371results about How to "Fully integrated" patented technology

The invention discloses a medical visual question-answering method and system based on multi-task modeling, and the method comprises the steps: carrying out the feature extraction of a to-be-analyzed medical image through a visual encoder, obtaining an image feature, inputting a preliminary question instruction and historical question-answering data into a dialogue large model to generate a question instruction, and obtaining a text feature through a text encoder; carrying out feature fusion on the obtained text features and image features, and carrying out self-attention image importance weighting processing and image-text alignment to obtain image-text fusion representation; and inputting a multi-target output projection layer, carrying out multi-task prediction, outputting question answers, image categories and important region masks, inputting the question answers and the image categories into a dialogue large model, and finally generating diagnosis opinions with detailed description in combination with dialogue context and multi-round interaction data and in combination with the important region masks. According to the invention, the interaction experience and the performance of the question and answer system are improved, and more accurate medical diagnosis in multi-modal complex information is facilitated.

The invention discloses a micro-expression recognition method and system based on a state space model and double-flow fusion, and relates to the technical field of expression recognition, and the method comprises the steps: extracting a starting frame image and a top frame image of a micro-expression continuous video frame sequence; inputting the initial frame image and the top frame image into a micro-expression recognition model based on a double-flow fusion network, extracting an optical flow image, and performing face key point detection and clipping operation on the top frame image and the optical flow image respectively; carrying out image block embedding operation to obtain texture features of the top frame image after cutting and motion features of the optical flow image; the two features learn the global dependency relationship and the local correlation between the image blocks from the positive direction and the negative direction through a parallel multi-layer bidirectional Mama fusion module, and deep image features and motion features are extracted after feature fusion; and the deep features of the two modes are interacted through a double-flow cross attention classification module, comprehensive features are learned, and a plurality of classifiers are combined to identify and output the micro-expression type more accurately.

The invention provides a method and a system for reconstructing a hyperspectrum from an RGB image. The method comprises the following steps: acquiring a three-channel RGB image; performing embedding and channel expansion on the RGB image by using a convolutional neural network to obtain an initial feature map; inputting the initial feature map into an encoder, respectively extracting channel dimension features and space dimension features by utilizing VisionTransform and VisionMama in the initial feature map, fusing the channel dimension features and the space dimension features, and respectively inputting the channel dimension features and the space dimension features into jump connection and separable convolution to obtain feature maps of different levels; inputting the feature map into a decoder to obtain a final feature map; mapping the final feature map by using a convolutional neural network, and adding the final feature map with an initial feature map before the final feature map is input into an encoder to obtain a hyperspectral image in a compressed form; and splitting the hyperspectral image in the compressed form according to the number of channels, and then performing color mapping to finally obtain a reconstructed hyperspectral image. According to the method, the precision and generalization performance of reconstructing the hyperspectrum from the RGB image can be improved.

The invention discloses a three-dimensional geological implicit modeling method based on global constraint features, a medium and equipment, and relates to the technical field of geological modelling, the three-dimensional geological implicit modeling method based on global constraint features mainly comprises the following steps: carrying out vertical dimension sampling and horizontal dimension sampling on original geological data; a profile data set and a drilling data set are obtained, global feature extraction is carried out on the profile data set and the drilling data set to obtain global vertical features and global horizontal features, global feature fusion and coordinate matching are carried out on the global vertical features and the global horizontal features to obtain a machine learning data set, and a Stacking machine learning model is trained to obtain a trained Stacking machine learning model; and predicting a research area by using the trained Stacking machine learning model to obtain a three-dimensional geologic model. According to the three-dimensional geological implicit modeling method based on the global constraint features, the medium and the equipment provided by the invention, the global expression of the underground space in the three-dimensional geological modeling process can be enhanced.

The invention belongs to the technical field of additive manufacturing, and particularly relates to an additive manufacturing method of an ultrahigh-strength steel-stainless steel interlayer heterogeneous structural part. Comprising the following steps that (1) a model is obtained, and an additive model is designed according to performance requirements needed by a structural part; (2) additive preparation is conducted, specifically, the type and specification of a base plate are selected, mechanical grinding and welding wire filling are conducted on the base plate, an additive path is planned according to the model, and a robot walking program is written; and (3) material adding is started, the optimal technological parameters are matched according to the planned material adding path, and material adding of the structural part is started and completed. The ultrahigh-strength steel-stainless steel heterostructure part obtained through the method has the excellent strength-toughness comprehensive mechanical property, the material adding efficiency is high, the interface of the ultrahigh-strength steel-stainless steel heterostructure part is free of the defects of holes, microcracks, unbonding and the like, and the surface quality is excellent.

The invention belongs to the technical field of unit fault diagnosis, and discloses a fault diagnosis method and system for a fault large-scale unit based on sharing and specific representation learning. The method not only can be applied to multi-measuring-point fault diagnosis of a large-scale unit, but also has wide applicability. In other multi-sensor systems or complex equipment, the fault detection effect can also be improved through multi-view representation learning. The invention provides a weighted sharing-specific multi-view depth representation learning fault diagnosis method for fault diagnosis of multi-measuring-point data of a large-scale unit. The method consists of two parts, namely sharing and specifically representing a learning network and weighted multiple similarity loss and orthogonal loss. A multi-measuring-point data fault diagnosis problem of a large-scale unit is solved by sharing a specific representation learning network, and weighted multiple similarity loss and orthogonal loss are used to further improve the feature extraction capability of the network and reduce redundancy.

The invention provides a bamboo strip defect detection method based on improved YOLOv9, and the method comprises the steps: employing a bamboo strip defect detection model based on YOLOv9, introducing a BottleneckAey module in a GELAN structure, and forming a RepNCSPELAN4Aey feature extraction module, so as to enhance the feature extraction capability of a Backbone network; and performing defect detection on the bamboo cane image after the model is trained.

The invention relates to the technical field of remote sensing image intelligent interpretation, in particular to a remote sensing image building extraction method and system based on a depth feature fusion network, and the method comprises the steps: firstly extracting the depth information of a remote sensing image through a depth information extraction module; fusing the depth information and the corresponding remote sensing image by using a depth feature fusion module to obtain a fused feature map; and finally, inputting the fused feature map into a U-Net network to carry out building extraction of the remote sensing image. According to the invention, the building extraction accuracy of the model is improved through depth information extraction and depth feature fusion; and extra multi-source registered remote sensing image data is not needed, so that the limitation of a traditional multi-source data fusion strategy is solved, and the extra multi-source registered remote sensing image data usually needs a large amount of time and energy to obtain.

The invention relates to a multi-modal survival prediction method fusing a digital pathological image and omics data. The method comprises the following steps: acquiring a full-view digital pathological image and omics data of an esophageal squamous cell carcinoma patient; carrying out preprocessing and feature extraction on the full-view digital pathological image and the omics data to respectively obtain a pathological embedding representation and an omics embedding representation; fusing the pathological embedding representation and the omics embedding representation by using a multi-head attention mechanism of Transform to obtain a preliminary fusion feature; performing deep interaction on the pathological embedding representation and the preliminary fusion feature to obtain a final fusion feature; and predicting a survival result based on the final fusion feature. According to the method, bimodal sharing of pathological images and omics data and full fusion of features are realized, and prognosis survival prediction of esophageal squamous cell carcinoma cancer patients can be better realized.

The invention discloses a false news detection method and system based on a partition filtering network, and the method comprises the steps: carrying out the fusion of news text features and comment text features, and obtaining event features; inputting the news text and the comment text into a dual-emotion acquisition module to obtain dual-emotion features; inputting the word vector of the comment text into a partition filtering network to obtain an initial vertical feature and an initial rumor feature; fusing the event feature with the initial vertical feature to obtain a corrected vertical feature; classifying the corrected vertical features to obtain a first classification result; fusing the initial rumor feature, the dual emotion feature and the corrected standing feature to obtain a corrected rumor feature; classifying the corrected rumor features to obtain a second classification result; and fusing the first classification result and the second classification result to obtain a final classification result.

The invention discloses a power distribution network fault classification method, terminal equipment and a storage medium, and the method comprises the steps: collecting fault recording data from a power distribution network, carrying out the preprocessing of the fault recording data, and obtaining the preprocessed data; performing category labeling on the preprocessed data to obtain a data labeling result; inputting the training data set into a data classification model, and training the data classification model to obtain a classification model; wherein the training data set comprises the preprocessed data and the corresponding data labeling result. The method can automatically classify the fault data into the category of the labeled data, reduces the time of manual observation and analysis, and facilitates the guarantee of the stable operation of a power grid. In addition, various fault types can be identified, and the generalization performance is excellent.

The invention provides a high-strength blooming-free PVC (polyvinyl chloride) cable material as well as a preparation method and application of the high-strength blooming-free PVC cable material. The environment-friendly bio-based plasticizer, the organic tin and rare earth composite heat stabilizer, the efficient compound antioxidant, the surface treatment modified nanofiller, the multifunctional processing aid, the phosphorus-nitrogen compound flame retardant and the cross-linking accelerant are matched in a specific proportion. The preparation method comprises the steps of raw material drying, mixing plasticizing, granulation forming and the like. The prepared cable material has the excellent properties of high strength, no blooming, low temperature resistance, high salt resistance and the like, is suitable for preparing various high-performance cables, is particularly applied to cables in a high-temperature-resistant and high-salt-resistant environment, and shows a wide application prospect.

The invention relates to the technical field of valve casting, and discloses a casting production device and method for valve casting, the casting production device comprises a box body, the top of the box body is fixedly connected with a first supporting seat, the top of the first supporting seat is fixedly connected with a motor, the output end of the motor is fixedly connected with a first electric telescopic rod, and the inner wall of the box body is fixedly connected with a hydraulic cylinder; a protective sleeve is fixedly connected to the inner wall of the box body, a pushing mechanism is arranged in the box body, a valve subjected to mold fixing can be subjected to extrusion force of an ejection block to be demolded and separated from a mold, the demolding efficiency of the device is improved, the ejection block makes contact with the mold by directly utilizing pushing force of a hydraulic rod, the service time of the device is shortened, and the device is convenient to use. And the parts can fall into the external collecting box, so that the time cost of manual collection is reduced, potential safety hazards caused by too high temperature of the parts during manual collection are avoided, and the use efficiency of the device is improved.

The invention relates to the technical field of prefabricated part production, and discloses a concrete prefabricated part pouring production line for a tunnel. The stirring assembly and the stirring barrel are arranged in a penetrating manner; the driving assembly is arranged in the stirring barrel, the driving assembly is mounted on the inner wall of the stirring barrel, and the driving assembly is connected with the stirring assembly; and the tiny stirring part is arranged in the stirring barrel. Through arrangement of the stirring assembly, especially cooperative cooperation of a first wedge-shaped column, a second wedge-shaped column, a limiting sliding block and a first spring, when stirring is started, an auxiliary circular plate rotates to promote the first wedge-shaped column to rotate along with the first wedge-shaped column and extrude the inclined surface of the second wedge-shaped column, and the first wedge-shaped column slides in a limiting sliding groove in cooperation with the limiting sliding block and elastic buffering of the first spring; the auxiliary circular plate is successfully driven to transversely reciprocate, so that the stirring rod is driven to realize reciprocating transverse stirring; the dynamic stirring mode greatly improves the stirring flexibility.

The invention relates to the technical field of computer vision, in particular to a method and a device for constructing a binocular stereo matching recursive network based on deep learning, in a feature extraction stage, a self-adaptive spatial pyramid pooling layer based on a kernel normalization convolution layer is provided, multi-scale feature extraction is performed on an image, and in a matching cost calculation stage, a self-adaptive spatial pyramid pooling layer based on a kernel normalization convolution layer is provided. A new attention approach is utilized in combination with a regularized cost volume to produce a more accurate similarity measure. And in the main network framework, a GRU (Gated Recirculation Unit) based on kernel convolution is adopted as an iterative update operator, features are retrieved from the stereo matching cost body, and parallax is updated through regression iteration. The GRU can consider the wider context around each pixel, the fusion of local and global information is promoted, the more comprehensive understanding of the feature space is provided, and the reasoning ability of the network for the scene structure of a larger area is enhanced.

The utility model relates to the field of extraction equipment, in particular to a composite plant anti-allergy preparation device capable of fully stirring, which comprises an extraction tank, the top of the extraction tank is provided with a feeding port and a steam output port, the position of the extraction tank close to the bottom is provided with a steam outlet, and the steam outlet is communicated with a condensing device; the condensing device comprises a shell and tube condenser and a coil condenser; an output port of the coil type condenser is connected with an oil-water separator through a pipeline, an oil collector is arranged at the top of the oil-water separator, a liquid discharge port is formed in the bottom of the oil-water separator and connected with a return tank through a pipeline, and a discharge port of the return tank is communicated to input ports of the tubular condenser and the coil type condenser through pipelines; a stirring paddle is arranged in the extraction tank, the stirring paddle is rotatably mounted in the extraction tank, and the stirring paddle ascends and descends in the vertical direction while rotating. The purpose of the utility model is to effectively guarantee the quality of subsequent extraction of an extracting solution.

The invention discloses equipment for preparing lipoic acid particles, and belongs to the field of lipoic acid preparation.The equipment comprises a tank body, an upper cover is fixedly connected to the upper surface of the tank body, a feeding port is connected to one side of the upper surface of the upper cover, a discharging port is connected to the lower surface of the tank body, and an evaporation bin is fixedly connected to the middle of the inner surface of the tank body; second electromagnetic valves are symmetrically and fixedly connected to the lower surface of the evaporation bin, a stirring assembly used for stirring raw materials is connected to the upper surface of the upper cover, and a drying assembly used for drying the raw materials is connected to the position, close to the lower surface, of the outer surface of the tank body. The swing blades can continuously swing left and right in the dissolving frame, so that a solution in the dissolving frame is stirred in different directions, the solution forms turbulent flow in the dissolving frame, lipoic acid powder moves irregularly and is fully fused with the solution, the dissolving speed of the powder is increased, and compared with an existing stirring mode, the stirring efficiency is improved. Stirring and dissolving time is short, and working efficiency is high.

The invention relates to the technical field of food processing and functional drinks, and discloses a peptide-component-rich health-preserving liqueur which comprises the following components in percentage by weight: 50-65% of Maotai-flavor liquor, 10-20% of yellow wine, 3-7% of water, 10-20% of peptide components, 1-5% of plant extract and 5-10% of other components, the invention also discloses a preparation method of the health-preserving liqueur rich in peptide components, which comprises the following steps: S1, peptide treatment: carrying out directional enzymolysis on the peptide components, controlling the enzymolysis temperature at 40-50 DEG C and the enzymolysis time at 2-4 hours; and S2, microcapsule coating: coating the treated peptide components in cyclodextrin, gelatin or Arabic gum. By adopting a compound formula of multiple peptide components and combining technologies such as microcapsule coating, directional enzymolysis, low-temperature co-fermentation and nano-emulsification, the stability and bioavailability of the peptide components are remarkably improved, meanwhile, the flavor and taste of a wine body are optimized through a two-stage fermentation process, the wine body is more mellow and smoother, and the taste of the wine body is improved. The requirements of the market on the high-quality health-preserving liqueur are met.

The invention discloses a high-speed rail perimeter intrusion alarm method and system based on thunder-vision fusion, and solves the problem of low reliability of high-speed rail perimeter intrusion alarm in the prior art, and the method comprises the steps: carrying out the weather detection and rail surface foreign matter detection based on radar data, and generating radar weather information and radar alarm information; weather detection and rail surface foreign matter detection are carried out based on video data, object tracking is carried out on a moving target and a low-confidence target by adopting a similar SORT algorithm in combination with IOU, gradient and LBP, and video weather information and video alarm information are generated; decision-level fusion is carried out on the weather information and the alarm information, and effective alarm information is reserved; and performing priority weighting on the effective alarm information in a period of time, and taking the alarm information with the highest priority as final alarm information. According to the method, the reliability of object tracking is improved, the considered influence factors are more comprehensive, and the reliability of high-speed rail perimeter intrusion alarm is further improved.

The invention belongs to the technical field of NH3 catalytic oxidation catalysts, and particularly relates to a preparation method and application of a surface defect-rich catalyst for NH3 catalytic oxidation, and the preparation method comprises the following steps: (1) mixing heteropolyacid, a molecular sieve and deionized water, grinding, carrying out vacuum drying, and calcining in an inert atmosphere to obtain catalyst carrier powder; (2) weighing a manganese precursor salt and an auxiliary agent precursor salt, dissolving the manganese precursor salt and the auxiliary agent precursor salt in water to prepare an aqueous solution, adding the catalyst carrier into the aqueous solution of the manganese precursor salt and the auxiliary agent precursor salt, carrying out vacuum drying, and calcining in an inert atmosphere to obtain a supported catalyst; and (3) reducing the supported catalyst in a mixed reducing atmosphere containing CO / inert gas to obtain the catalyst with rich surface defects. According to the method disclosed by the invention, rich defect sites can be constructed on the surface of the catalyst, the NH3 catalytic oxidation activity of the catalyst under a wide temperature window is promoted, and meanwhile, the N2 selectivity is improved.

The embodiment of the invention relates to a wireless communication technology, and discloses a low-delay service transmission method, an access point and a storage medium. In the invention, the access point equipment which is also in a multi-AP cooperative transmission mechanism is used as a transmission opportunity holder; a multi-user request sending frame used for inquiring low-delay service information is sent to each transmission opportunity waiting party in a multi-AP cooperation set, and distribution of transmission opportunities is determined through feedback clear sending frames carrying the low-delay service information; and when serving as a transmission opportunity waiting party, generating low-latency service information according to the low-latency service existing in the station equipment or the station equipment associated with the station equipment, and sending a clear sending frame carrying the low-latency service information to the transmission opportunity holder. By reasonably distributing the transmission opportunities, the actual demand of current low-delay service processing is met, the characteristics of multi-AP cooperation and the low-delay service are fully fused, and efficient and stable transmission of the low-delay service in a wireless network is achieved.

The utility model provides a folding drill boom and an engineering truck. The folding drill arm comprises a basic arm, a lifting arm installed at the upper end of the basic arm in a pitching mode, a folding arm hinged to the end, away from the basic arm, of the lifting arm, and a drilling mechanism hinged to the end, away from the lifting arm, of the folding arm. The drilling mechanism, the folding arm and the lifting arm sequentially rotate in the direction close to the basic arm to form a folded state, and otherwise, the folding arm is in an unfolded state. And a swinging mechanism for driving the basic arm to swing left and right is arranged on the basic arm. The swing mechanism capable of driving the basic arm to swing left and right is additionally arranged on the basic arm, so that the drilling operation range is adjusted, and the transverse perpendicular relation between the folding drill arm and the operation ground is adjusted; besides, the lifting arm is connected with the basic arm in a pitching mode, so that the requirement of the folding drill arm for adjusting drilling positioning in the longitudinal direction is met, and the obstacle crossing operation capacity of the folding drill arm is also achieved.

The invention discloses a resistivity-while-drilling logging real-time image super-resolution method, system and equipment based on a generative adversarial network, and a medium, and the method comprises the steps: firstly, reading resistivity-while-drilling storage data of a certain region, obtaining a high-resolution image from the storage data according to pixels, carrying out the sampling of the resistivity-while-drilling storage data at equal intervals, and carrying out the imaging, obtaining a corresponding low-resolution image, wherein the high-resolution image and the corresponding low-resolution image form a paired image data set; secondly, a generative adversarial network model is built, the generative adversarial network model comprises a generator and a discriminator, the generator takes a low-resolution image as input and outputs a high-resolution image through stacked SE-RRDB modules and up-sampling operation, and the discriminator evaluates the generated image through multilayer convolution and jump connection and judges the authenticity of the image; thirdly, training the generative adversarial network model by the image data set, optimizing the network, and storing the optimal generative adversarial network model; and finally, acquiring resistivity logging-while-drilling real-time data, processing the data, and inputting the processed data into the trained generative adversarial network model to obtain a real-time image after super-resolution. The method has the advantages of rapidness, high efficiency, good reconstruction effect and high practicability.

The utility model relates to the field of reactors, and discloses a wastewater denitrification nitrogen removal reactor, which comprises a reaction tank, a filter tank and a clamping plate, the filter tank is arranged in the reaction tank, grooves are arranged on both sides of the top and the bottom of the filter tank, and the groove on the top of the filter tank is connected with the bottom of the clamping plate in a clamping manner. One side of the clamping plate is connected to the wall of the filter tank in a clamped mode, the side, close to the reaction tank and the wall body of the filter tank, of the clamping plate is provided with wall brushes, the wall brushes are integrally connected to the clamping plate, and the wall brushes are movably connected with the wall body of the reaction tank and the wall body of the filter tank. The stirring shaft rotates to drive the first stirring blades in the filter tank and the clamping plate, so that waste water drives internal impurities to rotate in the filter tank, filtration and flowing of the waste water are accelerated, meanwhile, the clamping plate drives the filter tank brush to wash filter holes of the filter tank, and the filter tank brush is driven by the clamping plate to brush the filter holes of the filter tank. The filter holes are prevented from being blocked by impurities to influence sewage flowing.

The invention provides a multi-source ground feature classification method based on a semi-supervised CNN-Transform hybrid model, provides a semi-supervised ground feature classification model based on transform and convolution, fully utilizes the advantages of two backbone networks, generates a pseudo tag with high confidence, assists the networks in training, and improves the classification accuracy of the ground features. According to the new model, good experimental results are obtained in terms of overall accuracy, average accuracy and Kappa coefficients, a multi-layer feature fusion strategy is designed in the module, shallow and deep features are fused after being spliced, full fusion among different modal features is realized by using a gating network, and the fusion accuracy of the model is improved. And modal redundant information with relatively high similarity is eliminated, so that the utilization efficiency of the model on the modal specific information is improved, and an obvious contribution is made to the improvement of the model performance.

The utility model discloses a powder water-soluble fertilizer mixing device which comprises a pretreatment mechanism, a sealed transmission mechanism and a turbulent flow stirring mechanism, the pretreatment mechanism comprises a driving motor, a block removing roller, an interaction gear and a treatment box, and the sealed transmission mechanism comprises a transmission motor, a spiral blade, a conveying pipe, a feeding pipe and a discharging pipe. The turbulent flow stirring mechanism comprises a stirring motor, a stirring frame, a mixing tank, a supporting rod and a turbulent flow plate, and through cooperation of the mechanisms, the device can crush caked powder fertilizer and can transport the powder material in a sealed mode, so that the powder fertilizer is prevented from being affected with damp and caked in the transportation process, dissolution is not affected, and the service life of the fertilizer is prolonged. Meanwhile, floating of powder in the air is reduced, the environment is protected, meanwhile, the human body is protected, finally, when the materials are mixed, fusion between the powder fertilizer and water is faster and more sufficient, and the dissolution rate of the powder fertilizer is increased.

The invention discloses a method for monitoring a blood vessel and a stent in the blood vessel. The method comprises the following steps: acquiring a target blood vessel fusion image of a target blood vessel at a preset angle; wherein the target blood vessel fusion image is an image obtained after a CT angiography image and an X-ray coronary angiography image of a target blood vessel are fused, and the target blood vessel fusion image comprises the intima features of the blood vessel; acquiring stent transmission images of the stent at different positions in the moving process in the target blood vessel at a preset angle; segmenting the support transmission image to obtain a support segmented image; and performing fusion display on the target blood vessel fusion image and the stent segmentation image to obtain a multi-modal fusion image of the target blood vessel and the stent. By means of the method, display of the intima characteristics in the target blood vessel on the image is fully considered, and the condition of the stent in the blood vessel can be observed more clearly from the image. The invention further discloses a monitoring device for the blood vessel and the stent in the blood vessel, electronic equipment and a computer readable storage medium.

The invention discloses raw material cleaning equipment for EPS foam material processing, and relates to the technical field of foam cleaning, and the raw material cleaning equipment comprises a composite mechanism. The raw material cleaning equipment for EPS foam material processing is designed through a composite mechanism. Raw materials enter from the top of the composite shell, water flow enters from the water inlet pipe, then the motor drives the center rotating shaft to rotate, the stirring mechanism stirs the raw materials, the effect of cleaning the raw materials is achieved, the effects of removing impurities and improving the product quality are achieved, and meanwhile subsequent machining equipment is prevented from being damaged by the impurities; in the rotating process of the central rotating shaft, the friction mechanism and the rotating mechanism are driven to rotate, and the friction mechanism rubs the inner wall surface of the equipment, so that the effect of cleaning materials on the inner wall is achieved, material adhesion is reduced, the adhered materials participate in cleaning operation, it is guaranteed that the materials are uniformly cleaned, and the influence on subsequent cleaning equipment is avoided; and then the materials are discharged from the output mechanism to facilitate subsequent operation.

The invention discloses an infrared and visible light remote sensing image deep learning fusion method based on a visual state space model. The method comprises the following steps: acquiring an infrared remote sensing image and a visible light remote sensing image to be fused; the infrared remote sensing image and the visible light remote sensing image are spliced and then input into a pre-constructed and trained image fusion network model, a fusion image output by the image fusion network model is obtained, and the image fusion network model comprises a first convolution module, a double-attention residual module, a multidirectional scanning module and a second convolution module which are connected in sequence. The first convolution module is used for extracting image shallow layer features, the double-attention residual module is used for extracting and fusing important features of an infrared remote sensing image and a visible light remote sensing image, the multi-direction scanning module is used for obtaining global features of the images in multiple directions, and the second convolution module is used for recovering image dimensions and obtaining a fused image; and outputting the fused image. According to the invention, infrared remote sensing image information and visible light remote sensing image information can be fully fused.

The utility model relates to the field of waterproof construction, and discloses a building roof waterproof paint smearing device which comprises a cart, the top of the cart is fixedly connected with a mixing barrel, the top of the mixing barrel is fixedly connected with a first motor, the output end of the first motor is fixedly connected with a first sleeve rod, and the interior of the first sleeve rod is slidably connected with a rotating rod. A second sleeve rod is slidably connected to the bottom of the rotating rod, two stirring rods are fixedly connected to the outer side of the rotating rod, a swash plate is fixedly connected to the upper side of the rotating rod, a fixing rod is fixedly connected to the inner top wall of the mixing barrel, two guide wheels are fixedly connected to the right side of the fixing rod, and a pressurizing assembly is installed at the top of the cart. According to the stirring device, the first motor drives the first sleeve rod, the first sleeve rod drives the rotating rod, the guide wheel limits the swash plate, and the stirring rod is driven to work, so that normal stirring and up-down moving stirring can be achieved, waterproof paint can be fused more sufficiently, time and labor are saved, and the working efficiency is effectively improved.