摘要：【中文】本发明涉及一种光稳定剂的制备方法，步骤如下：将没食子酸和甲醇反应生成没食子酸甲酯；向生成的没食子酸甲酯中加入四甲基哌啶醇、溶剂，并在催化剂作用下，通过酯交换反应生成没食子酸四甲基哌啶醇酯；将生成的没食子酸四甲基哌啶醇酯进行降温、脱色、过滤、冷却、结晶，即可得到光稳定剂；本发明公开了一种光稳定剂的制备方法，合成过程相对简单，反应条件温和，加工和使用时与聚合物相容性好，加工性能优异，且综合了没食子酸作为多酚类物质的抗氧化性和四甲基哌啶醇的光稳定效果，可放心作为食品包装材料添加助剂使用。 【EN】The invention relates to a preparation method of a light stabilizer, which comprises the following steps: reacting gallic acid with methanol to generate gallic acid methyl ester; adding tetramethyl piperidinol and solvent into the generated methyl gallate, and generating the tetramethyl piperidinol gallate through ester exchange reaction under the action of a catalyst; cooling, decoloring, filtering, cooling and crystallizing the generated gallic acid tetramethyl piperidyl alcohol ester to obtain the light stabilizer; the invention discloses a preparation method of a light stabilizer, which has the advantages of relatively simple synthesis process, mild reaction conditions, good compatibility with polymers during processing and use and excellent processing performance, integrates the oxidation resistance of gallic acid as polyphenols and the light stabilization effect of tetramethyl piperidinol, and can be used as an additive of food packaging materials without worry.

摘要：【中文】为了解决现有技术推断真实场景困难而导致的场景推断结果不准确的问题，本发明提供一种基于视觉相关判别网络的室外场景安全监控方法，其特征在于：包括如下步骤：S1：采集室外不同场景的图片；S2：基于Yolov3网络搭建深度学习网络；S3：根据YOLO3采用K‑means聚类得到先验框的尺寸；S4.得到训练后的数据集；S5.把S4步骤获得的经过训练的数据集进行视觉相关网络VD‑Net进行判别训练，得到图像判别结果；S6、根据S5步骤所获得的图像判别结果，利用场景真实语义的符合度进行对比，判断是否真正获取了场景的真实内容。本发明具有识别准确率高的优点。 【EN】In order to solve the problem of inaccurate scene inference result caused by difficulty in inferring a real scene in the prior art, the invention provides an outdoor scene safety monitoring method based on a vision correlation discrimination network, which is characterized in that: the method comprises the following steps: s1: collecting pictures of different outdoor scenes; s2: building a deep learning network based on a Yolov3 network; s3: obtaining the size of a prior frame by adopting K-means clustering according to YOLO 3; s4, obtaining a trained data set; s5, performing discrimination training on the trained data set obtained in the step S4 through a visual correlation network VD-Net to obtain an image discrimination result; and S6, comparing the image discrimination results obtained in the step S5 by using the conformity of the scene real semantics, and judging whether the real content of the scene is really acquired. The invention has the advantage of high identification accuracy.

摘要：【中文】本发明提供一种基于Hadoop的清洁型云机器人系统，包括依次连接进行信息交互的机器人、通信模块和Hadoop云服务端；通信模块负责机器人与Hadoop云服务端之间的信息传输；机器人负责执行清洁任务以及采集所处环境的图像信息、环境中的实物距离信息、机器人所在的位置信息，并将信息通过通信模块发送给Hadoop云服务端；Hadoop云服务端用来实现实时信息的采集和传输以及确定小车与周围事物的实际距离，然后为机器人规划最优路线；并将最优线路信息通过通信模块发送给机器人。本发明采用了Hadoop为云服务架构，可以根据机器人个数对树莓派集群增加数量进行扩张，由于树莓派开发板价格低廉，所以能够降低成本。 【EN】The invention provides a cleaning type cloud robot system based on Hadoop, which comprises a robot, a communication module and a Hadoop cloud server, wherein the robot, the communication module and the Hadoop cloud server are sequentially connected for information interaction; the communication module is responsible for information transmission between the robot and the Hadoop cloud server; the robot is responsible for executing cleaning tasks, collecting image information of an environment, physical distance information of the environment and position information of the robot, and sending the information to the Hadoop cloud server through the communication module; the Hadoop cloud server is used for acquiring and transmitting real-time information, determining the actual distance between the trolley and the surrounding objects and planning an optimal route for the robot; and the optimal line information is sent to the robot through the communication module. The invention adopts Hadoop as a cloud service framework, can expand the increased number of raspberry group clusters according to the number of robots, and can reduce the cost due to the low price of the raspberry group development board.

摘要：【中文】本申请公开了一种WLAN室内定位方法和装置，对获取的目标室内的第一环境信息数据进行主成分分析，选取满足预置条件的累计贡献率对应的第一环境信息数据为第二环境信息数据；基于第二环境信息数据构建目标室内模型；在目标室内模型中布置AP和设置参考点，获取每个参考点的信号强度RSSI值；基于每个参考点的信号强度RSSI值，根据模糊聚类算法对目标室内的环境进行区域划分，得到若干个子区域；将获取的子区域中的测试点的信号强度RSSI值输入到子区域对应的预置卷积神经网络模型，输出测试点的定位结果，解决了现有的室内定位方法采用欧氏距离进行搜索定位所存在的定位速度慢和定位精度低的技术问题。 【EN】The application discloses a WLAN indoor positioning method and a device, which are used for carrying out principal component analysis on acquired first environmental information data in a target room and selecting the first environmental information data corresponding to the accumulated contribution rate meeting preset conditions as second environmental information data; constructing a target indoor model based on the second environmental information data; arranging APs and setting reference points in the target indoor model, and acquiring a signal strength RSSI value of each reference point; based on the signal strength RSSI value of each reference point, carrying out region division on the environment in the target room according to a fuzzy clustering algorithm to obtain a plurality of sub-regions; the obtained signal strength RSSI value of the test point in the subregion is input into a preset convolutional neural network model corresponding to the subregion, and the positioning result of the test point is output, so that the technical problems of low positioning speed and low positioning precision existing in the existing indoor positioning method for searching and positioning by adopting Euclidean distance are solved.

摘要：【中文】本发明涉及信号处理技术领域，提出一种基于欠定卷积混叠模型的稀疏盲分离方法，包括以下步骤：获取欠定混叠语音信号；对所述欠定混叠语音信号进行短时傅里叶变换，得到频域上的稀疏混叠信号；对所述频域上的稀疏混叠信号进行数学建模，得到欠定卷积混叠模型；在所述欠定卷积混叠模型下建立稀疏代价函数，利用盲分离技术对混叠通道进行实时更新，得到估计的源信号；对所述估计的源信号进行尺度和排序处理，再利用傅里叶变换的逆运算得到时域上的完成分离的源信号。本发明利用源信号的稀疏约束以及欠定卷积混叠模型的构造，在处理真实环境下的高混响混叠信号具有更明显的优势。 【EN】The invention relates to the technical field of signal processing, and provides a sparse blind separation method based on an underdetermined convolution aliasing model, which comprises the following steps: acquiring an underdetermined aliasing voice signal; carrying out short-time Fourier transform on the underdetermined aliasing voice signal to obtain a sparse aliasing signal on a frequency domain; performing mathematical modeling on the sparse aliasing signal on the frequency domain to obtain an underdetermined convolution aliasing model; establishing a sparse cost function under the underdetermined convolution aliasing model, and updating an aliasing channel in real time by using a blind separation technology to obtain an estimated source signal; and carrying out scale and sequencing processing on the estimated source signals, and then obtaining the source signals which are separated in a time domain by utilizing the inverse operation of Fourier transform. The method has more obvious advantages in processing the high reverberation aliasing signal under the real environment by utilizing the sparsity constraint of the source signal and the construction of the underdetermined convolution aliasing model.

摘要：【中文】本发明涉及一种变速箱壳体试漏封堵机构，包括定位夹具，以及盖设于所述定位夹具上的封堵装置；所述定位夹具包括夹具底座，突出设于所述夹具底座顶部上的环形限位块，对应设于所述环形限位块顶部上的环形密封圈，以及突出设于所述夹具底座顶部上、并围设于所述环形限位块中的封堵底座结构；所述封堵装置包括与所述夹具底座对应的封堵顶座，突出设于所述封堵顶座底部四周、并与所述夹具底座顶部抵接的多个支撑柱，以及突出设于所述封堵顶座底部、并与所述封堵底座结构对应的封堵组件结构。本发明旨在解决传统技术中封堵可靠性差，合格件被误判为泄漏件的问题。 【EN】The invention relates to a leakage test plugging mechanism for a gearbox shell, which comprises a positioning clamp and a plugging device covered on the positioning clamp; the positioning clamp comprises a clamp base, an annular limiting block, an annular sealing ring and a plugging base structure, wherein the annular limiting block is arranged on the top of the clamp base in a protruding mode, the annular sealing ring is correspondingly arranged on the top of the annular limiting block, and the plugging base structure is arranged on the top of the clamp base in a protruding mode and is enclosed in the annular limiting block; the plugging device comprises a plugging top seat corresponding to the clamp base, a plurality of support columns protruding and arranged on the periphery of the bottom of the plugging top seat and abutted against the top of the clamp base, and a plugging assembly structure protruding and arranged on the bottom of the plugging top seat and corresponding to the plugging base structure. The invention aims to solve the problems that the plugging reliability is poor and qualified parts are misjudged as leakage parts in the prior art.

摘要：【中文】本申请公开了一种基于RBF神经网络的自动对焦方法及装置，包括：获取物镜所处位置拍摄的图像，根据图像计算对应的聚焦评价值以及平均灰度值；将所述聚焦评价值以及平均灰度值输入到预设的RBF神经网络模型中，RBF神经网络模型输出物镜处于最佳聚焦点的位置；利用变焦电机驱动物镜调节至所述最佳聚焦点的位置。本申请能有效避免现有对焦搜索算法中局部峰值的影响，且能有效减少变焦电机来回运动的频率，缩短了对焦时间。 【EN】The application discloses an automatic focusing method and device based on a RBF neural network, comprising the following steps: acquiring an image shot by the position of an objective lens, and calculating a corresponding focusing evaluation value and an average gray value according to the image; inputting the focusing evaluation value and the average gray value into a preset RBF neural network model, and outputting the position of an objective lens at the optimal focusing point by the RBF neural network model; and driving the objective lens to adjust to the position of the optimal focusing point by using a zooming motor. The method and the device can effectively avoid the influence of local peak values in the existing focusing search algorithm, can effectively reduce the frequency of the back-and-forth movement of the zooming motor, and shorten the focusing time.

摘要：【中文】本发明公开了一种离线指纹库区域划分方法、计算机设备、存储介质，所述方法包括以下步骤：在离线数据库构建阶段，在目标定位区域内，按比例划分出若干个一级分区，每个一级分区中都能在离线数据库中找到唯一标识该分区的至少一个一级特征AP；将每个一级分区划分若干个二级分区，每个二级分区中都能在离线数据库中找到唯一标识该分区的至少一个二级特征AP；将每个二级分区划分若干个三级分区，每个三级分区中都能在离线数据库中找到唯一标识该分区的至少一个三级特征AP；经过多级分区后，将目标定位区域划分出N级分区；在在线定位阶段，定位算法首先将待定位终端收集到的AP信息逐级与分区中的特征AP信息进行匹配，实现完成AP指纹数据库的遍历。 【EN】The invention discloses an off-line fingerprint database area dividing method, computer equipment and a storage medium, wherein the method comprises the following steps: in the off-line database construction stage, a plurality of primary partitions are divided in proportion in a target positioning area, and at least one primary feature AP which uniquely identifies the partition can be found in the off-line database in each primary partition; dividing each primary partition into a plurality of secondary partitions, wherein each secondary partition can find at least one secondary feature AP uniquely identifying the partition in an offline database; dividing each secondary partition into a plurality of tertiary partitions, wherein at least one tertiary characteristic AP uniquely identifying the partition can be found in an offline database in each tertiary partition; after multi-level partitioning, dividing the target positioning area into N-level partitions; in the on-line positioning stage, the positioning algorithm firstly matches the AP information collected by the terminal to be positioned with the characteristic AP information in the partition step by step, so as to complete traversal of the AP fingerprint database.

摘要：【中文】本发明公开了一种基于spark框架的短期电力负荷快速预测方法，该方法将训练两个模型，一个是使用BIRCH并行化算法对历史负荷数据和天气数据进行聚类，得到一个用于异常检测的模型，另一个是使用基于spark技术的lightGBM算法对历史负荷数据和天气数据进行训练，得到一个负荷预测模型，然后这两个模型发送至Spark Streaming集群，用于对实时数据流的聚类和预测；在对实时数据流的聚类和预测中使用kafka集群接收从各种终端发送过来的电力负荷数据流，并将数据流传送到Spark Steaming集群处理，在Spark Steaming集群上完成实时特征提取及归一化处理，使用异常检测模型进行实时聚类，以发现是否有异常数据，然后利用非异常的负荷数据使用负荷预测模型预测下一个时间段的负荷值。 【EN】The invention discloses a Spark framework-based short-term power load rapid prediction method, which comprises the steps of training two models, wherein one model is used for clustering historical load data and weather data by using a BIRCH parallelization algorithm to obtain a model for anomaly detection, the other model is used for training the historical load data and the weather data by using a lightGBM algorithm based on Spark technology to obtain a load prediction model, and then the two models are sent to a Spark Streaming cluster to be used for clustering and predicting real-time data streams; in the process of clustering and predicting the real-time data streams, a kafka cluster is used for receiving power load data streams sent from various terminals, the data streams are transmitted to a Spark Steaming cluster for processing, real-time feature extraction and normalization processing are completed on the Spark Steaming cluster, an anomaly detection model is used for real-time clustering to find whether abnormal data exist or not, and then the load prediction model is used for predicting the load value of the next time period by using non-abnormal load data.

摘要：【中文】本发明涉及一种双功能型紫外线吸收剂及其制备方法，其克服了现有技术中光稳定剂和紫外线吸收剂需要复配使用的问题，使化合物具有紫外线吸收和自由基捕获两种功能，起到一剂多用的效果。本发明结构如下式所示，该结构通过紫外线吸收剂UV‑326为原料，采用2,2,6,6‑四甲基‑4‑哌啶醇、4‑氨基‑2,2,6,6‑四甲基哌啶及其衍生物作为取代基，修饰得到一类含有受阻胺结构的紫外线吸收剂。该化合物具有苯并三唑和受阻胺结构，导致其具有紫外线吸收和自由基捕获的双重功能。本发明的制备方法：将四甲基哌啶醇/胺衍生物、紫外线吸收剂UV‑326、催化剂和溶剂混合，反应24h，冷却洗涤，蒸除溶剂，重结晶即可得到产物。。 【EN】The invention relates to a bifunctional ultraviolet absorbent and a preparation method thereof, which overcome the problem that a light stabilizer and an ultraviolet absorbent need to be compounded and used in the prior art, and enable a compound to have two functions of ultraviolet absorption and free radical capture, thereby achieving the effect of one agent with multiple purposes. The structure of the invention is shown as the following formula, and the ultraviolet absorbent containing hindered amine structure is obtained by modifying an ultraviolet absorbent UV-326 serving as a raw material by using 2,2,6, 6-tetramethyl-4-piperidinol, 4-amino-2, 2,6, 6-tetramethyl piperidine and derivatives thereof as substituents. The compound has a benzotriazole and hindered amine structure, so that the compound has double functions of ultraviolet absorption and free radical capture. The preparation method comprises the following steps: mixing the tetramethyl piperidinol/amine derivative, the ultraviolet absorbent UV-326, the catalyst and the solvent, reacting for 24 hours, cooling and washing, evaporating the solvent, and recrystallizing to obtain the product.。