摘要：【中文】本申请实施例提供了一种投屏方法，用于计算机设备中，所述方法包括：获取所述计算机设备的第一能力信息；获取目标终端的第二能力信息，其中，所述目标终端被配置为所述计算机设备的投屏对象；根据所述第一能力信息和所述第二能力信息，确定所述计算机设备中的多媒体内容的渲染操作执行方；如果所述计算机设备为所述渲染操作执行方，则对所述多媒体内容进行渲染操作；发送渲染后的多媒体内容至所述目标终端。本实施例所述的投屏方法可以根据投屏方和投屏对象的能力，动态确定特效渲染操作的执行方，避免了传统投屏操作中特效渲染不佳导致的播放效果不佳的问题，从而有效地提高了观影体验。 【EN】The embodiment of the application provides a screen projection method, which is used in computer equipment and comprises the following steps: acquiring first capability information of the computer equipment; acquiring second capability information of a target terminal, wherein the target terminal is configured as a screen projection object of the computer equipment; determining a rendering operation executing party of the multimedia content in the computer equipment according to the first capability information and the second capability information; if the computer equipment is the rendering operation executing party, performing rendering operation on the multimedia content; and sending the rendered multimedia content to the target terminal. The screen projection method can dynamically determine the execution party of the special effect rendering operation according to the screen projection party and the capability of the screen projection object, and avoids the problem of poor playing effect caused by poor special effect rendering in the traditional screen projection operation, so that the viewing experience is effectively improved.

摘要：【中文】本发明提供一种基于Labview的产品缺陷检测方法。本发明的方法利用Labview创建一套机器视觉检测系统，该系统将Labview与PLC进行通信，当检测到产品到达相应的位置后，会发出相应的信号给计算机，将信号传给Labview系统后，系统自动对产品进行拍照，并将获得的照片与模板图像进行比较，得出他们的相似度，如果相似度达到了用户的设定值，则认为产品合格，否则认为产品不合格。本发明方法可以全自动的实现对产品缺陷的检测，彻底改变了以往的人工检测方法，可以有效提高产品的质量，减少次品率。 【EN】The invention provides a product defect detection method based on Labview. The method of the invention utilizes Labview to create a set of machine vision detection system, the system communicates Labview and PLC, when detecting that the product reaches the corresponding position, the system sends corresponding signals to the computer, and after sending the signals to the Labview system, the system automatically takes pictures of the product, and compares the obtained pictures with the template image to obtain the similarity of the products, if the similarity reaches the set value of the user, the product is qualified, otherwise the product is unqualified. The method can realize the full-automatic detection of the product defects, thoroughly changes the prior manual detection method, can effectively improve the product quality and reduce the defective rate.

摘要：【中文】本发明涉及一种提高便携设备定位精度的方法。本发明利用多个普通GPS/BDS芯片同时进行定位，将每个普通GPS/BDS芯片的数据不同的权重进行融合，然后用卡尔曼滤波器对融合后的数据进行滤波，从而提高便携设备定位精度。本发明的多GPS/BDS芯片结构的定位精度有明显的优化，并且大大克服了普通GPS/BDS芯片受到干扰产生漂移的问题。加权方法用GPS/BDS芯片间的相对几何位置对检测数据进行约束，从而解决了多个GPS数据间的加权问题，更加可以削弱漂移数据的影响。没有复杂的数学计算，既有利于工程实现，又有利于减轻处理器的运算负担。 【EN】The invention relates to a method for improving the positioning accuracy of portable equipment. According to the invention, a plurality of common GPS/BDS chips are used for positioning at the same time, different weights of data of each common GPS/BDS chip are fused, and then a Kalman filter is used for filtering the fused data, so that the positioning precision of the portable equipment is improved. The positioning accuracy of the multi-GPS/BDS chip structure is obviously optimized, and the problem that the common GPS/BDS chip is interfered to generate drift is greatly solved. The weighting method uses the relative geometric position between the GPS/BDS chips to restrain the detection data, thereby solving the weighting problem among a plurality of GPS data and weakening the influence of drifting data. And no complex mathematical calculation is needed, so that the method is not only beneficial to engineering realization, but also beneficial to reducing the operation burden of the processor.

摘要：【中文】本发明提供一种基于多尺度特征注意机制的快速显著性检测方法。本发明方法首先通过深度卷积网络对图像进行处理，获得基于不同卷积层的特征，包括较浅层特征和高级语义特征，然后分别对获得基于不同卷积层的特征进行处理，最后将处理后的较浅层特征和高级语义特征输入至解码器中，生成显著性检测图。本发明消除了大部分背景特征的干扰，增加了计算效率并有效地抑制背景信息，对高级语义特征使用金字塔扩张卷积更好的利用语义信息，并采用双解码器结能对特征进行进一步细化，最终生成的显著性图能够以明晰的边界完整的凸显图像中的显著性区域，并有效地抑制背景区域。 【EN】The invention provides a rapid significance detection method based on a multi-scale feature attention mechanism. The method comprises the steps of processing an image through a deep convolutional network to obtain features based on different convolutional layers, wherein the features comprise shallow layer features and high-level semantic features, processing the features based on the different convolutional layers respectively, and inputting the processed shallow layer features and the high-level semantic features into a decoder to generate a significance detection graph. The method eliminates the interference of most background features, increases the calculation efficiency, effectively inhibits the background information, better utilizes the semantic information by using pyramid expansion convolution for high-grade semantic features, can further refine the features by adopting a double-decoder structure, and can finally generate the saliency map to highlight the saliency area in the image with clear and complete boundaries and effectively inhibit the background area.

摘要：【中文】本发明公开了一种提高TOF相机对运动模糊鲁棒性的方法。本发明步骤如下：1、利用深度传感器的双电容在一个积分时间内采集到的4个电容值Q₁～Q₄进行模糊检测；Q₁～Q₄分别为控制信号C₁～C₄在积分时间内收集到的电荷值；由于四相控制信号C₁～C₄相位差为90°，若没发生运动模糊，则需满足Q₁+Q₂＝Q₃+Q₄；2、若检测到的电容值满足Q₁+Q₂＝Q₃+Q₄，则使用Q₁～Q₄进行步骤5的深度计算；反之进入步骤3；3、通过反射回来的红外信号的斜率，并比较Q₁+Q₂和Q₃+Q₄的大小，得到由于运动模糊导致相位混合而错误的一对电容值；4、根据结果分别进入对应的电容值替换器，通过另一对正常电容值替换错误电容值进行后续的深度计算；5、使用Q₁～Q₄和三角函数进行深度计算。本发明简单快速且不会丢失任何原始深度图像细节。 【EN】The invention discloses a method for improving motion blur robustness of a TOF camera. The invention comprises the following steps: 1. 4 capacitance values Q acquired in one integration time by using double capacitors of depth sensor₁～Q₄Carrying out fuzzy detection; q₁～Q₄Are respectively a control signal C₁～C₄The charge value collected over the integration time; due to the four-phase control signal C₁～C₄The phase difference is 90 DEG, and if no motion blur occurs, Q is satisfied₁+Q₂＝Q₃+Q₄(ii) a 2. If the detected capacitance value satisfies Q₁+Q₂＝Q₃+Q₄Then use Q₁～Q₄Performing the depth calculation of the step 5; otherwise, entering the step 3; 3. by the slope of the reflected infrared signal and comparing Q₁+Q₂And Q₃+Q₄Is obtained due to movementBlurring a pair of capacitance values that are wrong due to phase mixing; 4. respectively entering corresponding capacitance value replacers according to the results, and performing subsequent depth calculation by replacing the error capacitance value with the normal capacitance value; 5. using Q₁～Q₄And performing depth calculation by using a trigonometric function. The method is simple and quick and does not lose any original depth image details.

摘要：【中文】本发明公开了一种基于卷积神经网络的自适应分组卷积模块设计方法。本发明步骤如下：1：对全局特征进行加和求平均处理；2：通过全卷积操作将输入特征的通道数转换为指定大小的通道数，输出特征Ⅰ；同时通过全卷积操作将输入特征通道数不变，但大小改变的特征Ⅱ；3：将步骤2的特征Ⅰ作为softmax函数输入，得到一组概率分布；并按照此概率分布将原始输入特征和得到的卷积核进行分组；4：将分组后的原始输入特征通过分组后的卷积核进行特征提取；对完成特征提取的各个分组得到的特征通过通道拼接进行融合，且融合后的通道数与原始输入特征的通道数一致。本发明设计的模块在卷积神经网络对特征进行提取的过程中，能够自适应的将输入特征进行分组提取。 【EN】The invention discloses a design method of a self-adaptive grouping convolution module based on a convolution neural network. The invention comprises the following steps: 1: carrying out addition and averaging processing on the global features; 2: converting the channel number of the input feature into a channel number with a specified size through full convolution operation, and outputting a feature I; meanwhile, inputting a feature II with the same number of feature channels and changed size through full convolution operation; 3: inputting the characteristic I in the step 2 as a softmax function to obtain a group of probability distributions; grouping the original input features and the obtained convolution kernels according to the probability distribution; 4: performing feature extraction on the grouped original input features through the grouped convolution kernels; and fusing the features obtained by each group after the feature extraction is completed through channel splicing, wherein the number of the fused channels is consistent with that of the original input features. The module designed by the invention can adaptively extract the input features in groups in the process of extracting the features by the convolutional neural network.

摘要：【中文】本发明提供一种基于语义分割和深度学习的图片背景风格迁移方法。本发明首先选择内容图片和风格图片并进行图片预处理；然后通过ResNet网络由内容图片和风格图片直接计算的到一张相对比较接近结果的图片；然后通过VGG‑19网络获得风格约束和内容约束，根据损失函数进行梯度下降，通过多次迭代的方式获得背景风格迁移结果，最后将迁移结果放回图片上。本发明速度提高了上百倍，可扩展性强，对局部区域进行风格迁移，保留了图像主体内容，以达到突出主体，增强图像艺术表现力的效果，代码易读性和可移植性强。 【EN】The invention provides a picture background style migration method based on semantic segmentation and deep learning. Firstly, selecting content pictures and style pictures and preprocessing the pictures; then, directly calculating a picture with a relatively close result by the content picture and the style picture through a ResNet network; and then obtaining style constraint and content constraint through a VGG-19 network, performing gradient descent according to a loss function, obtaining a background style migration result through a plurality of iterations, and finally putting the migration result back on the picture. The invention has the advantages of hundreds of times of speed improvement, strong expandability, style migration to local areas, retention of the main content of the image, prominent main body and enhanced artistic expressive force of the image, and strong code readability and transportability.

摘要：【中文】本发明公开了一种基于TDOA的目标辐射源定位方法。本发明定义相应的非视距传感器向量，根据接收到的时差参量对该非视距传感器向量进行初始化，以此对原时差参量进行更正。选取中间变量，构造目标辐射源TDOA伪线性估计模型，运用加权最小二乘估计，得到目标辐射源的位置估计。运用梯度下降算法迭代修正结果。最后，利用中间变量与目标位置估计坐标的关系，再次运用加权最小二乘估计给出最终的目标位置估计。本发明通过引入合适的中间变量，使得不易直接求解的目标辐射源TDOA量测方程转化为伪线性估计模型同时结合梯度下降算法，降低了对可能存在非视距误差对定位性能的影响。最后进一步优化了位置坐标的最终解。 【EN】The invention discloses a TDOA-based target radiation source positioning method. The invention defines the corresponding non-line-of-sight sensor vector, initializes the non-line-of-sight sensor vector according to the received time difference parameter, and corrects the original time difference parameter. And selecting intermediate variables, constructing a pseudo linear estimation model of the target radiation source TDOA, and obtaining the position estimation of the target radiation source by using weighted least square estimation. And (5) iteratively correcting the result by using a gradient descent algorithm. And finally, giving out the final target position estimation by using the relation between the intermediate variable and the target position estimation coordinate and applying the weighted least square estimation again. According to the method, by introducing appropriate intermediate variables, the TDOA measurement equation of the target radiation source which is difficult to directly solve is converted into a pseudo linear estimation model, and the gradient descent algorithm is combined, so that the influence on the positioning performance caused by possible non-line-of-sight errors is reduced. And finally, the final solution of the position coordinates is further optimized.

摘要：【中文】本发明提供本发明一种基于图注意力网络的社交机器人识别方法。本发明方法基于图注意力网络，通过对社交网络上发布的内容进行自然语言处理构建节点特征，各社交账号之间的转发、评论关系来构建图，然后进行分类，从而判断出该账号是否为社交机器人。首先社交网络数据，进行数据集的创建，然后构建图注意力网络，通过创建的数据集进行图注意力网络的训练和测试。针对复杂的社交网络机器人识别问题，本发明方法能够自动高效的识别社交机器人，减少不法分子的可乘之机，从而限制机器人发布的言论，削弱不良社会舆论影响，有利于维护社会和谐稳定。 【EN】The invention provides a social robot identification method based on a graph attention network. The method is based on a graph attention network, the natural language processing is carried out on the content published on the social network to construct node characteristics, the forwarding and commenting relations among the social account numbers are used for constructing a graph, and then the graph is classified, so that whether the account number is a social robot or not is judged. The method comprises the steps of firstly, social network data are generated, a data set is created, then a graph attention network is built, and training and testing of the graph attention network are conducted through the created data set. Aiming at the problem of complex social network robot identification, the method can automatically and efficiently identify the social network robot and reduce the opportunities of lawless persons, thereby limiting the language issued by the robot, weakening the influence of bad social public opinion and being beneficial to maintaining social harmony and stability.

摘要：【中文】本发明属于多糖提取技术领域，具体涉及一种褐藻糖胶的提取方法及其应用。所述方法包括粗提和提纯，所述粗提过程为将原材料粉碎，用醋酸水溶液先浸泡提取，过滤，保留滤液，并将滤饼在醋酸水溶液中加热至30‑40℃搅拌二次提取，过滤，将两次滤液合并，调至中性，进一步加入乙酸乙酯萃取，其中有机层提取了色素，分离的水层用乙醇沉淀法，获得褐藻糖胶的粗提物。本发明提供的技术方案在粗提阶段，在较温和的温度和pH值条件下最大限度避免了褐藻糖胶的破坏，硫酸盐含量较高，另外在粗提阶段进行了脂溶性特别是色素物质的萃取，这一操作不仅使得褐藻多糖得率有所提升，并且为褐藻黄素的回收利用提供了可能，增加了原材料的综合利用度。 【EN】The invention belongs to the technical field of polysaccharide extraction, and particularly relates to an extraction method and application of fucoidan. The method comprises crude extraction and purification, wherein the crude extraction process comprises the steps of crushing raw materials, soaking and extracting with acetic acid aqueous solution, filtering, retaining filtrate, heating filter cakes in the acetic acid aqueous solution to 30-40 ℃, stirring and extracting for two times, filtering, combining the two filtrates, adjusting to be neutral, further adding ethyl acetate for extraction, extracting pigment from an organic layer, and precipitating a separated water layer with ethanol to obtain a crude extract of fucoidan. The technical scheme provided by the invention avoids the damage of fucoidan to the maximum extent under the conditions of mild temperature and pH value in the crude extraction stage, has higher sulfate content, and extracts fat-soluble and particularly pigment substances in the crude extraction stage, so that the brown algae polysaccharide yield is improved, the possibility of recycling fucoxanthin is provided, and the comprehensive utilization rate of raw materials is increased.