53results about How to "Improve fusion effect" patented technology

The application discloses a kind of millimeter wave radar and the fusion method and system of visual identification target, the method comprises: obtaining the data information of multiple first target points and multiple second target points;The first target point represents the target object detected by millimeter wave radar sensor, and the second target point represents the target object detected by camera;With the first target point as the center, the absolute threshold set is expanded longitudinally and transversely, to form the first confidence interval;With the second target point as the center, the relative threshold is expanded longitudinally and transversely to form the second confidence interval;The size of the relative threshold is linearly related to the distance and speed of the second target point;KM algorithm is executed to fuse, as many as possible millimeter wave radar detected first target points and camera detected second target points are one-to-one associated as the same target object.The method can improve the target fusion effect and reduce the generation of virtual target.

The present application relates to computer vision and underwater detection technology, provide a kind of underwater small biological target detection method and system based on deep learning, the WLDA-YOLOv8 deep learning model constructed includes main network, neck network and detection head, main and neck embed light multi-scale module, detection head introduces dynamic detection head and adaptive sample selection strategy;Training stage uses improved WIoUv3s loss to carry out boundary box regression, the model is trained using underwater biological data set, finally realizes the accurate detection of small target in underwater image, enhances the capture and fusion ability to small target subtle features.

The utility model discloses an auxiliary heating structure of an injection mold, which comprises a heating rod arranged in a fixed mold core, the heating rod penetrates through a cavity of the fixed mold core, and a space between a bulge structure in the cavity and the inner side wall closest to the bulge structure in the cavity is set as a heating section needing to be heated. The projection of the heating section in the depth direction of the cavity of the fixed mold core is partially or completely overlapped with the projection of the heating rod in the depth direction of the cavity of the fixed mold core, the cavity penetrates through the top wall of the fixed mold core to form a discharging port for taking out a product, the heating rod is located on the side, away from the discharging port, of the fixed mold core, and the heating rod is detachably connected with the fixed mold core. According to the utility model, the overall temperature is high when the injection molding material flows to the fusion position, so that the fusion position is not easy to form a fusion line.

The invention discloses a PHSM-YOLO-based multi-target wearing identification method. The PHSM-YOLO-based multi-target wearing identification method comprises the following steps: constructing a PHSM-YOLO target detection model used for identifying whether a constructor correctly wears a safety appliance; collecting a target image containing constructors in the construction environment; and inputting the target image into a PHSM-YOLO target detection model, and outputting a wearing identification result of the safety wearing tool of each construction worker in the target image by the PHSM-YOLO target detection model. According to the multi-target wearing identification method, the wearing of the safety wearing tool can be identified on the target image containing the construction personnel through the constructed PHSM-YOLO target detection model, so that the personal safety of the construction personnel is guaranteed.

The application discloses a smart community police management system and method based on Internet of Things security, and belongs to the technical field of smart communities, and solves the problem that existing system data processing cannot utilize edge computing nodes for low-delay real-time analysis, and cannot meet timely early warning requirements, the method comprising preprocessing of multi-source heterogeneous community monitoring information, analysis and processing of the community monitoring information based on edge computing, and generation of a risk propagation path in combination with front-end analysis results and heterogeneity characteristics; in the application, analysis and processing of the community monitoring information based on edge computing avoids the delay of centralized processing, can quickly trigger early warning in the early stage of a risk event, thereby gaining golden disposal time for risk event disposal, and a risk analysis model realizes a leap from single-node anomaly detection to composite risk quantification through multi-dimensional feature fusion and dynamic space-time reasoning, and realizes multi-dimensional feature fusion and dynamic reasoning capability of the risk event.

This application relates to the field of children's playground technology and discloses a children's activity area based on a suspended walkway. The activity area includes a ground surface, a suspended walkway on the ground surface, and a movable house on the ground surface, located on one side of the suspended walkway. The movable house includes a single-story building with a second-story building on top. This children's activity area based on a suspended walkway, through the movable house, the first climbing frame inside the movable house, and the slide on one side of the suspended walkway, fully utilizes vertical space. The three-dimensional design of the activity area based on the suspended walkway allows for more diverse playground facilities within a limited ground area. The activity area features various playground facilities, providing children with more challenging and fun play experiences. The suspended walkway within the park enhances the interaction and integration with the natural environment.

The invention relates to the technical field of deep mineral detection, and discloses a deep mineral rapid detection method based on multiband spectrum fusion, which comprises the following steps: step 1, building a detection system; 2, carrying the detection system on a detection platform, and synchronously collecting three types of spectral data and course angle, pitch angle, roll angle and speed parameters; step 3, data preprocessing; and step 4, a cross-modal fusion module realizes alignment of three types of data through cross-modal feature matching, and then weighted fusion is carried out. According to the deep mineral rapid detection method based on multi-band spectrum fusion, an attitude sensing module collects a course angle, a pitch angle, a roll angle and speed parameters in real time, a data preprocessing module cooperates with a cross-modal fusion module, and an attitude correction matrix is constructed based on Euler angle transformation to carry out geometric correction on multi-band spectrum data; the detection platform advances according to a preset height and speed working condition, a high-precision inertial navigation system is not needed, and the cost and the detection precision are both considered.

The invention discloses a multi-modal recommendation method for cross-modal semantic alignment and graph relation enhancement. The multi-modal recommendation method comprises the following steps: S1, initializing a model and setting parameters; s2, constructing an article heterogeneous relation graph fusing semantic information and behavior information; s3, executing cross-modal local-global multi-granularity semantic alignment; and S4, performing model optimization and preference prediction based on the aligned multi-modal representation. Through the integrated design of heterogeneous relation fusion and cross-level alignment, the problem that in an existing method, article relation mining is one-sided and cross-modal feature alignment is insufficient is effectively solved, and recommendation accuracy and robustness are remarkably improved.

The utility model relates to a medicine and food homology effect improving and stabilizing device which comprises a conveying belt assembly, an installation frame is connected to the conveying belt assembly, an air cylinder is installed on the upper portion in the installation frame, a first upper installation plate is connected to the telescopic end of the air cylinder, and second upper installation plates are slidably connected to the left portion and the right portion of the first upper installation plate. The front portion of the second upper mounting plate is connected with a heating plate, the front portion of the first mounting plate is also connected with a heating plate, the rear portion of the mounting frame is connected with three fans, and three mold bases are placed on the conveying belt assembly. The heating plate is arranged and heats the first upper mounting plate and the second upper mounting plate which are adjacent, the shielding curtain can enable the interior of the mounting frame to be in a relatively sealed state, so that heat loss is reduced, meanwhile, the fan can promote warm air in the mounting frame to flow, the flowability of medicine powder particles can be enhanced, and the service life of the medicine powder particles is prolonged. Powder adhesion is reduced, powder loss is reduced, and the fusion effect is improved.

The invention discloses an asynchronous federated continuous learning method for a large model, which belongs to the field of federated learning and continuous learning, is applied to a client, and comprises the following steps: receiving an initial model parameter of a global model issued by a server and a global feature extraction module, and performing feature representation on local service task data based on the module, constructing a corresponding feature prototype and a feature covariance statistic; uploading the feature prototype and the feature covariance statistics to a server in an asynchronous mode to perform asynchronous federal fusion, and updating global feature statistics; and receiving the global feature statistics after the server fuses and updates, wherein the global feature statistics are used for local decision modeling of subsequent tasks. Through the asynchronous federal aggregation strategy, incremental feature knowledge from different financial institutions and different task stages can be continuously integrated, the problem of historical knowledge forgetting is effectively relieved, and the stability and consistency of global feature representation in the long term evolution process are kept.

This invention discloses a collaborative generative recommendation method based on multimodal semantic enhancement tagging. The method includes: constructing an entity database; extracting text semantic embeddings, image semantic embeddings, and collaborative embeddings from entities in the entity database; performing semantic enhancement collaborative alignment fusion expression processing on a semantic enhancement collaborative alignment fusion expression model to obtain quantized text embeddings and quantized image embeddings, which are then stored in the entity database; extracting text semantic embeddings and image semantic embeddings from entities of interest and inputting them into the semantic enhancement collaborative alignment fusion expression model; and searching the entity database for entities or / and combinations of entities associated with the entity of interest to generate recommendation results. This invention achieves collaborative enhancement of semantic understanding and structure awareness, enabling a unified representation of semantic richness, structure awareness, and generative guidance capabilities, providing key technical support for intelligent recommendation applications.

This disclosure defines a method and apparatus for determining intent. The method includes: acquiring a voice command sent by a target object and determining the text features corresponding to the voice command; monitoring data frames on a target bus of a vehicle and determining first state data corresponding to the vehicle based on the data frames; collecting facial key points of the target object and determining second state data corresponding to the target object based on the facial key points; calculating a first confidence level of the text features for each candidate intent, a second confidence level of the first state data for each candidate intent, and a third confidence level of the second state data for each candidate intent; and determining the target intent corresponding to the voice command from among multiple candidate intents based on the first confidence level, second confidence level, and third confidence level corresponding to each candidate intent. According to this disclosure, the problem of in-vehicle voice assistants' insufficient ability to fuse multimodal information, leading to the inability to accurately determine the user's deep intent, is solved in related technologies.

The utility model discloses a micro-thread implant, which relates to the technical field of implants and comprises a cylinder, a rubber sleeve is arranged at the lower end of the cylinder, straight threads are arranged on the outer side of the lower end of the cylinder, and an inner hexagonal groove is arranged on the top surface of the cylinder; according to the utility model, through the matching of the rubber sleeve and the cylinder, the cylinder can conveniently extrude the rubber sleeve, so that a hydroxyapatite material in the rubber sleeve is extruded out, and the fusion efficiency of the cylinder and new tissues is improved; through the cooperation of the rubber sleeve and the hydroxyapatite, gaps around the cylinder can be conveniently filled, and the fusion effect of the cylinder and new tissues is improved; and through the matching of the straight thread and the fine thread groove, the tooth body can be better mounted, and the tooth body mounting effect is improved.

The present application belongs to the technical field of calibrating inertial measurement unit, and particularly relates to a low-complexity self-calibration method for IMU array. The present application firstly aligns the coordinate system of IMU on the IMU array by using iterative closest point algorithm to obtain the initial attitude of each IMU, and then establishes signal model for the accelerometer and gyroscope respectively, and solves the geometric distance of each IMU by using maximum likelihood estimation. The present application can effectively solve the problems of sensor installation error and unknown physical position of sensor chip internal sensing elements, thereby improving the applicability and positioning accuracy of complex array. Therefore, the low-complexity self-calibration method for IMU array proposed by the present application is a stable calibration method which can be well applied to different array types.

The application discloses a lithium battery data generation method and system based on an equivalent mileage and diffusion model. The method first acquires battery-related data in the running process of a battery pack of an electric vehicle, obtains a local charging and discharging segment sample, constructs an equivalent mileage calculation model, obtains a current equivalent aging mileage and a life end equivalent mileage, and combines the local segment starting SOC to obtain a joint condition label for controlling the data generation process. Then, the local charging and discharging segment sample and the joint condition label are extracted and fused, and through an encoder and reparameterization, a latent variable representation is obtained. Finally, based on the latent variable representation and the conditional features, through a conditional latent space diffusion model and a decoder of a conditional variational autoencoder, local charging and discharging data are generated. The application helps to alleviate the problems of lithium battery real vehicle data scarcity, insufficient life stage coverage and insufficient SOC interval sample and the like.

The high dynamic range imaging network method based on pyramid structure spatial feature transformation is based on three sub-networks: a deformable convolutional alignment network, a PSFT conditional network, and a fusion network of deformable convolutional residual dense blocks. The process is as follows: 1) The exposure of the input image is aligned using gamma correction, and then all images with and without exposure alignment are input into two convolutional layers of the deformable convolutional alignment network with shared offsets to obtain preliminary image features; 2) The exposure-aligned image features are geometrically aligned in the deformable convolutional alignment network with shared offsets, and the offset obtained in the geometric alignment process is also applied to the LDR image features without exposure alignment for geometric alignment; only gamma correction is used to align the exposure; 3) The geometrically aligned LDR image features without exposure alignment are input into the PSFT-based conditional network to obtain optimized features and obtain a high dynamic range image.

The application provides a small target detection method and system based on an improved YOLOv8 structure, which comprises the following steps: pre-processing a to-be-detected image to obtain an input standardized image; inputting the standardized image into a backbone network of YOLOv8 to extract feature maps of different scales; inputting the feature maps into a neck network of YOLOv8 for feature fusion, introducing an ASF mechanism into the neck network, integrating a scale sequence feature fusion module and a three-feature encoder module, cross-scale enhancing and integrating features of different resolutions to generate multi-scale fusion feature maps; based on the fusion feature maps, constructing P2, P3, P4 and P5 detection layers; the P2-P5 detection layers respectively predict candidate target boxes and their class confidence; adopting soft non-maximum suppression to process the candidate target boxes, adjusting the confidence by using a linear or Gaussian attenuation function, suppressing redundant boxes and outputting final detection results. The application significantly improves the recognition ability of small targets in dense and complex scenes.

The present application relates to biological signal processing technology, and is a non-invasive continuous blood pressure estimation system based on a subject adaptive feature modulation mechanism, an individual prior embedding vector is generated by an individual prior embedding vector generation module; a subject adaptive feature extraction module performs real-time multi-level dynamic modulation on the preprocessed time series physiological signal according to the individual prior embedding vector, and outputs a high-dimensional time series feature map; a sequence processing module performs time series dependence modeling and context refining on the high-dimensional time series feature map, and outputs a time series feature; a cross-modal attention fusion module splices the individual prior embedding vector and the time series feature; the spliced feature vector is input into a multi-head attention fusion module to realize dynamic cross-modal feature weighted fusion and obtain a high-dimensional feature vector; and a blood pressure regression output module maps the high-dimensional feature vector through a full connection layer to regress to obtain an estimated value of systolic pressure and diastolic pressure. The present application solves the problem that the existing static mapping paradigm cannot cope with individual heterogeneity.

The invention discloses a radiology report generation method and system based on focus guide mask and knowledge graph enhancement, and belongs to the technical field of medical artificial intelligence. The method comprises the following steps: preprocessing a chest medical image and extracting local features; constructing a lesion knowledge graph containing an organ-lesion-disease category triple; obtaining a candidate organ and a mask image thereof through a pre-trained focus detection model; retrieving related knowledge from the knowledge graph based on the candidate organs to generate focus enhanced knowledge features, and inputting the organ mask and the original image into a mask image attention layer together to obtain mask enhanced image features; using a multi-branch gating cross-modal fusion module to carry out adaptive weighted fusion on the two types of features to obtain knowledge mask enhanced fusion representation; and finally, a radiology report is generated through a Transform encoder-decoder. According to the method, image-text alignment of organ granularity can be realized, language illusion in a generated report is effectively reduced, and clinical consistency and interpretability of the report are improved.

The utility model relates to the technical field of adhesive mixing, in particular to an adhesive preparing, mixing and blending device which comprises a base body and a stirring assembly, the stirring assembly comprises a driving part, a center stirring assembly and an eccentric stirring assembly, the driving part is fixedly installed on the top of the base body, the center stirring assembly is coaxially and fixedly installed on the driving part, and the eccentric stirring assembly is coaxially and fixedly installed on the center stirring assembly. The eccentric stirring assembly is mounted on the horizontal side of the central stirring assembly, and the driving part drives the central stirring assembly to rotate and drives the eccentric stirring assembly to reciprocate for displacement stirring at the same time; therefore, the central stirring assembly drives the eccentric stirring assembly to continuously change the stirring position in the horizontal circumferential direction while enabling the adhesive and the additive to continuously turn up and down, so that the adhesive and the additive form a complex flow field; furthermore, under the mutual cooperation of the central stirring assembly and the eccentric stirring assembly, the effect of mixing, stirring and blending the adhesive and the additive is improved, and the problem that in the adhesive mixing and blending process of the adhesive mixing and blending device, uneven mixing is likely to happen is solved.

This invention relates to the field of *Dictyophora indicum* cultivation technology, and in particular to a rapid propagation method for *Dictyophora indicum* using a tiered cultivation system and its application. The method includes: a cultivation rack for cultivating *Dictyophora indicum*; a cultivation box fixed inside the cultivation rack; a collection mechanism located at the bottom of the cultivation box inside the cultivation rack; and a temperature regulating mechanism located inside the cultivation rack. The collection mechanism includes a first rectangular plate fixedly installed at the bottom of the inner cavity of the cultivation rack. The beneficial effect is that: manually rotating the gear via a throttle handle, since the second circular block on the gear is inside the second slot on the swing arm, the rotation of the gear can drive the swing arm to rotate. Furthermore, since the first circular block on the slider is inside the first slot on the swing arm, the rotation of the swing arm drives the slider to move along the first sliding rod, carrying the first stirring rod. This facilitates rapid stirring of the water and ice in the hollow tank through the movement of the first stirring rod, allowing the ice and water to fuse and create new cooling water.

The invention discloses a parasitic ovum detection method based on IDF-YOLO, and belongs to the technical field of image processing. The method comprises the following steps: constructing an IDF-YOLO target detection module with an enhanced feature fusion characteristic, and selecting CSPDarkNet64 as a backbone feature extraction network; a C3k2-IDWC module is designed in combination with the ideas of multi-scale convolution and partial convolution, and a C3k2 module in the deep layer of the model is replaced by the C3k2-IDWC module; the v7DSP module is used for replacing most of down-sampling modules; a WPFIoU loss function is used for replacing an original CIoU loss function; and obtaining an optimal model to detect a to-be-detected parasitic ovum image. According to the method, a small number of model parameters are adopted, the higher lightweight parasite ovum recognition accuracy is achieved, the calculation burden is greatly relieved, the method can be deployed to limited equipment for efficient detection, and therefore auxiliary parasite diagnosis can be achieved.

The application discloses a multi-modal medical image fusion method and system based on function-anatomy Transformer. First, a multi-modal medical fusion model is constructed: the SPECT image is subjected to RGB-YUV color space conversion to obtain three components Y, U and V; the Y component is input into a function multi-scale branch to extract multi-scale function features; meanwhile, the MRI is input into an anatomy multi-scale branch to extract multi-scale anatomy features; a function-guided Transformer module and an anatomy-guided Transformer module are designed to interact the multi-scale function features and the multi-scale anatomy features; the extracted multi-scale function features and the multi-scale anatomy features are integrated and reconstructed to obtain the Y component of the fusion image; pixel loss and gradient loss are constructed to constrain the fusion image to have similar pixel information as the source image and guide the fusion image to have similar gradient distribution as the source image; the pixel loss and the gradient loss are combined to jointly optimize the constructed multi-modal medical image fusion model; in the test stage, the optimized multi-modal medical image fusion model is used to obtain a fusion result.

The invention discloses a multi-modal image fusion method and device based on a graph neural network and a storage medium in the technical field of medical image processing and artificial intelligence. The method comprises the following steps: acquiring a multi-modal image; performing depth feature extraction on the multi-modal image to obtain a high-dimensional feature map of each modal; dividing the high-dimensional feature graph of each modal into a plurality of graph nodes, calculating the similarity between each graph node and an adjacent node thereof, and constructing an adjacent matrix of a multi-modal graph structure according to the similarity; taking an adjacent matrix as graph topology guidance, capturing a global association relationship and a long-distance dependency relationship of node features in graph nodes, and performing global aggregation according to an attention pooling graph reading function to generate global features; and inputting the global features into the reconstruction network to generate a final fusion image. The technical problems that in the prior art, the modeling capacity for irregular data is limited, and complex spatial association and semantic dependence between multi-modal images are difficult to effectively capture can be solved.

The utility model relates to the technical field of soil detection pretreatment shake-up, and discloses a rotary shake-up device for soil detection pretreatment, which comprises a placing seat, a base and a connecting seat, a servo motor is arranged on the outer surface of the connecting seat, a worm is fixedly mounted at the output end of the servo motor, and the worm is connected with the base. The outer surface of the worm is in meshed connection with a worm wheel part, and the outer surface of the worm wheel part is in meshed connection with a first rack plate and a second rack plate. According to the rotary shaking-up device for soil detection pretreatment, the servo motor is controlled to operate, so that the worm drives the worm wheel part to conduct linkage, and the first fixing part is driven to be close to the rear fixing part; the placing seat and the protective cover are correspondingly connected, the clamping piece and the clamping groove are clamped and fixed, then the first motor is controlled to operate, and the connected placing seat and protective cover are rotated left and right, so that the effects of detecting multiple samples, improving the fusion effect of soil and reaction liquid and reducing the influence on the experimental result are achieved.

This invention relates to a prosthesis for implantation into bone defects. Specifically, it discloses a cranioplasty component, comprising: a bone flap body for filling a defect in the skull, the bone flap body including a disc-shaped filling body and multiple stimulation units arranged circumferentially between the outer and inner disc surfaces of the filling body, and each stimulation unit being capable of protruding from the sidewall of the filling body through radial movement; a force-applying component for driving each stimulation unit to protrude radially from the sidewall of the filling body, thereby, after the bone flap body fills the defect, the force-applying component drives each stimulation unit to protrude from the sidewall of the filling body, causing the radially outer ends of the stimulation units to press against the sidewall of the defect to stimulate the bone of the sidewall of the defect; and a positioning component for attaching and positioning the bone flap body to the skull.

The application provides a remote sensing image fusion method considering image distortion and a computer readable medium. The application solves an intersection polygon of an image participating in fusion, constructs a minimum circumscribed rectangle of the intersection polygon according to a geographical range of the intersection polygon. After grid division of the minimum circumscribed rectangle, initial matching is performed, and affine transformation processing is performed on the corrected hyperspectral resolution image based on the matching result. The intersection polygon of the image participating in fusion and the corresponding minimum circumscribed rectangle are updated, grid division is performed to obtain a plurality of blocks, and fine matching processing is performed, and further polynomial correction is performed on the hyperspectral resolution remote sensing image subjected to the affine transformation processing based on the matching result. Each block of the image participating in fusion is subjected to fusion processing, so that a fused remote sensing image is obtained. The application can effectively solve the problem that the fusion effect is poor due to the internal distortion of the image, and can realize fusion between large-width remote sensing images.

The invention discloses a pod rapid self-alignment method based on multi-sensor joint estimation, and relates to the technical field of safety management, and the method comprises the following steps: I, initializing a multi-source sensor and collecting environment information; iI, constructing an adaptive filtering structure and estimating the initial attitude of the pod; iII, analyzing the causal relationship of the multi-source observation data and evaluating the credibility of the multi-source observation data; iV, correcting the postures of the pods and eliminating redundant information in the multi-source observation data; the method can significantly shorten the overall alignment time, improve the system starting efficiency, avoid the influence of single sensor failure or noise increase on the system, improve the data fusion effect, dynamically adjust the filtering structure according to the actual operation condition, improve the adaptability to the complex environment, and improve the reliability of the system. A non-linear noise mode which is difficult to find by a traditional statistical method can be identified, abnormal data are stripped through a noise index positioning and data reconstruction mode, and the observation data quality is improved.

The invention provides an underground cable three-dimensional imaging method based on terahertz data and X-ray data, and the method comprises the steps: solving dielectric constant distribution and material density distribution according to the terahertz data and the X-ray data, building the incidence relation between the dielectric constant distribution and the material density distribution, and determining projection reconstruction initial parameters; a fitness function is constructed according to the projection data and the projection reconstruction initial parameters, dielectric constant distribution and material density distribution are reconstructed through a quantum genetic optimization algorithm based on the fitness function and the incidence relation, and an initial reconstruction result is obtained; performing cross iterative optimization on the dielectric constant distribution and the material density distribution by using a regularization item constructed by an initial reconstruction result until convergence; and based on the converged dielectric constant distribution and material density distribution, after an X-ray reconstruction image and a terahertz reconstruction image are generated, data fusion and feature optimization are carried out, and a three-dimensional image of the underground cable is obtained. Therefore, a high-precision three-dimensional image of the underground cable is generated by fusing the optimized bimodal image data.

The invention discloses a zero sample image restoration and enhancement method and system based on text guidance, and the method comprises the steps: obtaining an original degraded image, and constructing a positive and negative text pool; respectively generating a restoration map and a degradation map through the restoration network and the degradation network; prior loss is constructed; obtaining a reconstructed restored image and a reconstructed degraded image through the restoration model and the degradation model, and calculating reconstruction loss; converting the original degraded image, the restored image and the reconstructed restored image into image embedding vectors, converting texts in the positive and negative text pools into text embedding vectors, and screening to obtain positive and negative text embedding vectors; mapping an image embedding vector of the restored image and positive and negative text embedding vectors obtained by screening to a multi-mode embedding space to calculate and compare learning loss; and constructing a total loss function based on the prior loss, the reconstruction loss and the contrast learning loss to carry out iterative training, and outputting a restored image output by the restoration model as a final result. According to the invention, the fusion capability of the multi-modal information and the image restoration and enhancement quality can be improved.