44results about How to "Improve feature utilization" patented technology

The invention discloses a non-uniform motion blurred image adaptive restoration method based on an attention model, and belongs to the field of digital image / video signal processing. The invention designs a conditional generative adversarial network combined with an attention mechanism. The generation network is of a coding and decoding structure, in the coding stage, a dense connection network isadopted to extract features, the feature utilization rate is increased, feature propagation is enhanced, and a visual attention mechanism is added, so that the network can adaptively modulate networkparameters and dynamically remove image blurring for different input images. According to the method, the clear image can be effectively restored from the non-uniform motion blurred image. The technology has a wide application prospect in the fields of target tracking, traffic detection, military reconnaissance and the like.

The invention provides a method and a device of connecting multiple intelligent household electric appliances quickly, a method and a device of connecting a router quickly, a router and an intelligent household electric appliance. The method of connecting multiple intelligent household electric appliances quickly comprises the steps of sending network distribution information packets through a private protocol for multiple intelligent household electric appliances to monitor; judging whether equipment identifiers of the multiple intelligent household electric appliances are in a trusted list respectively; and when the judgment result is negative, adding the equipment identifier information into a blacklist to hold back and / or cut off network connection of trustless intelligent household electric appliances. Through the technical scheme, the network distribution process of the intelligent household electric appliances can be simplified, and multiple intelligent household electric appliances can be connected quickly without one-by-one operation of a user.

The invention belongs to the technical field of computed tomography image processing. According to the specific technical scheme, the CT image super-resolution reconstruction method based on the generative adversarial network comprises the following specific steps: 1, establishing a dense connection relationship among different residual blocks based on a multi-stage dense residual block generatornetwork; 2, adding a bottleneck layer to the front end of each dense residual block; 3, optimizing the global network by adopting the Wasserstein distance loss and the VGG feature matching loss; 4, arranging a multi-path generator based on the sequence from thick to thin; 5, generating an image based on conditional expression generative adversarial learning; 6, reconstructing a CT image super-resolution reconstruction framework of the generative adversarial network based on multiple paths of conditions from coarse to fine; 7, reconstructing a loss function. According to the method, network redundancy is reduced, feature multiplexing among different residual blocks is realized, the maximum information transmission of the network is realized, the feature utilization rate is improved, and thereconstructed image quality is greatly improved.

The invention discloses a face recognition model construction method and device, a face recognition method and device and a storage medium. The method comprises steps of preprocessing face pictures, extracting four local areas, namely a left eye area, a right eye area, a nose area and a mouth area, of the preprocessed face image to perform face partitioning; amplifying the data on the basis of a partitioning result, respectively amplifying the data by considering the shielding of 1 / 2 and 1 / 4 conditions, and finally training a deep learning neural network by using the amplified data to construct a training feature data set, thereby realizing the face recognition based on feature matching. The method is good in shielded face recognition effect, does not need a large number of shielded samples, and is small in occupied memory.

The invention discloses a tunnel diffusion draught fan, which comprises a tower, a worktable arranged on the top of the tower, a lower bearing disk arranged on the worktable, an upper bearing disk positioned on the lower bearing disk and a cabin arranged on the upper bearing disk, wherein the upper bearing disk is provided with a supporting bracket which is provided with a tunnel diffusion draught fluid channel with a round cross section; a diffuser is arranged in the tunnel diffusion draught fluid channel; a front diffusion and acceleration section three-dimensional channel, a three-dimensional fluid acting channel and a rear post-action air flow three-dimensional fluid channel, which are communicated in turn, are arranged in the diffuser; and the three-dimensional fluid acting channel is provided with at least two impellers along the axial direction of the fan, and a working machine positioned in the cabin is connected with the impellers through a transmission mechanism. The tunnel diffusion draught fan has high kinetic energy utilization rate and reduced cost and is convenient and safe in manufacturing, transportation, installation and maintenance.

The invention discloses a target detection method based on a dense connection deep network, belonging to the technical field of target detection. According to the target detection method based on thedense connection deep network, a dense connection mode is fused into the yolo-tiny network, a convolution layer is increased, and a feature extraction network is improved; and the improved network firstly normalizes an input image into a fixed size, then uses a DenseBlock module to extract and fuse the characteristics of each channel, and then uses different priors anchors for prediction on different scales to complete the classification and positioning of a target. Compared with an original algorithm, the improved algorithm has the advantages that precision is improved by 15%, and the requirement of real-time detection can still be met; and the size of a model is only 44.7 MB, so the requirements on memory occupation and real-time performance in actual use can be met.

The invention discloses a CycleGAN-based image training network structure ArcGAN and a method, the network structure ArcGAN is composed of a generator and double discriminators, and the double discriminators comprise a rough discriminator and a fine discriminator, wherein the encoder comprises an input layer and three down-sampling convolution layers, and each down-sampling convolution layer is connected with two flat convolution layers with the same structure as the input layer; a converter comprises five dense convolution blocks without pooling layers, each dense convolution block comprisesfive dense convolution layers with bottleneck layers, and a compression layer is arranged between the blocks, wherein the decoder comprises three upsampling deconvolution layers and an output layer, and each upsampling deconvolution layer is connected with two flat convolution layers with the same structure as the input layer; and each layer of downsampling in the encoder and the upsampling in thedecoder corresponding to the layer of downsampling are in replicated connection. A rough discriminator is used for processing high-level visual information and consists of six down-sampling layers and an output layer; the calculation loss of the fine discriminator is combined with the calculation loss of the rough discriminator, and the fine discriminator and the rough discriminator jointly complete confrontation consistency training.

The invention relates to a method for accessing WiFi through scanning a two-dimensional code. The method comprises the steps of: initiating a scanning program to scan a WiFi two-dimensional code which contains access information for accessing the WiFi and guiding information for guiding the initiation of a target scanning program; analyzing the WiFi two-dimensional code to obtain access information if the current scanning program is a target program, and entering a page authorizing connection to the WiFi; analyzing the WiFi two-dimensional code to obtain the guiding the initiation if the current scanning program is not the target program, initiating the target scanning program to analyze the WiFi two-dimensional code to obtain access information, and entering the page authorizing connection to the WiFi according to the access information; and receiving a connection command of a user, and accessing the WiFi according to the access information of the WiFi. In addition, the invention further provides a device for accessing WiFi through scanning the two-dimensional code. The method and the device for accessing the WiFi through scanning the two-dimensional code can add a sharing function to the scanning program, and enhance the function utilization rate of the scanning program.

The invention discloses a hyperspectral remote sensing image classification method and system based on multi-saliency feature fusion, and relates to the technical field of image classification, and the technical scheme is that the method comprises the steps of: carrying out the dimension reduction of an original hyperspectral remote sensing image; obtaining a morphological feature map of the dimension-reduced image by using an extended morphological method EMP; for the dimension-reduced image and the morphological feature map, extracting texture features and gradient features by using a local binary pattern (LBP) and a histogram of gradient (HOG), and obtaining texture and gradient-based saliency features by using a saliency BMS based on Boolean mapping; and carrying out operations of fusing the texture and gradient-based saliency features and then integrally fusing all the saliency features, and carrying out pixel point classification by using a random forest classification algorithm according to a final fusion result. According to the method, the feature utilization rate in the classification process can be improved, and the problems of difficulty in model parameter adjustment, easiness in overfitting, high training cost and the like in the current hyperspectral remote sensing image classification method based on the deep neural network are solved.

The invention is applicable to the technical field of intelligent segmentation of pipeline diseases, and relates to a Mask RCNN-based underground drainage pipeline disease pixel level detection method, which comprises the following steps of: making an underground drainage pipeline instance segmentation data set by utilizing an acquired drainage pipeline disease video; a loss function and an ROI pooling layer in a Mask R-CNN deep learning architecture being optimized, and a ResNet101 network being used as a pipeline disease feature extraction network, so that the detection precision of an instance segmentation algorithm is improved; initializing network model parameters by using a transfer learning technology, performing a hyper-parameter tuning test on the network model, and starting network model training; evaluating the performance of the training network model, and analyzing the intersection-combination ratio of the network model prediction disease area and the real disease area; and judging whether the network model can achieve a pixel-level segmentation effect or not. According to the method, the Mask R-CNN instance segmentation framework, the ResNet101 residual neural network and the drainage pipeline disease big data are combined, so that rapid, accurate and automatic identification and positioning of the underground drainage pipeline disease are realized.

The invention relates to a preparation method of broad bean aqueous extract feed, a product and application of the broad bean aqueous extract feed in embrittlement of fish muscles. The method comprises the following steps: preparation of broad bean powder, preparation of a saturated calcium hydroxide solution, preparation of 4M hydrochloric acid, extraction of a broad bean aqueous extract, preparation of broad bean aqueous extract powder, preparation of compound feed powder, preparation of a mixture of a compound feed and the broad bean aqueous extract, and preparation of a broad bean aqueousextract feed. The broad bean extract feed extracted by the method is used for fishes, can significantly improve the muscle hardness, chewiness and cohesiveness of the fishes and promote the growth performance of the fishes, and is an innovative application in fish culture.

The invention relates to a remote sensing-based system and method for simulating watershed hydrology. The method comprises the steps of: processing data by utilizing a data normative processing unit;constructing a model to ensure that model input data can be automatically extracted; integrating the data normative processing unit, the constructed model and an automatic model input data extractingunit; and carrying out demonstration application on a watershed hydrology simulating integration unit. The system comprises: the data normative processing unit, the automatic model input data extracting unit, the watershed hydrology simulating integration unit and a demonstration application unit. Through the method and the system, diversified information is effectively integrated and optimized and an efficient data processing method is proposed, so that the feature utilization rate of various data is increased, the data processing speed is increased, the integration level of simulation technique and the model is enhanced and the development of model simulation study is promoted.

The invention discloses a low-exposure vein image enhancement method based on cross-scale feature fusion. The method comprises the following steps: making a low-exposure vein image data set, and constructing a cross-scale feature fusion module; based on a channel attention mechanism, constructing a vein cross-scale fusion residual block by adopting a residual structure; connecting and stacking a plurality of residual blocks end to end to form a vein image cross-scale fusion model, constructing a test set, and inputting the low-exposure vein image in the test set into the vein image cross-scale fusion model to obtain an enhanced vein image. According to the multi-scale branch cross-scale feature information fusion method provided by the invention, spatial structure information between different scales of the vein image is fully utilized, the representation learning ability of the network model for detail information such as vein veins is enhanced, and the enhancement effect of the low-exposure vein image is improved.

The invention relates to a main bearing member used for optical minimal satellite and a common structure of an optical camera. The main bearing member comprises an optical camera, an on-satellite standalone device, a carbon fiber common structure, a diaphragm and a main reflector; the diaphragm can eliminate the stray light of a main reflector of the optical camera; and the on-satellite standalone device is installed on the fiber common structure. The invention provides a common structure for the main bearing structure of the optical small satellite and provides a common structure of the optical camera, both the main bearing structure and the common structure have functions of the camera light shielding housing and the satellite main bearing, improves the function utilization rate of the on-satellite structure member, and thus improves the space utilization rate in order to reduce the enveloping size and reduce the weight of the integral satellite.

A construction method for jacking up a building with a steel slideway, comprising the following steps: reinforcing the original foundation (1) of the building, calculating the displacement to be lifted for each frame column (2), and A steel corbel (4) is arranged around the frame column (2), and a jack (5) is respectively installed under each steel corbel (4). Appropriate range Use a wire saw to saw off the frame column (2) until it is completely cut, and insert a thin steel plate at the fracture of the frame column (2) sawed off by the wire saw, and set the angle steel (6) to form a steel slideway (8). The upper part of the truncated frame column (2) is lifted to the specified height, and fine stone concrete is poured between the steel slideway (8) below the upper part of the frame column (2) and the lower part of the frame column (2) to form a permanent load-bearing column, thus completing the Lifting of buildings. The construction speed is fast, the cost is low, and the impact on the surrounding environment is small.

The invention discloses an image feature extraction method capable of being relied on for a long time, and the method comprises the following steps: analyzing an overlapping region of images continuously shot by a camera through estimating the motion state of the camera, and carrying out the feature extraction of the overlapping region; constructing a pyramid layer, extracting angular points on the pyramid layer layer-by-layer, selecting the number of the angular points according to the image resolution, wherein the larger the image resolution is, the more the number of the angular points needing to be extracted is; adjusting the mean value and variance of the Gaussian function through the motion state to achieve the effect of controlling the threshold distribution of the extracted features, and thus controlling the distribution of the extracted features on the image. According to the invention, the universality property is good; the extracted features come from the overlapping region,and the feature matching is the feature of the needed part of the image, so that the extracted features can be used for feature matching, and the feature utilization rate is high; the method can be accurately used for image feature extraction of intelligent video search in target recognition of pedestrians and the like while complete autonomous extraction is achieved, and the method is easier toimplement in engineering.

The invention relates to a hot water pipeline temperature adjusting device for family heating. The device comprises a temperature control module and an air power module. The temperature control module is used for detecting the temperature of a hot water pipeline and controlling running and stopping of a hydrodynamic module. The air power module is connected with the hot air pipeline and used for driving water in the hot water pipeline to circulate and flow more quickly. The air power module is an air multiplier on the basis of the air multiplication principle. The temperature control module comprises an intelligent controller and a temperature sensor connected with the intelligent controller, and the temperature sensor is attached to the outer wall of the hot water pipeline. When the temperature on the hot water pipeline is lower than a set temperature, the intelligent controller controls the air multiplier to be automatically started and drive the water in the pipeline to flow, hot water in the pipeline starts to circulate, and heat is evenly conveyed to all places of the hot water pipeline. The hot water pipeline temperature adjusting device for family heating of the structure has the advantages of effectively adjusting heating uniformity of the hot water pipeline and being quick in temperature adjusting, convenient to use and affordable.

The invention discloses a new coronal pneumonia X-ray image classification method and system based on a lightweight model. According to the method, an image classification model comprising a multi-path multi-scale feature fusion module and a depth separable convolution dense block is constructed, receptive fields obtained by multiple paths in the multi-path multi-scale feature fusion module are different, the feature utilization rate can be effectively improved, the representation capability of the model is enhanced, and over-fitting is effectively prevented; the depth separable convolution dense block performs space and channel association on the features in the feature map, and performs channel splicing on the associated feature map and the input feature map, thereby achieving feature reuse, improving the model characterization capability, and effectively solving the problem of model underfitting caused by gradient disappearance. The image classification model provided by the invention is a lightweight model, can be applied to mobile equipment or embedded equipment with limited storage space and computing power, and is high in X-Ray image classification speed and high in accuracy.

The invention relates to an image data processing method and a system. The method comprises the following steps: receiving high-quality satellite data through a sensor to obtain a global image after removing interference; using the filter to remove the background to form the target image database; feature extraction and classification of the image in the target image database, verification and adjustment of the classification; outputting and displaying the verification results and the adjusted classification. The system comprises a receiving device, a global image database forming device, a target image database forming device, a feature extraction and classification device, a verification device, and an output and display device. The method and the system can remove the interference and the background of the diversified information, so that the data processing is efficient and accurate, and the precision and the accuracy are improved through the classification verification, thereby the feature utilization of the multi-data is improved, and the processing efficiency and the quality of the data are improved.

The invention discloses a face recognition model construction method, recognition method, device and storage medium, including preprocessing a face picture, and extracting four local areas of the preprocessed face picture, namely left eye, right eye, nose and mouth Carry out face segmentation, amplify the data based on the segmentation results, and consider the occlusion of 1 / 2 and 1 / 4 cases to perform data amplification respectively, and finally use the amplified data to train the deep learning neural network to construct Train the feature dataset to achieve face recognition based on feature matching. This method has a good recognition effect on occluded faces, does not require a large number of occluded samples, and occupies less memory.

The invention discloses a rotating machinery fault diagnosis method and system based on a multi-scale network structure. The method includes: collecting the operation data of the rotating machinery, inputting the operation data into the fault diagnosis convolutional neural network, and the fault diagnosis convolutional neural network outputs the diagnosis result; in the fault diagnosis convolutional neural network: The features are fused in the channel dimension, and the fused features are used as the input of the subsequent network; the weight of all or part of the output features of the convolutional layer is obtained through the spatial attention mechanism, and the weight is multiplied by the output feature of the convolutional layer to obtain The scaled features, the scaled features are used as the input of the pooling layer connected after the convolutional layer. The convolutional neural network for fault diagnosis can have high migration ability on different distribution training data and test data, and obtain quite high accuracy and migration on data sets under different rotation speeds and different loads.

The invention relates to a remote sensing-based system and method for simulating watershed hydrology. The method comprises the steps of: processing data by utilizing a data normative processing unit;constructing a model to ensure that model input data can be automatically extracted; integrating the data normative processing unit, the constructed model and an automatic model input data extractingunit; and carrying out demonstration application on a watershed hydrology simulating integration unit. The system comprises: the data normative processing unit, the automatic model input data extracting unit, the watershed hydrology simulating integration unit and a demonstration application unit. Through the method and the system, diversified information is effectively integrated and optimized and an efficient data processing method is proposed, so that the feature utilization rate of various data is increased, the data processing speed is increased, the integration level of simulation technique and the model is enhanced and the development of model simulation study is promoted.

The invention relates to a method for accessing WiFi through scanning a two-dimensional code. The method comprises the steps of: initiating a scanning program to scan a WiFi two-dimensional code which contains access information for accessing the WiFi and guiding information for guiding the initiation of a target scanning program; analyzing the WiFi two-dimensional code to obtain access information if the current scanning program is a target program, and entering a page authorizing connection to the WiFi; analyzing the WiFi two-dimensional code to obtain the guiding the initiation if the current scanning program is not the target program, initiating the target scanning program to analyze the WiFi two-dimensional code to obtain access information, and entering the page authorizing connection to the WiFi according to the access information; and receiving a connection command of a user, and accessing the WiFi according to the access information of the WiFi. In addition, the invention further provides a device for accessing WiFi through scanning the two-dimensional code. The method and the device for accessing the WiFi through scanning the two-dimensional code can add a sharing function to the scanning program, and enhance the function utilization rate of the scanning program.

The invention provides an operation key structure and a mobile terminal device having the same. The operation key structure includes a housing and a key cap that is threaded on the housing. The key cap is rotatable around its axis and movably penetrated on the housing along its axis. The operation key structure also includes: a first switch assembly , the keycap is set opposite to the first switch assembly, and when the keycap moves along its axis, the first switch assembly is turned on or off; the second switch assembly, the keycap is connected to the second switch assembly, and the keycap rotates around it When the axis rotates, the second switch assembly is turned on or off. The operation key structure in the present invention solves the problem that the function of the operation key structure in the prior art is relatively single.

The invention discloses an SKW real control network field bus, which comprises a master machine and one to fifteen slave machines, wherein serial ports are formed on the master machine; and the master machine is connected with the slave machines through the serial ports. The SKW real control network field bus is simple in structure, lower in cost, small in occupation space of equipment, high in functional utilization ratio, and quite applicable to a field control unit of a small hydropower station or other industrial automatic systems.

Embodiments of the present invention provide a terminal application processing method, device, and terminal, wherein the method includes: receiving a call request from a first application in a terminal to a target function module in a second application; The invocation of the target function module will not generate a call conflict, then in response to the call request, call the target function module in the second application for the first application; wherein, the invocation of the target function module will not generate a call The conflict includes: detecting that the second application does not call the target function module, and / or determining that the call to the target function module does not generate a conflict according to a preset conflict list. By adopting the present invention, function sharing can be quickly realized, and the function utilization rate of the function modules can be improved, and at the same time, the development cost of the terminal application can be greatly reduced, the development efficiency can be improved, and the calling conflict of the function modules can be accurately avoided, thereby improving the The accuracy of the call.