20 results about "Event retrieval" patented technology

The invention discloses an operation and maintenance knowledge base analysis system and method based on deep learning, and is applied to unified operation and maintenance management of a hospital information system and medical equipment. The system collects operation and maintenance data such as operation logs, monitoring indexes, alarm events, work orders, maintenance records and the like, constructs a minute-level time window through equipment identification and technology triggering time, and merges multi-source heterogeneous data into operation and maintenance sub-events and operation and maintenance events. Log, index and text feature extraction is performed on each operation and maintenance sub-event, multi-modal feature fusion is performed to generate an operation and maintenance event feature vector, the operation and maintenance event feature vector is stored in a vector index database and is associated with an operation and maintenance knowledge graph, and similar historical event retrieval and root cause and processing scheme recommendation of a new operation and maintenance event are realized. In combination with the adoption and processing results of recommended schemes by operation and maintenance personnel, the system performs hierarchical management on the operation and maintenance knowledge and performs incremental updating on the feature modeling network, so that the operation and maintenance knowledge retrieval accuracy and the fault positioning efficiency are improved.

The invention discloses a bank data analysis processing method and system, particularly relates to the technical field of bank data processing, and is used for solving the problem that multi-source data is difficult to trace uniformly. Multi-service system records are accessed through a unified time reference and a subject identifier, scattered records are recombined into event frames with event types, time and subject sets based on event anchor point rules, conflict resolution is performed on multi-source same-name fields according to authoritative source priorities and delay propagation thresholds, main field values and alternative field values are reserved, and multi-source same-name fields are obtained. On this basis, an event frame table, a state sequence table and a participating entity table are constructed, indexes are established according to subject identifiers, time intervals, event types and state codes, retrieval and statistics around customer events are achieved, data consanguinity among source fields, event frame fields and analysis fields is synchronously established when the analysis fields are generated, and the data consanguinity among the source fields, the event frame fields and the analysis fields is established; and in combination with the version information, tracing of index sources in a rule change and late data scene is supported.

The invention relates to the technical field of information engineering knowledge graph construction, in particular to a supervision knowledge graph dynamic modeling method based on multi-source heterogeneous data fusion, and the method comprises the steps: designing a four-dimensional knowledge graph modeling framework, and combining the entity alignment of a graph neural network, the relation extraction of remote supervision and the attribute processing of weight fusion. A dynamic evolution and cross-source consistent knowledge network is constructed, virtual features are generated through a virtual feature compensation mechanism (historical event retrieval and multi-dimensional feature weighted fusion) for a multi-modal data missing scene and mapped to a graph, data are effectively filled, the problem of incomplete knowledge coverage in a complex engineering scene is solved, a compliance entropy evaluation model is introduced, and the knowledge coverage is evaluated. The compliance state of the knowledge graph is quantitatively supervised from the three dimensions of rule coverage rate, data consistency and tracing integrity, and the hysteresis and subjectivity limitation of traditional manual compliance check is broken through.

The present application relates to the field of large-scale industrial automation system. An event collection method of large-scale industrial automation system classifies the equipment involved in the large-scale industrial automation system and defines four-level standard events, and takes an event subject as the minimum collection unit when an event corresponding to each event subject occurs. When the submission period is met in time or the number of events in the event stack meets the batch submission number, each field of each event in the event stack is serialized, and the events in the event stack are packed to generate a data frame by a protocol stack program and sent to a central event server. In the central event server, the received events are stored in the time sequence database, the event subscription and query steps, the event retrieval terminal is connected with the central event server, and the events archived in the central event server are pushed to the event retrieval terminal through the subscription or query mode.

The present application relates to the technical field of intelligent transportation, and more particularly to a traffic event image-text mutual search method and system based on cross-modal retrieval, which comprises the following steps: obtaining query data of a traffic event, structurally processing the query data to obtain query semantic units; performing feature coding on the query semantic units to generate query semantic vectors; constructing a weak common semantic space, and setting an image index channel and a text index channel for the weak common semantic units; selecting an index channel according to the query data, mapping the query semantic vectors to the weak common semantic space for similarity matching, and generating a traffic event preliminary search result set; and performing fine sorting on the traffic event preliminary search result to output a traffic event search result set. Through the present application, the problems of low search efficiency and insufficient search accuracy in large-scale search are effectively solved, the matching deviation caused by modal difference is reduced, and the efficiency and business adaptability of traffic event search are effectively improved.

The application belongs to the technical field of computers and provides a fault diagnosis method and system based on equipment maintenance knowledge base. The method comprises the following steps: constructing an equipment maintenance knowledge base, wherein the equipment maintenance knowledge base comprises phenomenon events, fault causes and the relationship between the phenomenon events and the fault causes, the relationship is a fault scene, and the fault scene comprises a single cause scene and a combined cause scene; acquiring a last node in a fault diagnosis path as a phenomenon event, retrieving a fault scene of the phenomenon event, and if the retrieval result is empty, the phenomenon event is a fault diagnosis result. The single cause scene can realize the reasoning ability of the traditional decision tree rule, the combined cause scene is combined with the single cause scene, the fault combination condition traversal with exponential complexity can be automatically expanded based on a small amount of structured expression, the knowledge representation is simplified, and the complexity and workload of the construction and maintenance of the knowledge base are reduced.

The invention discloses a video event retrieval method based on space-time interlacing, which is characterized by comprising the following steps of: firstly, acquiring a hash code of a training sample, a video event classification prediction result of the training sample and action classification prediction of a single object through a feature extraction model to be trained, a space-time interlacing model to be trained and a hash retrieval model to be trained; performing training based on the constraint of the total loss function to obtain a trained feature extraction model, a trained space-time interlacing model and a trained Hash retrieval model; enabling a query sample and a retrieval sample to pass through each trained model to obtain a corresponding hash code of the query sample and a corresponding hash code of the retrieval sample; searching a Hash code closest to the Hash code of the query sample in a retrieval set Hash code retrieval library as a retrieval result, and displaying the retrieval result to complete the retrieval process of the query sample; the method has the advantages that the output hash codes are efficient and compact, and the retrieval accuracy is high.

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.

One embodiment comprises a non-transitory storage medium storing program code to receive one or more sensory events from a sensory event analytics module that receives sensory data about a physical environment from one or more sensors and processes the sensory data from the one or more sensors to detect the one or more sensory events. The one or more sensors include at least an Internet Protocol (IP) video camera. The one or more sensory events include one of a face detected, a vehicle detected, and a license plate detected. The program code includes code to store the sensory events for later retrieval as stored sensory events; and evaluate one or more historical correlations among the stored sensory events by evaluating the stored sensory events for the one or more historical correlations across at least one of time and space based on at least one weight vector.

The invention provides a short video false news detection method based on a thinking chain and mixed experts. The short video false news detection method comprises the steps of structured thinking chain construction, mixed verification expert module construction, mixed theme expert module construction, model training and parameter optimization and short video false news detection reasoning. According to the method, a structured thinking chain is designed, a multi-modal large language model is guided to simulate a human rumor recognition thinking process, news element extraction, event retrieval and multi-angle reasoning are completed in sequence, and fine-grained reasoning explanation is generated from three aspects of cross-modal consistency, fact verification and logic verification; meanwhile, a hybrid verification expert module is introduced, multi-angle evidences are integrated through a dynamic routing mechanism, key reasoning features are captured, and the judgment credibility is improved; and in combination with a mixed topic expert module, a discrimination standard is adaptively adjusted according to a short video topic scene, and the generalization ability and discrimination precision of the model under different topics are enhanced, so that interpretable robustness detection of the short video false news is realized.

One embodiment comprises a non-transitory storage medium storing program code to receive one or more sensory events from a sensory event analytics module that receives sensory data about a physical environment from one or more sensors and processes the sensory data from the one or more sensors to detect the one or more sensory events. The one or more sensors include at least an Internet Protocol (IP) video camera. The one or more sensory events include one of a face detected, a vehicle detected, and a license plate detected. The program code includes code to store the sensory events for later retrieval as stored sensory events; and evaluate one or more historical correlations among the stored sensory events by evaluating the stored sensory events for the one or more historical correlations across at least one of time and space based on at least one weight vector.

The present application relates to the technical field of information engineering knowledge graph construction, and particularly relates to a supervision knowledge graph dynamic modeling method based on multi-source heterogeneous data fusion. The method constructs a dynamic evolution and cross-source consistent knowledge network by designing a four-dimensional knowledge graph modeling framework, combining entity alignment of a graph neural network, relationship extraction of remote supervision, and attribute processing of weighted fusion. For a multi-modal data missing scenario, a virtual feature compensation mechanism (historical event retrieval, multi-dimensional feature weighted fusion) is used to generate virtual features and map them to the graph, effectively filling in the data and solving the problem of incomplete knowledge coverage in complex engineering scenarios. A compliance entropy evaluation model is introduced to quantify the compliance state of the supervision knowledge graph from three dimensions of rule coverage, data consistency, and traceability integrity, breaking through the limitations of the traditional artificial compliance check in terms of lag and subjectivity.

The invention discloses a potential evapotranspiration extreme event intelligent identification method for time sequence big data, and relates to the technical field of big data mining, and the method comprises the steps: collecting meteorological data, carrying out the quality control and restoration, calculating the daily potential evapotranspiration amount, writing in a column storage table of time and space partitions, and forming a potential evapotranspiration time sequence data set; obtaining a conditional quantile baseline value and a scale parameter according to the potential evapotranspiration time sequence data set, performing streaming updating, calculating a standardized residual error through the conditional quantile baseline value and the scale parameter, and generating a standardized residual error sequence; according to the standardized residual sequence, identifying continuous high-value time periods and generating event objects, calculating extreme strength indexes and sorting the extreme strength indexes to obtain an event curve and an event list; and constructing an event signature through the event curve and the event list and establishing an inverted index, and performing intelligent identification and result archiving on the potential evapotranspiration extreme event through similar event retrieval. And a comparable, searchable and reusable effect can be achieved for potential evapotranspiration extreme events.

The invention relates to the technical field of intelligent traffic, in particular to a traffic event image-text mutual search method and system based on cross-modal retrieval, and the method comprises the steps: obtaining query data of a traffic event, and carrying out the structural processing of the query data, and obtaining a query semantic unit; performing feature coding on the query semantic unit to generate a query semantic vector; constructing a weak common semantic space, wherein a weak common semantic unit is provided with an image index channel and a text index channel; selecting an index channel according to the query data, mapping the query semantic vector to a weak common semantic space for similarity matching, and generating a traffic event initial detection result set; and carrying out refined sorting on the traffic event initial inspection results, and outputting a traffic event retrieval result set. According to the method and the device, the problems of low large-scale retrieval efficiency and insufficient retrieval precision are effectively solved, the matching deviation caused by modal difference is reduced, and the efficiency of traffic event retrieval and the business suitability are effectively improved.

The invention discloses an event report generation method and device, equipment, a storage medium and a program product, and relates to the technical field of artificial intelligence. Based on a user input event, event retrieval is carried out in the internet resources and the private knowledge base, and event related information is obtained; screening and integrating the event related information to obtain event refining data; generating a plurality of event subtopics based on the event refining data; for each event subtopic, content supplementation is carried out based on event related information, and an event context containing all the subtopics is obtained; based on the output style framework, event contexts are sorted, a complete event report is obtained, the complete event report is input into a webpage generation model, and an event report visual page is generated in combination with cue words. The method has the technical effect of improving the generation stability of the event report.

A method is generally provided for adjusting a pattern projection rate in an active perception system, comprising the steps of observing a scene with an event-based sensor array; projecting a current light pattern of a sequence of light patterns on the scene for a given interval of time; retrieving events generated by the event-based sensor array in response to the projection of the current light pattern; monitoring a property of the event retrieval progress; and dynamically controlling the projection of the light pattern based on the property.

The present disclosure relates to a smart feeder bottle accessory system and method for enhancing infant feeding experiences. The system comprises a capping component containing an integrated sensor for detecting fluid dispensation, and a sleeve component containing electronic components. These components work in tandem to transform standard feeding bottles into interactive devices. The method involves detecting fluid dispensation, registering it as a dispensation event, retrieving associated feedback, and emitting this feedback through an output system. The system includes a memory device for storing various types of feedback and at least one hardware processor for event processing. Feedback may be audio, visual, mechanical, or sensory, providing a perceptible response to feeding events. The system aims to encourage proper feeding habits, engage infants, and provide valuable data to caregivers, thereby revolutionizing the infant feeding process.

The invention discloses a health monitoring system for a bridge engineering structure. The health monitoring system comprises an acquisition module, an anomaly detection module, a bridge anomaly judgment module, a previous event retrieval and analysis module, a matching module and an accident prediction and calculation module, the bridge anomaly judgment module is used for receiving and processing the anomaly information of the previous bridge engineering information structure acquired by the anomaly detection module, calculating an anomaly prediction value, judging the state of the current bridge according to the anomaly prediction value, and judging that the current bridge engineering structure is in an abnormal state when the anomaly prediction value is smaller than an anomaly threshold value; the bridge engineering structure health monitoring system overcomes the defect that in the prior art, when the bridge engineering structure is abnormal during health monitoring of the bridge engineering structure, accidents caused by abnormal information of the current bridge engineering structure cannot be analyzed; and economic loss caused by accidents caused by abnormal information of the current bridge engineering structure cannot be predicted.

This invention belongs to the field of intelligent video surveillance technology and discloses a factory hazardous area monitoring storage and intrusion behavior recognition and early warning system. By incorporating a distributed hierarchical storage module, it automatically configures differentiated storage duration and multi-copy backup strategies based on the intrusion hazard level and area risk level. The first-level solid-state drive enables real-time data and second-level abnormal event retrieval and playback, the second-level mechanical hard drive stores regular video, and the third-level offline archiving medium archives high-risk event data for a long time. At the same time, it generates an immutable hash verification chain for medium and high-risk events, and automatically cleans up abnormal data in batches on a periodic basis, realizing intelligent management of security data throughout its entire life cycle. This not only significantly reduces storage costs but also ensures the complete retention and secure traceability of critical data. The multi-modal preprocessing module completes the spatiotemporal alignment and adaptive interference filtering of video, infrared thermal imaging, millimeter-wave radar, and vibration sensor data.

The present application relates to an event pre-training method for Chinese-Vietnamese cross-language event retrieval, and belongs to the technical field of natural language processing. The present application uses two pre-training methods to additionally pre-train mBERT. First, event knowledge is integrated into the model by using event element mask pre-training to improve the model's representation of extremely low-resource events. Then, cross-language contrast learning is used to make sentences with similar meanings in different languages have a closer distance in the representation space. Then, a Chinese-Vietnamese cross-language event pre-training model is obtained, which is fine-tuned to make the downstream task have better performance. The experiment of the event pre-training method for Chinese-Vietnamese cross-language event retrieval on the self-built Chinese-Vietnamese bilingual news event retrieval dataset proves the effectiveness of the method.