25results about How to "Consider speed" patented technology

The invention provides an SQL (Structured Query Language) generation and error correction method based on multi-stage feedback, which comprises the following steps: S1, collecting user demands, and greatly improving the accuracy of generating SQL while keeping quick response by introducing an expert thinking model and by means of a mode of demand disassembly and task gradual combination, so that the user experience is improved; and it is ensured that effective query statements can be quickly generated in the face of simple or medium-complexity tasks. Common requirements can be efficiently processed, so that the query requirements of the user can be quickly met. And secondly, aiming at a task which is complicated or has a high-precision requirement, through a multi-round interaction and error correction mechanism, the accuracy of a generated result is ensured. When the SQL execution statement is generated for the first time, if an error occurs or a user is not satisfied, multiple rounds of error correction processes can be started, the SQL execution statement is corrected step by step, and finally high-precision output meeting the requirement is achieved. Through the double-link design, the speed and the accuracy are both considered, and the requirements in different scenes are met.

The invention belongs to the technical field of optical communication, and relates to a flexible silicon-based photoelectric device, a receiving end and a visible light communication system. The flexible silicon-based photoelectric device comprises a back electrode layer, a hole transport layer, a P-type silicon light absorption layer, an electron transport layer and a front electrode layer. The valence band side energy difference between the electron transport layer and the P-type silicon light absorption layer is greater than the conduction band side energy difference; the conduction band side energy difference between the P-type silicon light absorption layer and the hole transport layer is greater than the valence band side energy difference; the forbidden bandwidth of the electron transport layer is greater than 3.0 eV. The front electrode layer comprises a plurality of grid electrodes which are separately arranged on the surface of the electron transport layer. According to the scheme provided by the invention, the heterojunction is formed on the surfaces of the two sides of the P-type silicon light absorption layer, so that the intensity of a built-in electric field is effectively enhanced, and the flowing speed of photo-generated electrons and photo-generated holes in the device is accelerated, thereby improving the response speed of the device; and meanwhile, the collection efficiency of electrons and holes is improved, so that the photoelectric conversion efficiency of a photoelectric device is improved.

The invention relates to the technical field of radiation dose calculation, in particular to a radiation dose calculation method and device based on a V-Net network. The method comprises the following steps: acquiring a CT image and light beam flux information for a focus area of a patient to be treated, wherein the light beam flux information represents parameter information of a light beam for performing radiotherapy on the patient to be treated; and on the basis of the CT image and the beam flux information, through a pre-trained V-Net network, performing radiotherapy dose calculation for a focus area of a patient to be treated, and generating a dose distribution diagram. According to the method, the dose distribution diagram can be automatically generated, human intervention is reduced, and the clinical working efficiency and the adaptability of a treatment plan are improved. Moreover, the pre-trained V-Net network can effectively capture features in the CT image, the calculation speed is greatly improved, the precision of the generated dose distribution diagram is equivalent to the calculation precision of the Monte Carlo method, and both the speed and the precision are taken into account.

The invention provides a real-time safe distance monitoring method and system for field operation and a storage medium, and the method comprises the following steps: calculating the distance between a field operator and a potential dangerous object or region as a first distance based on a real-time image of the field operation, and executing the subsequent steps if the first distance is smaller than a first set value; otherwise, continuing to calculate the first distance in real time; based on the real-time three-dimensional position data of the field operation, calculating the distance between the field operation personnel and the potential dangerous object or area as a second distance; performing data fusion on the first distance and the second distance to obtain a third distance; if the third distance is smaller than the second set value, alarm operation is executed; otherwise, continuing to calculate the first distance in real time. According to the invention, the technical advantages and advancement of various intelligent detection methods are fully combined, and the safety distance of field operation is effectively monitored in real time.

The invention relates to a self-adaptive control method, system and device for environment test equipment, and the method comprises the following steps: S1, data collection: collecting environment detection data and data of a corresponding actuator according to a to-be-controlled physical quantity; s2, selecting or constructing a control object according to the physical quantity to be controlled; s3, performing online identification operation when an identification condition is met, otherwise, directly executing the step S4; s4, feedforward and feedback: calculating feedforward compensation and feedback compensation according to the value of the to-be-estimated quantity, and superposing the feedforward compensation and the feedback compensation to obtain a control instruction; and S5, performing constraint allocation on the actuator: performing saturation constraint processing on the control instruction to obtain a control quantity of the corresponding actuator. Joint online identification can be carried out on the to-be-estimated quantity only through perturbation excitation; different loads and test object states are automatically adapted; saturation constraint processing is adopted; the model and the algorithm are low in calculation amount, and precision, speed and engineering realizability are considered.

The invention provides a small target detection method and system based on an improved YOLOv7 algorithm, a storage medium and electronic equipment, and relates to the technical field of target detection. According to the technology, the expression effect of a YOLOv7 model in an urban traffic road end detection scene is improved from two aspects: an adaptive bounding box loss weight is introduced under the condition that a model network structure is not changed, and the small object detection precision of the model is improved; a detection head for a small object is added under the condition that the reasoning time is slightly increased, and the small object detection precision of the model can also be improved. On the premise that the detection cost is hardly increased, the small target detection precision of the YOLOv7 target detection algorithm is improved, and the speed and precision of small target detection are well considered.

The invention discloses a hierarchical topology aggregation visual display method for network threat associated data, and the hierarchical topology of threat association refers to the associated data of mechanisms, devices, software and threats, and the method comprises the following steps: (10) determining four layers of network topologies, i.e., dividing data into four categories according to a platform layer, a logic layer, an application layer and a threat layer; (20) determining a specific node classification display icon of each layer and performing four-layer visualization; (30) determining node and edge aggregation display of the application layer and the threat layer; (40) determining expansion of aggregation nodes and edges of the application layer and the threat layer; (50) determining a red marking and hiding visualization method of important and non-important nodes of the logic layer and the application layer; and (60) determining a visual layout algorithm and a visual tool of each layer. According to the method, the problem that rendering is too slow due to the fact that the data size of nodes of an application layer and a threat layer is too large during network topology visual display and multi-layer display is solved, and therefore a user can obtain network topology information including a platform, equipment of the platform, software on the equipment and threats on the software more visually and rapidly; and displaying through a graphic visualization tool g6.

The utility model discloses a photovoltaic pile foundation suitable for rocky desertification high abrupt slope soil-rock combined geology, which mainly comprises anchor bars, an anchor bar foundation small bearing platform and a steel sleeve, the upper end of the steel sleeve is provided with an adjusting bolt, and the lower end of the steel sleeve is arranged in the anchor bar foundation small bearing platform; the anchor bars penetrate through the bottom of the anchor bar foundation small bearing platform, the upper parts of the anchor bars are arranged in the anchor bar foundation small bearing platform, and the lower parts are arranged in the anchor bar holes; the steel sleeve is welded and connected with an anchor bar in the anchor bar foundation small bearing platform through round steel; and concrete is poured into the anchor bar holes and the small anchor bar foundation bearing platform. The photovoltaic pile foundation has the advantages of being simple and safe in structure, wide in application range, high in adaptability to local conditions and the like, can be widely applied to various terrain environments, and can be constructed on soil steep slopes, especially on rocky desertification high steep slope soil-rock combined geology which is difficult to construct through a conventional foundation.

The invention discloses a five-degree-of-freedom electromagnetic positioning system and method, and the system comprises a transmitting unit which comprises a plurality of single-axis transmitting coils and is used for generating a magnetic field in a three-dimensional space; the receiving unit is used for collecting field intensity information in a three-dimensional space; the control and data processing unit is used for inputting the field intensity information collected by the receiving unit into a pose initial value prediction network and outputting initial pose information of the receiving unit; inputting the initial attitude information into a coil magnetic field modeling network to obtain theoretical field intensity; and the iterative optimization module acquires a field intensity difference according to the theoretical field intensity and the acquired field intensity information, updates the attitude information according to the field intensity difference, and returns to execute the step of acquiring the theoretical field intensity until a preset iteration termination condition is reached, thereby acquiring the final attitude information. The method can reduce the complexity of a traditional electromagnetic positioning algorithm, improves the positioning precision and speed, can be widely applied to the fields of auxiliary medical treatment, man-machine interaction and motion tracking, and achieves a multi-degree-of-freedom electromagnetic positioning function.

The invention discloses a zooming method, a shooting module, terminal equipment and a storage medium, the zooming method is applied to the shooting module, the shooting module comprises a plurality of cameras and an image processing module, and the zooming method comprises the following steps: obtaining a zooming instruction; wherein the zoom instruction carries a mixed zoom ratio k, and k is greater than 0; and in response to the zooming instruction, carrying out mixed zooming processing through the plurality of cameras and the image processing module, and adjusting the focal length of the shooting module to be a k-time focal length.

The utility model discloses a multi-layer insulating glue pressing device for a dispensing driving plate, which is applied to the technical field of electronic engineering, adopts a two-side clamping design, can effectively improve the pressing stability and precision of parts, can uniformly distribute pressure, avoids deviation or deformation of a glue layer caused by single-side stress, and improves the product quality. Two-side clamping can adapt to parts of different sizes, the production compatibility and efficiency are improved, the manual adjustment time is shortened, and the clamping device is suitable for the automatic production scene of electronic elements; according to the vertical multi-layer synchronous pressing design, the production efficiency can be remarkably improved, accurate pressing of multiple parts can be completed at a time, the time consuming problem of layer-by-layer operation is avoided, it is ensured that all layers are evenly stressed through the structure, the adhesive layer thickness deviation or dislocation risk is reduced, meanwhile, the consistency of insulating performance is kept, and space is saved through vertical layout.

The present invention discloses a distance protection method applicable to a differential electronic voltage transformer, the method comprising the steps of: analyzing the circuit of the differential electronic voltage transformer and establishing a model; constructing a virtual digital electronic voltage transformer based on the model; using the virtual digital electronic voltage transformer to process the current sampling signal to obtain a current signal in the differential domain; using a low-pass filter to process the voltage signal and the current signal output by the differential electronic voltage transformer; using the voltage signal and the current signal processed by the low-pass filter to calculate the reactance, and comparing the calculated reactance with the distance protection reactance constant, so as to determine whether the distance protection is activated. The method is applicable to an electronic voltage transformer whose output is a differential signal, and can effectively avoid the influence of the integral loop on the protection accuracy, thereby improving the accuracy of the protection.

The present application belongs to the technical field of digital publication processing, and provides a multi-modal PDF content purification and intelligent proofreading system and method based on visual perception enhancement, which comprises a visual perception module, a semantic understanding module and an intelligent decision module; the visual perception module eliminates physical layer interference by generating multi-resolution images, extracting fine structure features and locating invisible characters; the semantic understanding module constructs a knowledge graph and a DST-Transformer model to realize character recognition, semantic coding and image-text association repair; the intelligent decision module outputs a purified document through weighted confidence calculation, dynamic threshold adjustment and cross-platform verification; the method corresponds to the three module steps, realizes full-process processing from physical layer analysis to logical layer reconstruction and then to intelligent output, improves the PDF content purification accuracy and cross-platform compatibility, and is suitable for electronic document deep purification and proofreading.

The present application relates to the special protection technology of electronic components, and proposes an ESD static surge protection structure and method, which comprises a plurality of I / O pins, a detection circuit, a transient voltage suppressor and a discharge channel. The I / O pin is connected with the detection circuit and is used for input signal. The detection circuit is used for analyzing ESD static event or surge event, and inputs the transient voltage suppressor and the discharge channel. The transient voltage suppressor comprises a side TSV path and a center TSV path. The detection circuit comprises a voltage sensor and a current sensor. The voltage sensor is connected with a first comparator and an amplitude detector respectively. The first comparator is used for monitoring the voltage slope of the input signal of the I / O pin. The current sensor is connected with a second comparator. The second comparator is used for monitoring the pulse energy and pulse duration of the input signal of the plurality of I / O pins. Through the multi-parameter dynamic analysis of the voltage slope, pulse width, voltage peak value and pulse energy, accurate event recognition and classification protection are realized.

The invention relates to the technical field of cable manufacturing, in particular to an intelligent tension regulation and control method and system in a cable twisting process, and aims to solve the problems of low regulation and control precision, response lag and weak interference resistance in the prior art. The method comprises the following steps: firstly, introducing a multi-wire tension coupling coefficient according to cable specifications, material flexural rigidity and the number of stranded layers, calibrating a tension reference value by combining real-time environment temperature and humidity, and synchronously extracting roughness and correcting friction parameters through optical characteristics of wire surfaces so as to adapt to equipment speed; training historical data by using an LSTM (Long Short Term Memory) time sequence model, predicting the tension trend in the future 2-5 seconds in real time, performing classified regulation and control according to deviation statistical characteristics, and performing micro-step rotation speed regulation by a servo motor and torque compensation execution by a magnetic powder brake; updating the model when the error exceeds the threshold value or the specification is changed, and periodically finely adjusting the reference value by combining the quality data. The system comprises seven functional modules, the tension control precision and the anti-interference capability can be improved, the adaptability is high, and the cable quality and the production efficiency are effectively improved.

The invention discloses a power electronic module and system for a switch and a detection method. In the module, an insulating substrate is arranged in a packaging shell; the plurality of power semiconductor chip units are connected in parallel to the insulating substrate; the three-dimensional suspension type non-contact sensing structure is arranged in the packaging shell, and the sensor carrier plate is installed above a single target power chip in the multiple power semiconductor chip units in a crossing mode in a three-dimensional suspension mode through the insulating supporting piece and does not occupy a main power wiring area of the insulating substrate in the vertical projection direction. The chip-level current sensor is integrally mounted on one side, facing the single target power chip, of the sensor carrier plate in a surface-mounted manner and directly faces a bonding wire arching area of the single target power chip, and the chip-level current sensor is configured to acquire branch current flowing through the single target power chip in a non-contact magnetic coupling manner and output an analog voltage signal; and the analog signal output end of the chip-level current sensor is directly connected to the hardware straight-through protection terminal through a low-impedance straight-through path.

The invention relates to an insurance industry customer information analysis method and device and a storage medium. According to the method, a scene set and a result set are established based on historical data analysis logs; performing primary preliminary matching on all scenes in the'scene set 'and input analysis requirements by applying a logistic regression model, and reserving the first n scenes with the highest matching degree; inputting the preliminary matching result into a deep matching model for further matching, and outputting k final matching scenes from high to low according to the matching probability; and after a final matching scene is obtained, k analysis results of a'result set 'are generated corresponding to the final matching scene by applying a pre-packaged data query statement based on the association between the result and the scene. The insurance industry customer information analysis method combining logistic regression and deep matching is used, analysis requirements are understood, a scene is selected from a scene set, a result is output from a result set, and both speed and accuracy are achieved.

The present invention relates to the field of automatic navigation technology, and specifically to a navigation scheme planning method, system, and storage medium, wherein the method comprises: obtaining the location information of a departure place and a destination; searching for the departure public transportation station closest to the departure place and the destination public transportation station closest to the destination according to the location information of the departure place and the destination respectively; respectively calculating the total driving path generated according to the location information of the departure place and the destination and calculating the corresponding total driving distance, and generating the initial driving path according to the location information of the departure place and the departure public transportation station, and calculating the corresponding initial distance; judging whether the difference between the total distance and the initial distance is less than a first predetermined value preset by a user; if it is less than the first predetermined value, selecting the total driving path as the navigation path. The navigation scheme of the present invention includes a mixed mode of public transportation and driving; it can take both cost and speed into consideration, and effectively improve the user experience of citizens.

The invention discloses a power system transient time domain simulation switching method and system for adaptively analyzing and calculating a scene, and a storage medium. The method comprises the following steps: constructing an electromagnetic transient simulation system consistent with an electromechanical transient simulation system; respectively adopting an electromagnetic transient simulation system and an electromechanical transient simulation system to synchronously carry out electromagnetic transient simulation and electromechanical transient simulation considering the electromagnetic transient process of the simulation object until a preset time period passes after the fault is recovered; according to the simulation results of the electromagnetic transient simulation and the electromechanical transient simulation which are synchronously carried out, calculating a characteristic quantity for identifying whether the electromechanical transient simulation result is credible or not; and inputting the characteristic quantity into an offline pre-generated classifier, if an output result of the classifier is credible, continuing to adopt the electromechanical transient simulation and terminating the electromagnetic transient simulation at the same time, otherwise, switching to the electromagnetic transient simulation, carrying out a subsequent simulation process, and terminating the electromechanical transient simulation at the same time. The scene can be adaptively analyzed and calculated, and simulation is adaptively switched according to conditions.

This application provides a training method, apparatus, device, and storage medium for a game intent recognition model. The method includes: acquiring a basic sample set; calling a pre-trained large language model to perform deep semantic parsing on the data in each basic sample in the basic sample set to obtain the full intent annotation information corresponding to each basic sample; concatenating multiple pre-generated instruction templates corresponding to different intent tasks with each basic sample to obtain multiple input samples corresponding to each intent task; decomposing the full intent annotation information corresponding to each basic sample to obtain the intent annotation information corresponding to each input sample; and performing supervised fine-tuning training of the pre-trained lightweight language model for different intent tasks based on each input sample and the intent annotation information corresponding to each input sample to obtain the trained game intent recognition model. This solution can improve the accuracy of the model's game intent recognition and enhance the model's generalization ability.

The present invention relates to the technical field of audio recognition, and specifically to a method for selecting from multiple sets of speech recognition results using confidence scores, comprising: collecting speech audio of a person in a car; sending the speech audio to an offline recognition engine and an online recognition engine respectively, and immediately sending the analysis result of each engine to a scoring model after receiving it, to obtain a corresponding confidence score; judging whether the confidence score first output by the scoring model is greater than or equal to a first preset threshold; if the confidence score first output is greater than or equal to the first preset threshold, outputting the analysis result corresponding to the confidence score first output as a recognition result; if the confidence score first output is less than the first preset threshold, waiting for the scoring model to output the confidence score corresponding to the analysis result of each engine. The present invention can solve the defect that the prior art cannot take into account both speed and accuracy when recognizing speech.

The invention discloses a ship navigation test simulation method and simulation system based on a scene. The method comprises the following steps: A1, constructing a dynamic grid; a2, texturing the grid data; step A3, sampling of RayMarching fluid parameters is carried out; step A4, fluid force calculation; step A5, interaction type detection; step A6, carrying out ship-shore interaction calculation; step A7, updating the motion state of the ship; a8, state output and iteration judgment are carried out; according to the method, rapid fluid load calculation is achieved by constructing a local space grid and utilizing three-dimensional texture and GPU programming, parameters are provided for a ship motion model, meanwhile, the ship-shore interaction is solved by utilizing an image depth technology, the ship-ship interaction is calculated by utilizing a three-dimensional texture fusion method, and the ship-shore interaction is calculated by utilizing a three-dimensional texture fusion method. The method realizes fast calculation of the ship motion state with low sacrifice precision and fast calculation speed, and can be applied to ship navigation test simulation and control system design.

The application discloses a cluster resource self-adaptive expansion and contraction method and system based on flow mutation awareness, and relates to the technical field of cloud computing. The method comprises the following steps: S1, collecting cluster expansion and contraction awareness data in real time, performing data preprocessing, and extracting flow change characteristics, response change characteristics and queuing change characteristics; S2, identifying a flow mutation state, analyzing historical flow mutation records or growth extrapolation results, evaluating a flow peak value predicted arrival time and a predicted peak value request amount, and performing boundary checking; S3, performing time limit evaluation on a newly added instance available preparation process, and generating an expansion opportunity marker; S4, judging expansion instance demand, generating an instance expansion instruction according to the expansion opportunity marker and the expansion instance demand, and adding the newly added instance to a traffic forwarding link corresponding to a target service. The method solves the problem that in a flow surge scenario, the actual available time of expanded resources does not match the flow peak value arrival time, resulting in expansion lag, service overload or resource idle waste.

The present application relates to a kind of spectral level calibration real-time color measurement system and method, it is related to color measurement technical field, system includes viewfinder lens group, shutter, light beam spectrometer module, multi-channel light filter detection unit, dispersion module, main control board and angle biasing device;Method includes calibration stage, acquisition stage and measurement stage.By obtaining monochromatic light image sequence and light intensity data, calculate sensor response curve, filter transmittance curve and dispersion module response curve;Using reference board to calculate camera correction coefficient;To be measured object acquisition image and apply correction coefficient, combined with dynamic calibration compensation light source fluctuation, finally output high-precision color value.The present application solves the problem of prior art by light source interference, bench difference, high cost and low efficiency, realizes the real-time color measurement of low cost, high efficiency, high precision and high resolution, it is suitable for the online detection of textile, electronic display and other industries.

The present invention proposes a foreground-enhanced end-to-end small target detection method for drone aerial images, comprising: inputting the image taken by the drone into a lightweight backbone network ShuffleNetV2 to obtain feature maps of different sizes; adding an FPN feature pyramid structure after the feature map to fuse multi-scale features to obtain a fused multi-scale feature map; inputting the multi-scale feature map into a single-stage detection head to obtain a rough detection result, and obtaining a foreground map from the rough detection result; passing a layer of convolution on the foreground map to obtain a foreground-enhanced multi-scale feature map; inputting the foreground-enhanced multi-scale feature map into the single-stage detection head again to obtain a fine detection frame; using a dynamic soft-weighted redundant detection frame suppression algorithm to eliminate redundant redundant frames of the same target in the fine detection frame to obtain the final detection result. The method proposed by the present invention has a good detection effect for drone aerial scenes with dense local targets and sparse global foregrounds.