30results about How to "Complementary effect is good" patented technology

A traffic missing data completion method based on the spatio-temporal attention mechanism. First, through the attention mechanism, capture the degree of influence of all road sections in the road network on the traffic state of the road network at the current moment, and recapture the spatial data at different times. Relevance information to improve the accuracy of data completion. Secondly, considering the timing of traffic data, traffic data at different times have different influences on the data at the current moment. This inconsistent time correlation information is captured through the time attention mechanism and retained when completing the current missing data. The most effective information improves the completion effect of the model. Finally, while using the spatiotemporal attention mechanism to capture the spatiotemporal correlation of traffic data, considering that the correlation between data is attenuated by the increase of spatial distance and time interval, adding a spatiotemporal attenuation matrix improves the completion accuracy. The present invention not only greatly improves the completion accuracy in the case of low data missing rate, but also improves the completion accuracy in the case of high data missing rate.

The embodiment of the present invention discloses a data processing method, device, equipment and medium. The data processing method includes: performing grid processing on the target area according to the digital map information of the target area to obtain S grid areas of equal area , S is an integer greater than or equal to 1; according to the position data of the target area acquired by the base station in the preset first time period, determine the amount of moving points in each grid area; for all grids in the preset first time period The fragment moving point amount missing in the grid area is complemented to obtain the complete moving point amount of the target area. The fragment moving point amount is the moving point amount of a grid area within a preset time length, and the preset time length is A time period within the first time period is set.

The invention discloses a time series data completion method based on distance matrix, which mines and utilizes the inherent high-order time correlation of time series data to complete missing data with similar data points in the time series data; the method specifically Including: for the time series data, the distance matrix D of the time series is modeled based on a certain distance metric function, wherein the matrix element Dij located in the i-th row and the j-th column is the i-th data point and the j-th data point in the time series. The distance between data points; based on the obtained distance matrix D, find the k segments with the closest distance to the missing segment in the original time series; use the calculated k nearest neighbor segments to complete the missing segment data. This method can achieve a good completion effect in real time series data missing scenarios. At the same time, this method is highly interpretable and the physical meaning behind it is clearer, so it can be expanded on the basis of this method. , so that it can be effectively used in various real scenarios.

The invention discloses a three-dimensional spectrum situation completion method based on a generative confrontation network, which includes: acquiring and storing historical or empirical radio monitoring data off-line, and obtaining several complete three-dimensional spectrum situation or field strength training data; and iterating on the variant of the generative confrontation network And adversarial off-line training, get a variant of the generative adversarial network that has learned the 3D spectrum situation or field strength complement mechanism; collect the actual radio monitoring data of the current 3D target area online, and preprocess to get the 3D target area defect spectrum situation or field strength Measured data: Input the obtained measured data into the variant of the generated confrontation network to obtain the output data of the variant of the generated confrontation network, and process it to obtain the current three-dimensional target area to complete the spectrum situation or field strength. The invention is oriented to the application of the three-dimensional spectrum situation or field strength in the space-space-ground information network, and can effectively reduce the complement error of the three-dimensional spectrum situation or field strength.

The invention relates to an image completion method based on a stacked generation confrontation network. The network structure is mainly composed of three stacked generator network layers. First, the mask image is cut into multiple image blocks so that the network can extract the features of different image blocks; then, the present invention puts the completed multi-image block results into the generator of the next layer to further complete image; finally, the completion results of different blocks are applied to a whole mask image to obtain the final completion output. From coarse to fine completion, make full use of the high-level semantic information extracted by the convolutional neural network. And through the image block discriminator to distinguish the real and fake of the generated image and the original image. Experimental results show that the method of the present invention can generate high-quality completion results for images with irregular masks, and the completion results are closer to the original image.

The invention relates to a knowledge base completion method based on multi-modal representation learning. Given a knowledge base KB, the KB includes two parts, one is a known knowledge set, and the other is an unknown knowledge set; Data preprocessing of the data; a knowledge base completion model ConvAt is proposed, and the multimodal representation of the head entity and tail entity is first generated for the acquired data; then the multimodal representation of the head entity, the structural feature vector of the relationship and the tail entity The multi-modal representation of is concatenated by columns, processed by the convolutional neural network module, channel attention module and spatial attention module respectively, and finally multiplied with a weight matrix to obtain the score of the triplet (h, r, t); Use the loss function to train the completion model in step S2, and use the trained model to complete the knowledge base. The algorithm proposed by the invention can fuse external information and utilize richer semantic information.

The present invention provides a system and method for complementing a laser radar three-dimensional point cloud target. The global characteristics of the laser radar target are obtained through three layers of shared multi-layer perceptrons and three layers of point-by-point maximum pooling layers, so that the objects obtained by laser radar scanning The point cloud distribution is denser and more uniform, and the details of the outline of the object are more complete, so as to achieve the purpose of detection, recognition and measurement calculation.

The present invention discloses a method for completing tagged time series data, which is mainly used to solve the problem of missing a continuous segment of time series data common in real scenes. The core idea of ​​the data completion method includes two aspects: first, using Hankel matrix technology organizes low-dimensional time series into high-dimensional forms, introduces high-order time dependencies, and uses matrix decomposition to complete missing data on this basis, thereby effectively overcoming the problem of data loss; second, in the overall algorithm The tag information is modeled in the framework, and the tag information is used to support the data completion process, so that the completed data is more in line with the real scene. By rationally using the above two ideas, the method proposed by the present invention can achieve a better completion effect in the real time series data missing scene; at the same time, the method has strong interpretability, and can also be based on the method More expansions are carried out so that it can be effectively used in various real scenarios.

The invention discloses a dam safety monitoring data completion method based on spatiotemporal multi-view fusion. Abstract the view model; fuse the four models using lasso regression to generate a spatiotemporal multi-view fusion model; use the spatiotemporal multi-view fusion model to generate complementary data. In the case of strong correlation between temporal and spatial characteristics, this method can well solve the problems of block deletion and local deletion in dam safety monitoring data, and has been verified on real dam safety monitoring data. Algorithms and traditional spatiotemporal models have smaller errors and better completions.

The invention discloses a time series data complement method, device and electronic equipment. The method includes: constructing a data set according to a set format for the collected data, and the data set contains at least one series of data; The time point to be completed is inserted into the data set; the data set at the inserted time point is repartitioned according to the series, and the data of the same series are put into the same partition; the data in the partition is sorted to obtain the data list; Traverse the data in the partition and complete the data. Under the distributed computing framework, this method completes the missing time series data forward or backward. When performing data completion, only a small number of variables need to be traversed once to complete the data completion.