59 results about "Hash coding" patented technology

The invention discloses a dynamic airspace gridding management method and system for low-altitude economy, and belongs to the technical field of unmanned aerial vehicle traffic management, and the method comprises the steps: collecting airspace state data in real time through a multi-source sensing device, and constructing a four-dimensional space-time grid model; generating a four-dimensional space-time grid with a block chain hash code by fusing meteorological data, an airspace control rule and a real-time flight demand; receiving a space-time grid use request submitted by the aircraft through the smart contract, and calculating an optimal grid allocation scheme based on a deep reinforcement learning model; and the edge computing node executes local track prediction, issues a navigation instruction to the aircraft through the distributed account book, monitors a grid occupation state in real time, and triggers a dynamic grid recombination mechanism when sudden conflicts are detected. According to the method, the rigid constraint of static airspace division can be broken through, the cooperative conflict of multiple aircrafts is eliminated, and the marketization configuration of airspace resources is realized.

The invention discloses a general model pre-training and self-adaptive optimization system and method for a remote sensing image in the technical field of remote sensing images. The method comprises the following steps: acquiring a structured data set containing multi-modal remote sensing core data and corresponding auxiliary geographic information; based on the structured data set, utilizing a two-channel feature extraction architecture to synchronously process visual spectral features and geographic spatial features, and performing model training in combination with innovative training tasks of geographic context contrast learning and multi-temporal mask reconstruction to construct a pre-training model with general characterization capability; and when the pre-training model is applied to a target domain, performing cross-domain feature distribution alignment by utilizing a dynamic gradient inversion layer, and selectively unfreezing part of model parameters by adopting a task perception unfreezing strategy to complete adaptive migration. According to the method, through a self-supervised pre-training architecture of geographic coordinate hash coding and multi-temporal mask reconstruction, the spatial-temporal feature extraction robustness is enhanced, and the problem of generalization of a small sample scene is solved at the same time.

The invention provides an immersive video coding method and system based on 3DGS, and the method comprises the steps: carrying out the sparse reconstruction of each frame of multi-view video data, and obtaining an initial point cloud; determining anchor points of three-dimensional Gaussian distribution in each frame; extracting spatial context information of the anchor points through multi-resolution hash coding, and splicing the spatial context information with the features of the anchor points to form fusion features; predicting parameters of three-dimensional Gaussian distribution corresponding to each anchor point through a neural network by using the fusion features and camera parameters, and obtaining a rendered image; calculating the color loss between the rendered image and the original video frame, and optimizing the parameters of the anchor points; performing quantization and entropy coding on the optimized parameters of the anchor points; in combination with the color loss and the coding rate, performing rate distortion optimization on the quantization parameter to generate compressed three-dimensional scene representation; and repeating the steps for each frame of the multi-view video, and finally outputting a compressed immersive video code stream. According to the invention, high-quality representation and efficient compression of the immersive video are realized, and the rate-distortion performance is improved.

The application belongs to the field of three-dimensional vision and digitization, and relates to a three-dimensional human body reconstruction method and system based on three-dimensional Gaussian splashing. The method steps are as follows: based on a layered hash coding parameter field, the center position of each Gaussian primitive in the constructed three-dimensional Gaussian primitive set is corrected, and the color of the Gaussian primitive under the current observation angle is predicted; based on the human body posture parameters and shape parameters corresponding to the monocular video sequence, linear mixed skin transformation is performed on the corrected Gaussian primitive to map to the posture space; based on the Gaussian primitive parameters mapped to the posture space, three-dimensional Gaussian differentiable rendering is performed on the Gaussian primitive mapped to the posture space to obtain a rendering image consistent with the corresponding view angle of the monocular video; based on the constructed joint loss function, the Gaussian primitive parameters and the layered hash coding parameter field are optimized to obtain a three-dimensional human body model. The application can quickly reconstruct an animatable three-dimensional human body model with stable contours and clear textures from a monocular video.

The large-scale image search method based on deep hash comprises the following steps: acquiring a query image uploaded by a user and preprocessing the query image to obtain a to-be-searched image; the to-be-searched image is input into a pre-trained image Hash coding model, the image Hash coding model comprises a feature extraction network and a Hash mapping network, and the feature extraction network is constructed based on Vision Transform and is used for carrying out feature extraction on the to-be-searched image to obtain image high-dimensional features; the Hash mapping network is used for performing Hash coding on the image high-dimensional features to obtain search Hash codes; calculating the similarity between the search hash code and the in-library hash code of the inventory image in the image library based on the Hamming distance; and screening the inventory images based on the similarity to obtain a target image and feeding back the target image to the user. The invention provides a deep hash-based large-scale image search method, an image recommendation method, a recommendation system and computer equipment, which have better search performance.

The present application relates to a kind of asteroid global terrain three-dimensional reconstruction method for illumination change, comprising the following steps: input asteroid optical image, estimate camera pose and generate sparse point cloud, initialize Gaussian radiation field;Using image segmentation large model outputs the segmentation mask of asteroid optical image, prunes Gaussian radiation field;In combination with the mixed mask output by Gaussian radiation field, identify the shadow area of asteroid surface;The camera position, spherical harmonic coefficient and hash coding are combined into incident light angle, input into neural network, and the color value of Gaussian cell of illumination and shadow area is regressed;Add depth normal consistency loss and rotation variance and scale constraint loss, combine photometric loss to jointly supervise the optimization of Gaussian radiation field, calculate Gaussian opacity field, extract the three-dimensional model of asteroid global terrain.The present application can identify the shadow area on asteroid, improve the three-dimensional reconstruction precision of asteroid with surface illumination change.

The application discloses a kind of wisdom energy multi-energy evaluation visualization system and method for industry, it is related to energy big data visualization cross technical field, including, acquisition multi-source heterogeneous data and pre-processing, by space-time hash coding method, generate high-dimensional energy data of space-time alignment;Based on three-dimensional manifold coordinate model, the equipment energy efficiency parameter is converted into equivalent energy quality field, by energy field reconstruction and geometric optimization method, generate curved manifold space and identify optimal transmission path and distortion intensity cloud chart;Based on distortion intensity cloud chart, generate thermal color scale by thermal color scale mapping algorithm, and the three-dimensional gradient field distribution of equivalent energy quality field is rendered using ray tracing method, generate visual energy evaluation interface.The application generates space and time hash code by space-time hash coding method, establishes equipment-grid mapping relationship in combination with GeoHash coding, realizes the accurate space-time alignment of multi-source heterogeneous data.

The invention provides a point cloud frame searching and matching method and device, electronic equipment and a storage medium, and the method comprises the steps: determining first descriptor information of each point cloud frame in a point cloud frame sample set, and determining first hash coding information corresponding to each piece of first descriptor information; determining second descriptor information of the input point cloud frame, and determining second hash coding information corresponding to the second descriptor information; according to the first Hash coding information and the second Hash coding information, a similar candidate subset is determined from the point cloud frame sample set, the similar candidate subset comprises M point cloud frames, and M is a positive integer; and determining a target point cloud frame from the similar candidate subset according to the input point cloud frame. According to the method, descriptor hash search is added for screening, so that a large amount of time consumed by directly traversing and searching descriptor information can be avoided, and the effect of greatly reducing the calculated amount of point cloud frame search matching based on descriptors is achieved.

The application provides an optimization method and device of a hash encoder, electronic equipment and a storage medium, and belongs to the technical field of vector coding. The method comprises the following steps: determining a dimension importance score according to the information entropy and variance of each vector dimension of training data, selecting N key dimensions with the highest importance after eliminating highly relevant redundant dimensions to construct a rotation matrix, and configuring a plurality of pre-training hash encoders; determining a target hash encoder, encoding input data, and monitoring the coding quality in real time; when quality drift is detected, triggering a parameter optimization mechanism, adjusting the configuration parameters of the target hash encoder in multiple optimization rounds; until the optimization target is reached, applying the optimized target configuration parameters; when the incremental threshold is reached, constructing a rotation matrix increment based on the N key dimensions of the new data, weighting and fusing the current rotation matrix and the rotation matrix increment to obtain an updated rotation matrix, and updating the target hash encoder. By using the application, adaptive parameter optimization can be realized.

The application provides a similar vector existence query method, system, device and storage medium, relates to the similar vector retrieval technical field, and the method comprises an index construction process and a query process; the index construction process firstly carries out local sensitive hash coding on a preset vector library to obtain a coding library, and records coding parameters simultaneously; then, a plurality of integer number vectors of the coding library are input to a plurality of Bloom filters; the query process firstly calculates a coding search range based on a preset distance threshold and the coding parameters; then, local sensitive hash coding is carried out on a query vector to obtain a query coding vector; finally, the query coding vector is input to the plurality of Bloom filters, and the existence of a similar vector is judged in combination with the coding search range. The application effectively combines the two technologies of local sensitive hash coding and Bloom filter, can dynamically adjust the search range according to the distance threshold during the query, and achieves the query effect without missing detection.

The invention discloses an agile robust APT detection method based on traceability graph multi-view comparative learning, and aims to solve the problems that existing APT detection depends on fixed coding, generalization is weak, single sampling omits key information, and agility is insufficient. Screening core fields such as entity IDs and event types and filtering invalid logs; constructing a continuous time dynamic traceability graph, and defining processes, files and sockets as nodes; a read-write event and a connection event are edges, and node attributes are coded by adopting hierarchical feature hash; time and structure subgraphs containing positive and negative samples are generated through eta-BFS and-DFS sampling; generalization graph representation is obtained through multi-view comparative learning, and edge type prediction pre-training optimization is supplemented; and calculating edge reconstruction loss, a node deviation score and a queue abnormal score, and judging that the APT attack exists if a preset threshold value is exceeded. According to the method, detection is stable in a benign event injection confrontation scene, an attack core area can be quickly positioned, and robustness and agility are outstanding.

The invention discloses a multi-source data view feature extraction and quick matching system based on deep learning, and relates to the technical field of computer deep learning, a multi-modal data preprocessing module supports access of various heterogeneous data, technologies such as adaptive normalization are adopted to process data, a feature extraction module constructs a multi-branch attention network architecture, and a multi-source data view feature extraction and quick matching system is established. The trunk uses improved ResNet50 and introduces a channel attention mechanism, output of each branch generates a joint feature through a tensor fusion layer, the matching retrieval module adopts a layered hash coding strategy and a multi-index hash table, the weight learning module dynamically adjusts modal weight through a meta-learning framework and a context awareness mechanism, and the matching retrieval module performs matching retrieval on the modal weight through the meta-learning framework and the context awareness mechanism. The semantic enhancement module constructs a knowledge graph embedding layer and develops a semantic similarity calculation function. The multi-branch attention network improves the accuracy of feature extraction, distributed computing acceleration supports high-concurrency processing, and the method also has the functions of incremental learning, interpretable analysis and security privacy protection, and meets various requirements.

The invention relates to a lightweight stereo matching method based on weight sharing and channel attention. The left image and the right image pass through a weight sharing feature extractor to obtain a feature map, and an SE channel attention mechanism is inserted behind a residual block of the feature extractor for feature calibration; when a multi-scale correlation body is constructed, sparse indexing is carried out on parallax dimensions by adopting Hash coding based on space coordinates, and voxel sampling and trilinear interpolation with constant time complexity are realized; and obtaining a final disparity map through a variable-resolution iterative updating strategy. According to the method, the quantity of model parameters is effectively reduced, EPE and D1 indexes on a Middlebury data set are superior to those of an existing RAFT-Stereo method, and the method is suitable for real-time scenes such as automatic driving and robot navigation.

The invention discloses an all-weather scene adaptive reconstruction method based on non-uniform point cloud and depth map fusion, and the method comprises the steps: collecting multi-modal sensor data, carrying out the time-space alignment, and unifying the data to a world coordinate system; performing multi-scale sparse Hash coding on the data after space-time alignment to generate high-dimensional feature representation of the query point; implicit neural scene representation decoding based on weather perception: inputting the high-dimensional feature representation and the weather condition vector into a geometric decoder, and outputting a symbol distance function value and a geometric feature vector; gradient-level adaptive fusion is realized by using a dynamic prior weight network, a dynamic weight is generated according to a spatial position, an environmental condition and local uncertainty, and gradient back propagation in a training process is modulated; and obtaining a finally reconstructed three-dimensional scene model through end-to-end joint training and scene reconstruction. According to the method, high-precision and high-robustness unified reconstruction of non-uniform point cloud and depth map data under extreme weather interference is realized.

The application relates to the technical field of nasogastric tube intubation, and discloses a pressure regulation and monitoring method and system for nasogastric tube intubation, wherein the pressure regulation and monitoring method for nasogastric tube intubation comprises the following steps: acquiring real-time pressure monitoring data of a nasogastric tube intubation process, and generating a pressure mode feature vector; mapping the pressure mode feature vector to a low-dimensional hash space by using a region-sensitive hash algorithm, and generating a hash code; carrying out similar case retrieval in a multi-level hash index structure based on the hash code, and outputting a candidate similar pressure mode set; analyzing the similar pressure modes and corresponding historical control strategies retrieved, and generating a control strategy recommendation for a current intubation scene; the application solves the technical problem of real-time retrieval of similar pressure modes in a large-scale historical case library, realizes millisecond-level pressure mode identification and control strategy recommendation, and improves the decision-making efficiency and safety of the nasogastric tube intubation process.

The application discloses a large-scale image retrieval hash method focusing on the relationship between image blocks, characterized in that: firstly, multi-scale features are extracted from samples through a first self-attention calculation module and a second self-attention calculation module; then, a global average pooling layer, a hash layer and a classification layer are added to preliminarily extract features of images in a database and to use a trained hash retrieval model to perform hash coding to obtain a hash code retrieval library; features of an image to be retrieved are preliminarily extracted and the trained hash retrieval model is used to perform hash coding to obtain hash codes of the image to be retrieved; the hash code retrieval library is searched for data with the closest Hamming distance to the hash codes of the image to be retrieved, and the original sample corresponding to the data is displayed as a retrieval result, and the retrieval process is completed; the method has the advantages that the calculation amount is greatly reduced, rich multi-scale feature information can be extracted, the obtained hash codes are efficient and compact, and the image retrieval efficiency is relatively high.

The application discloses a kind of complex equipment system visual simulation method and system based on digital twinning, comprising: S1, obtains multi-source observation data and control data;S2, constructs digital twinning mechanism model, establishes state mapping relationship;S3, constraint perception hash addressing mechanism is introduced by Instant-NGP algorithm, and multi-resolution hash coding is executed;S4, state deviation is calculated;S5, input improved 4DGS model, introduce mechanism residual error gate projection mechanism, and execute gate projection;S6, based on SDS constraint algorithm, introduce feasible region constraint calculation of distillation gradient, and execute gradient update;S7, execute consistency determination, and output visual simulation result.The application realizes the high consistency, high stability and high real-time visual simulation output of complex equipment system under complex working condition, strong interference and multiple constraint conditions.

The invention relates to the technical field of big data analysis, in particular to a space-time big data acquisition and analysis system and method based on the Internet of Things, and the method comprises the following steps: inserting a drift compensation tag at the tail part of each time grid block, and outputting the time grid block with the drift compensation tag; combining the associated data items into a logic data unit; performing semantic isomorphic compression on the logic data unit and the data items with the same spatial hash codes at an edge node to generate a compression block with a spatial-temporal index; and the cloud server receives the compressed block, performs time conflict disambiguation and event-level causal reasoning on the spatio-temporal index by using a causal chain engine, and directly outputs drifting-free and redundancy-free spatio-temporal knowledge which is used for driving high-order application. According to the method, cross-space causal path matching and standard time sequence verification of the uploaded data are realized, and a plurality of actually occurring event-level causal chains can be automatically identified.

The invention discloses a pedestrian re-recognition system and method based on Hash coding learning and application. The system comprises a pedestrian feature extraction module, an uncertainty estimation and loss calculation module, a model training module and a retrieval module. The pedestrian feature extraction module comprises a Hash main network and a momentum auxiliary network which are consistent in network structure, and is used for performing feature extraction on an input pedestrian image and outputting corresponding activation features; the uncertainty estimation and loss calculation module is used for calculating bit-level uncertainty and image-level uncertainty based on the activation features and constructing a total loss function; the model training module is used for training and optimizing the Hash main network and the momentum auxiliary network based on the private training data set; and the retrieval module is used for carrying out offline library building on the pedestrian images in the base library and carrying out online retrieval on the to-be-queried images. According to the method, the storage pressure and the comparison speed pressure of the pedestrian re-recognition features can be remarkably reduced, and the storage and calculation efficiency of a pedestrian re-recognition application system is improved.

The invention belongs to the technical field of machine learning, and relates to a heterogeneous federal map learning method and system. The method comprises the following steps: in each client, constructing a globally shared symbiotic space, and generating a unified target semantic prototype through a label propagation mechanism; generating a prototype distribution matrix, and uploading the prototype distribution matrix to a server side; generating a global prototype by aggregating the prototype distribution matrix of each client, and distributing the global prototype to each client; embedding and mapping the nodes into a hash bucket by using a hash function, generating aligned local hash codes and local anchor point embedding, and uploading the local hash codes and the local anchor point embedding to a server side; aggregating the local hash codes and the local anchor point embedding to obtain global hash codes and global anchor point embedding; global hash coding and global anchor point embedding are optimized through a graph auto-encoder and consistency constraint, and alignment of the concentrated graph is optimized. According to the invention, the flexibility and efficiency of federal map learning are improved, the data privacy and security are ensured, and the communication overhead is reduced.

This invention proposes a multi-class unified 3D anomaly detection method based on hash-coded neural fields, belonging to the fields of computer vision and industrial quality inspection technology. It solves the problem that existing 3D anomaly detection methods struggle to simultaneously handle multi-class unified modeling, pose change adaptation, and anomaly repair. The method includes: 1. Acquiring multi-class normal point cloud samples and performing pose normalization on the samples; 2. Extracting multi-scale geometric features from the normalized point clouds and training a class prediction model; 3. Constructing a class-conditionalized multi-resolution hash-coded neural field and training a signed distance field model using normal samples; 4. Performing pose normalization, class prediction, and prototype registration on the point cloud to be tested; 5. Querying the trained neural field to complete point-by-point anomaly localization and target-level anomaly determination; 6. Reconstructing the surface of the anomaly region based on the zero isosurface of the signed distance field to obtain the repaired point cloud result. This invention can achieve multi-class point cloud anomaly detection, localization, and repair under a single model, and has advantages such as good pose robustness, strong fine-grained geometric expression capability, and high detection efficiency.

The invention discloses an IP camouflage identification anti-crawler method and system based on Transform, and the method comprises the steps: (1) a data collection and preprocessing stage: collecting access log data of a target network, extracting IP fingerprint data, and carrying out the Hash coding of the IP fingerprint information to form a fingerprint code; (2) a feature construction stage: constructing a three-layer dynamic time sequence feature sequence X containing basic behaviors, disguise attributes and historical tracing; (3) a model reasoning stage: inputting the feature sequence X into a pre-trained Transform encoder; and (4) a decision and feedback stage: outputting a probability value that the IP address is a camouflage crawler, if the probability exceeds a preset threshold value, triggering a countering measure, and injecting a countering effect score as a newly added dimension into a next round of feature sequence for online fine adjustment of the model. According to the method, the identification accuracy under large-scale network traffic is greatly improved, the problem of concealment of an IP camouflage crawler is effectively solved, and the method has extremely high real-time performance and expandability.

The invention provides a calculation control method and system based on position coding, a chip and a storage medium, and the method comprises the steps: receiving the space coordinate information of to-be-calculated data, carrying out the nonlinear hybrid Hash coding based on the space coordinate information of the to-be-calculated data, and obtaining the space position Hash code of the to-be-calculated data, and based on the spatial position hash code of the to-be-calculated data, the calculation rule corresponding to the to-be-calculated data is selected, so that the to-be-calculated data is calculated based on the calculation rule corresponding to the to-be-calculated data, and the system can automatically select the appropriate calculation rule according to the hash code obtained based on spatial coordinate information coding. Therefore, more efficient and more flexible calculation processing is realized, and changes of various calculation scenes can be adapted more intelligently.

The invention discloses a method for identifying the amount of motion of an image and electronic equipment, and relates to the technical field of terminals. The first image and the second image are sequentially compared according to pixel units such as a row or a column, the same pixel units in the first image and the second image are obtained, and the motion amount of the second image relative to the first image is obtained according to the same pixel units. When the pixel units are compared in sequence, after the pixel units are subjected to one-way hash coding, hash values corresponding to the pixel units are compared in sequence, and if the hash values are the same, the pixel units are determined to be the same pixel units. In this way, rapid and accurate comparison can be achieved, the calculated amount is remarkably reduced under the condition that accuracy is guaranteed, and the efficiency of recognizing the exercise amount is improved.

This invention discloses a method and system for local scene 3D reconstruction based on an edge computing platform. It acquires synchronous temporal data of the target scene using multi-source sensors and performs dynamic interference removal; performs real-time pose estimation and global trajectory optimization based on an improved tightly coupled vision-inertial-laser SLAM framework; constructs and dynamically maintains an explicit symbolic distance field based on a sparse octree as a geometric prior representation; implicitly encodes the leaf node regions of the sparse octree using multi-resolution hash coding, and generates geometric residuals and color information through lightweight neural network decoding, which are then fused with the explicit symbolic distance field prior to form a layered hybrid scene representation; a keyframe selection strategy maximizing coverage is used to select keyframes for optimization; and multi-task joint optimization of the layered hybrid scene representation is performed based on differentiable volume rendering technology to output a 3D scene model. This invention achieves near real-time reconstruction with millimeter-level geometric accuracy and high-quality textures on an edge platform.

The invention discloses a multi-level video group event retrieval method. The method is characterized by comprising the following steps: firstly, carrying out object-level visual feature reconstruction and establishing a space-time relation graph between objects; performing multi-level fusion through a multi-level space-time fusion module; an event classification prediction result corresponding to the multi-level fusion result and a corresponding Hash code are obtained through a Hash module retrieval module, and then network training is performed to obtain each trained model; the method has the advantages that by combining action classification loss, event classification loss, Hash loss, Hash filtering loss and reconstruction loss of a single object, feature representation and a Hash function can be learned from an input video at the same time, and an obtained Hash code is efficient and compact, so that video event retrieval is effectively performed; and finally, sequentially carrying out matrix multiplication on the obtained Hash codes through the filtering matrix, so that the Hash codes are close to the Hash codes output by the upper layer, and the storage space of the multi-layer Hash codes is effectively reduced.

This invention discloses an incomplete cross-modal hash retrieval method based on predictive completion and auxiliary code guidance, belonging to the field of cross-modal hash retrieval technology. This invention employs a pre-trained CLIP model as a cross-modal encoder to extract semantic features from image and text modalities, effectively enhancing the semantic consistency of cross-modal data. A bidirectional predictive network is designed, combined with a variational inference mechanism, to efficiently complete missing modal features using available modal features, successfully addressing the problem of modal loss caused by data acquisition failures or transmission interruptions. A residual contrastive network is introduced to improve the discriminative ability of the completed multimodal features and achieve feature alignment, effectively mitigating the distribution shift caused by missing data. In the hash encoding stage, a high-bit auxiliary code is used as a semantic teacher, guiding the low-bit hash code to learn richer semantic representations through knowledge distillation, significantly reducing information loss during compression. This invention significantly improves the efficiency and accuracy of incomplete cross-modal retrieval.

The invention provides a CT (Computed Tomography) reconstruction method, system and equipment of a three-dimensional sparse view angle and a medium. The method comprises the following steps: acquiring projection data formed under a sparse view angle scanning condition and corresponding imaging geometric parameters; determining a ray path set based on the projection data, and carrying out the following processing on each ray path: generating a plurality of three-dimensional sampling point coordinates on the ray path according to imaging geometric parameters corresponding to the ray path; performing feature mapping and interpolation processing on the coordinates of the three-dimensional sampling points based on Hash coding to generate point feature vectors corresponding to the three-dimensional sampling points; inputting each point feature vector into a neural network model to obtain an attenuation coefficient corresponding to each three-dimensional sampling point; carrying out integral accumulation on the attenuation coefficient of each sampling point along the ray direction to obtain a projection result of the ray path; and generating a three-dimensional reconstructed image based on the projection result of each ray path. According to the method, high-quality and low-power-consumption three-dimensional CT reconstruction processing can be realized under the sparse view angle scanning condition.

The invention relates to the technical field of network security, and discloses a virtual identity association method based on a behavior graph structure and iterative hash coding, which comprises the following steps: collecting network behavior data of a virtual identity, normalizing the network behavior data into a unified behavior tag, and extracting attribute features corresponding to behaviors; taking a single behavior as a node, constructing a directed behavior graph with a time weight according to a behavior occurrence time sequence, and limiting a behavior graph scale according to a preset range; performing initial coding on the nodes of the behavior graph, aggregating node neighborhood information and time weight through multi-round iterative hash, and generating multi-granularity codes and global hash fingerprints; the behavior graph structure similarity is calculated based on multi-granularity coding, and the final similarity is obtained by combining with auxiliary statistical feature similarity fusion; and inputting the final similarity, the behavior graph characteristics and the time sequence characteristics into a logistic regression model, and completing virtual identity association judgment according to a logistic regression model output probability and a threshold value. According to the method, behavior characteristics of virtual identities under different platforms can be effectively mined, and associated virtual accounts can be found.

The invention particularly relates to an efficient privacy protection cross-modal retrieval method based on hierarchical indexing. The method comprises the steps that firstly, cross-modal hash codes of data are obtained through an unsupervised deep hash code learning model; then constructing a hierarchical index structure to realize efficient retrieval; encrypting the data and the index in combination with a security inner product protocol and a pseudo-random function, and outsourcing the data and the index to a cloud server; and finally, in a cloud environment, enhancing the security protection capability of the data while ensuring smooth interaction, supporting ciphertext matching and efficient retrieval between images and texts, and returning a result to a search user. Compared with an existing retrieval method, the method is low in calculation complexity and higher in retrieval efficiency.