导致了产物中低碳烯烃的增加。浸渍Mg的ZSM-5催化剂是潜在氯甲烷转化生成低碳烯烃的催化剂。     

数字化蓄热式钢包烘烤器及自动控制系统

蓄热式燃烧技术和脉冲燃烧技术是新型的节能环保燃烧技术。数字化蓄热式钢包烘烤器控制系统利用数字化脉冲控制把这两项技术结合在一起。这大大降低了能耗，减少环境污染。本文针对数字化蓄热式钢包烘烤器，提出升温阶段采用自寻优算法，不断调整换向时间，并能记忆最优参数，保证经过几次烘烤后系统整体最优；保温阶段即采用脉冲控制的方法。     

Na+、K+离子交换对Co-Mo/MCM-41加氢脱硫催化剂的影响

 用偏硅酸钠和正硅酸乙酯作硅源制备了MCM-41（分别记作MCM-41(S)和MCM-41(T)）分子筛，并用Na₂C₂O₄和K₂C₂O₄对MCM-41(S)进行了碱金属离子交换改性。以质量分数0.8 %的二苯并噻吩（DBT）的十氢萘溶液为模型化合物，考察了不同MCM-41担载的Co-Mo硫化物催化剂对DBT的加氢脱硫反应性能。结果表明，MCM-41担载的Co-Mo催化剂加氢活性较低，DBT主要通过直接脱硫反应路径脱硫。其活性顺序为：Co-Mo/MCM-41(T)>Co-Mo/MCM-41(S)>Co-Mo/MCM-41(K)>Co-Mo/MCM-41(Na)。UV-Vis结果表明，部分Co与MCM-41(S)中少量Al发生相互作用，生成了CoAl₂O₄，是造成Co-Mo/MCM-41(S)活性降低的重要因素。而在Co-Mo/MCM-41(K)和Co-Mo/MCM-41(Na)中，除CoAl₂O₄物种之外，碱金属的引入还促进了Co₃O₄物种的形成，使其活性进一步降低。