摘要：【中文】本申请涉及一种模型训练方法、装置、计算机设备和存储介质。所述方法包括：获取预训练好的初始机器学习模型和样本图像；通过初始机器学习模型对样本图像进行处理，得到预测结果；将满足低置信度条件的预测结果对应的样本图像作为第一目标图像，并对第一目标图像进行标注处理得到对应的第一标注信息；将满足高置信度条件的预测结果对应的样本图像作为第二目标图像，并根据第二目标图像的预测结果确定与第二目标图像对应的第二标注信息；将第一目标图像和第二目标图像作为样本输入，并将相应的第一标注信息和第二标注信息作为训练标签，对初始机器学习模型进行再训练。采用本方法能够提升模型训练效率。 【EN】The application relates to a model training method, a model training device, computer equipment and a storage medium. The method comprises the following steps: acquiring a pre-trained initial machine learning model and a sample image; processing the sample image through an initial machine learning model to obtain a prediction result; taking a sample image corresponding to a prediction result meeting a low confidence condition as a first target image, and performing labeling processing on the first target image to obtain corresponding first labeling information; taking the sample image corresponding to the prediction result meeting the high-confidence condition as a second target image, and determining second annotation information corresponding to the second target image according to the prediction result of the second target image; and taking the first target image and the second target image as samples to be input, taking the corresponding first labeling information and the second labeling information as training labels, and retraining the initial machine learning model. By adopting the method, the model training efficiency can be improved.

摘要：【中文】本发明提出一种清洗装置，清洗装置能够对流水线上的工件进行清洗，清洗装置包括清洗机构、存料机构、推入机构及推出机构，清洗机构设置于流水线间，存料机构包括支撑架、存料驱动件、主动轴、从动轴、旋转链轮、传动链及固定框，支撑架固定设于清洗机构上，主动轴及从动轴分别转动地设置于支撑架上，主动轴及从动轴上分别设置有旋转链轮且旋转链轮上套设传动链，主动轴能够在存料驱动件的驱动下转动从而带动套设于旋转链轮上的传动链随之转动，固定框设置于传动链上，固定框用于存放工件，清洗机构用于对固定框上的工件进行清洗，推入机构用于将进料线上的工件推入固定框中，推出机构用于将清洗完的工件从固定框中推出落料于出料线上。 【EN】The invention provides a cleaning device, which can clean workpieces on a production line and comprises a cleaning mechanism, a stock mechanism, a pushing mechanism and a pushing mechanism, wherein the cleaning mechanism is arranged between the production lines, the stock mechanism comprises a support frame, a stock driving piece, a driving shaft, a driven shaft, a rotating chain wheel, a transmission chain and a fixed frame, the support frame is fixedly arranged on the cleaning mechanism, the driving shaft and the driven shaft are respectively and rotatably arranged on the support frame, the driving shaft and the driven shaft are respectively provided with the rotating chain wheel, the transmission chain is sleeved on the rotating chain wheel, the driving shaft can rotate under the driving of the stock driving piece to drive the transmission chain sleeved on the rotating chain wheel to rotate along with the rotating chain, the fixed frame is arranged on the transmission chain, the cleaning mechanism is used for cleaning the workpieces on the fixed frame, the pushing mechanism is used for pushing the workpieces on the feeding line into the, the pushing mechanism is used for pushing the cleaned workpiece out of the fixed frame and blanking the workpiece on the discharging line.

摘要：【中文】一种可配置有序键值类数据模拟发生方法，包括有：根据预设配置表信息，生成时间序列信息；针对生成的时间序列信息，结合预设配置表信息，判断是否需要将时间序列信息转换为频域有限数量离散信息；若需要转换，则根据预设配置表信息，将时间序列信息转换为频域有限数量离散子信息；若不需要转换为离散频域子序列信息，则根据预设配置表信息，将时间序列信息划分为有序键值类时间子序列信息；将频域有限数量离散子信息或有序键值类时间子序列信息转换为有序键值类数据。 【EN】A configurable ordered key value class data simulation generation method comprises the following steps: generating time sequence information according to preset configuration table information; aiming at the generated time sequence information, judging whether the time sequence information needs to be converted into frequency domain limited discrete information or not by combining preset configuration table information; if the conversion is needed, converting the time sequence information into frequency domain limited number discrete sub-information according to the preset configuration table information; if the time sequence information does not need to be converted into the discrete frequency domain subsequence information, dividing the time sequence information into sequential key value type time subsequence information according to preset configuration table information; and converting the frequency domain limited number of discrete sub-information or the ordered key value class time sub-sequence information into ordered key value class data.

摘要：【中文】一种时间序列结构体数据的模拟发生方法，包括有实际数据分类、结构体模式规范化处理、分析结构体化数据和评判模拟四个步骤；本发明将多类型数据组合的时间序列采用统一的数据结构进行记录，便于数据存储和分析；通过结构化处理使得多类型数据组合的时间序列在组成上进行统一，可以采用同类分析方法来分析多元时间序列；通过生成的模拟结构体数据，理解发生源数据的动态系统、预测将来系统走向、指导通过干预来控制将来事件。 【EN】A simulation generation method of time series structure data comprises four steps of actual data classification, structure mode standardization processing, structuralized data analysis and judgment simulation; the invention records the time sequence of the multi-type data combination by adopting a uniform data structure, thereby facilitating data storage and analysis; the time sequences of the multi-type data combination are unified in composition through structural processing, and the multi-element time sequences can be analyzed by adopting the same type of analysis method; through the generated simulation structure data, a dynamic system of the generation source data is understood, the trend of the future system is predicted, and the future event is controlled through intervention.

摘要：【中文】本发明公开了一种基于模糊分类器的模拟电路故障诊断方法，根据故障模型编写故障字典，然后构建模糊分离器，作为分类方法，给模糊分类器一个输入，它通过存储在模糊知识库中的隶属函数将其转换为语言变量，推理引擎获得模糊输入并使用IF‑Then规则转换为模糊输出，解模糊器中的隶属函数将模糊输出转换为清晰输出。在计算了所有输入特征参数集之后，模糊分类器可以输出元件的估计值，根据故障字典可以确定故障类型，因此可以获得更为准确的故障诊断结果。相比于传统的电子产品故障诊断方法，本发明所提出的方法具有应用广泛、识别准确率高、可视化效果好等优点。 【EN】The invention discloses a fault diagnosis method for an analog circuit based on a fuzzy classifier, which is characterized in that a fault dictionary is compiled according to a fault model, Then a fuzzy separator is constructed as a classification method, the fuzzy separator is used as an input of the fuzzy classifier, the fuzzy classifier converts the input into linguistic variables through a membership function stored in a fuzzy knowledge base, an inference engine obtains the fuzzy input and converts the fuzzy input into fuzzy output by using an IF-Then rule, and the membership function in the fuzzy classifier converts the fuzzy output into clear output. After calculating all the input feature parameter sets, the fuzzy classifier can output the estimation values of the elements, and the fault type can be determined according to the fault dictionary, so that more accurate fault diagnosis results can be obtained. Compared with the traditional electronic product fault diagnosis method, the method provided by the invention has the advantages of wide application, high identification accuracy, good visualization effect and the like.

摘要：【中文】本发明公开了一种基于吸收式热泵余热回收的超临界二氧化碳发电系统，包括超临界二氧化碳布雷顿循环系统、吸收式热泵循环系统、传统燃气型布雷顿循环系统和冷却系统。冷却系统负责提供超临界二氧化碳布雷顿循环系统和吸收式热泵循环系统冷凝所需的冷量，传统燃气型布雷顿循环在生产出了整个系统的大部分电能的同时还分别给超临界二氧化碳布雷顿循环系统和吸收式热泵循环系统提供了高温热源；在降低燃气轮机排烟温度的同时，还可提高系统整体热效率；超临界二氧化碳布雷顿循环系统通过在烟气回热器中利用烟气余热加热高压透平的乏汽并使之升温后继续推动低压透平做功，从而进一步提高循环热效率，提高发电效率。 【EN】The invention discloses a supercritical carbon dioxide power generation system based on absorption heat pump waste heat recovery, which comprises a supercritical carbon dioxide Brayton cycle system, an absorption heat pump cycle system, a traditional gas type Brayton cycle system and a cooling system. The cooling system is responsible for providing the cooling capacity required by the condensation of the supercritical carbon dioxide Brayton cycle system and the absorption heat pump cycle system, and the traditional gas type Brayton cycle produces most of the electric energy of the whole system and simultaneously provides high-temperature heat sources for the supercritical carbon dioxide Brayton cycle system and the absorption heat pump cycle system respectively; the exhaust gas temperature of the gas turbine is reduced, and the overall thermal efficiency of the system can be improved; the supercritical carbon dioxide Brayton cycle system heats the exhaust steam of the high-pressure turbine by using the waste heat of the flue gas in the flue gas heat regenerator, heats the exhaust steam and then continuously pushes the low-pressure turbine to do work, thereby further improving the cycle heat efficiency and improving the power generation efficiency.

摘要：【中文】本发明公开了一种选择含碳材料添加剂对褐煤微波脱水提质后液体产物性质产生影响的方法，将褐煤原煤进行初步研磨和筛分，制得煤样；取一定量的煤样与含炭材料添加剂混合均匀，得到混合煤；将混合煤放入微波加热装置中，在惰气保护下，进行微波加热处理；同时采用冰水浴对加热过程中产生的气体进行冷却，再对收集到的液体产物进行相关理化性质分析。与无添加剂时相比，加入不同的含炭材料添加剂后，液体产物的各理化性质均发生了有规律的变化。此方法操作简单，加热时间短，耗能少，并且通过对此液体产物的分析可以间接的推测出微波脱水提质过程中褐煤理化性质上的某些变化，具有很强的实用性和广泛的适用性。 【EN】The invention discloses a method for selecting a carbonaceous material additive to influence the properties of a liquid product obtained after microwave dehydration and upgrading of lignite, which comprises the steps of primarily grinding and screening raw lignite to prepare a coal sample; uniformly mixing a certain amount of coal sample with the additive containing the carbon material to obtain mixed coal; putting the mixed coal into a microwave heating device, and carrying out microwave heating treatment under the protection of inert gas; meanwhile, the gas generated in the heating process is cooled by adopting an ice-water bath, and then the collected liquid product is subjected to related physicochemical property analysis. Compared with the case without the additive, after different carbon-containing material additives are added, the physicochemical properties of the liquid product are regularly changed. The method is simple to operate, short in heating time and low in energy consumption, and certain changes in physical and chemical properties of the lignite in the microwave dehydration upgrading process can be indirectly inferred through analysis of the liquid product, so that the method has strong practicability and wide applicability.

摘要：【中文】本发明提供了一种基于深度学习的行人检测方法、装置及介质包括：使用深度学习神经网络对图像进行处理，获取图像中的多个关键点及所述关键点的嵌入特征，其中，所述多个关键点至少包括人的脖颈部位、人的头部、人的肩部；所述嵌入特征与所述关键点的位置相关；基于所述关键点的嵌入特征之间的距离，将所述关键点进行分组，形成第一组合和第二组合，将一个或多个第一组合形成的集合与一个或多个第二组合形成的集合，进行级联匹配，匹配成功，获得检测区域。本发明解决了已有的深度学习行人检测模型中，误检漏检现象较多的问题，从而更加准确地完成监控视频下的行人检测分析这一任务，为相关的安防监控分析与应用提供更好的技术支持。 【EN】The invention provides a pedestrian detection method, a device and a medium based on deep learning, which comprises the following steps: processing the image by using a deep learning neural network, and acquiring a plurality of key points in the image and embedded features of the key points, wherein the plurality of key points at least comprise a neck part of a person, a head of the person and shoulders of the person; the embedded features are associated with the locations of the keypoints; and grouping the key points based on the distance between the embedded features of the key points to form a first combination and a second combination, carrying out cascade matching on a set formed by one or more first combinations and a set formed by one or more second combinations, and obtaining a detection area after successful matching. The invention solves the problem that the existing deep learning pedestrian detection model has more false detection and missing detection phenomena, thereby more accurately completing the task of pedestrian detection and analysis under the monitoring video and providing better technical support for related security monitoring analysis and application.

摘要：【中文】本发明提供一种处理聚丙烯酰胺废水的生物菌剂及其制备方法和应用，涉及污水处理技术领域；一种处理聚丙烯酰胺废水的生物菌剂，包括混合比例为1:1～1:3的PAM‑De‑Yan‑1#和PAM‑De‑Yan‑3#。本发明所述的生物菌剂可以提高现有PAM生物处理单元中PAM的降解率、总有机碳和化学需氧量的去除率。通过生物强化后，PAM浓度为100mg/L时PAM降解率为75％～80％，TOC去除率为55％～60％，COD去除率为65％～75％。PAM浓度为400mg/L时，PAM降解率为60％～75％，TOC去除率为45％～55％，COD去除率为65％～75％。 【EN】The invention provides a biological agent for treating polyacrylamide wastewater and a preparation method and application thereof, relating to the technical field of sewage treatment; a biological agent for treating polyacrylamide wastewater comprises PAM-De-Yan-1# and PAM-De-Yan-3# which are mixed in a ratio of 1: 1-1: 3. The biological agent can improve the degradation rate of PAM in the existing PAM biological treatment unit and the removal rate of total organic carbon and chemical oxygen demand. After biological strengthening, the PAM degradation rate is 75-80% when the PAM concentration is 100mg/L, the TOC removal rate is 55-60%, and the COD removal rate is 65-75%. When the concentration of PAM is 400mg/L, the degradation rate of PAM is 60-75%, the removal rate of TOC is 45-55%, and the removal rate of COD is 65-75%.

摘要：【中文】本发明公开了一种选择阳离子添加剂促进褐煤微波脱水过程的方法，通过在定量的褐煤中添加定量的不同的阳离子添加剂，于不同的加热时间中，比较各阳离子添加剂对促进褐煤微波脱水过程的效果，进而选择阳离子添加剂，优化褐煤微波脱水效率，提高经济效益。通过添加阳离子添加剂，增大褐煤的介电损耗，增强煤样对微波的吸收利用微波能的能力。于微波加热过程中利用接地微波屏蔽型热电偶对煤样温度进行实时测量，获得了比传统过程使用红外热电偶更为准确的煤样温度数据。通过利用不同阳离子添加剂对褐煤微波加热过程产生影响，加快了脱水提质过程，加快脱水速率，增加脱水量，提高脱水提质效果。具有操作简单，加热时间短，耗能少，效率高等优点。 【EN】The invention discloses a method for promoting microwave dehydration process of lignite by selecting cationic additive, which is characterized in that quantitative different cationic additives are added into quantitative lignite, and the effect of each cationic additive on promoting the microwave dehydration process of lignite is compared in different heating time, so that the cationic additive is selected, the microwave dehydration efficiency of lignite is optimized, and the economic benefit is improved. By adding the cationic additive, the dielectric loss of the lignite is increased, and the microwave energy utilization capability of the coal sample for absorbing microwaves is enhanced. The temperature of the coal sample is measured in real time by using the grounded microwave shielding type thermocouple in the microwave heating process, and more accurate coal sample temperature data is obtained compared with the coal sample temperature data obtained by using an infrared thermocouple in the traditional process. Different cation additives are utilized to influence the microwave heating process of the lignite, so that the dehydration quality-improving process is accelerated, the dehydration speed is accelerated, the dehydration amount is increased, and the dehydration quality-improving effect is improved. The method has the advantages of simple operation, short heating time, low energy consumption, high efficiency and the like.