514 results about "Target distribution" patented technology

The invention discloses an unmanned aerial vehicle inspection control system based on binocular vision, and relates to the technical field of unmanned aerial vehicle inspection, a binocular vision acquisition module is used for acquiring a to-be-inspected target image, detecting shielding and degree, and extracting a target feature and a distribution rule when an unmanned aerial vehicle inspects according to a preset route; the information acquisition module is used for acquiring a target position, peripheral complementary target distribution, unmanned aerial vehicle pose and kinematics parameters when it is detected that the shielding degree of the target point exceeds a set value. Complete coverage of the two types of target sub-regions can be ensured, the flight trajectory curvature of the second type of target sub-regions can be minimum, and stable flight is ensured.

The invention provides a panoramic image processing method and device suitable for an unmanned aerial vehicle and a panoramic unmanned aerial vehicle. The method comprises the following steps: acquiring multi-view images in real time through a multi-camera array of an unmanned aerial vehicle, acquiring pose data by using an inertial sensor, converting each view image into a first projection image and a second projection image, and performing dynamic target detection and recognition on the second projection image to obtain a dynamic target, geometric scene features and target distribution. Differentiated processing areas are divided, specifically, a high-precision algorithm is adopted for the first processing area, a low-precision simplified algorithm is adopted for the second processing area, and computing power focusing and processing capacity optimization are achieved. In the splicing stage, a dynamic target historical trajectory optimization splicing algorithm is combined, and processing strategies are adapted to different areas, so that the quality of the key areas is guaranteed, and the process of the non-key areas is simplified to reduce the calculation load. According to the scheme, the problem of real-time panoramic splicing under the condition that unmanned aerial vehicle hardware resources are limited is effectively solved, and the processing real-time performance and efficiency are remarkably improved.

The invention provides an unmanned aerial vehicle cluster dynamic hunting method for a complex three-dimensional scene. The method comprises the steps that the state of an escaper unmanned aerial vehicle, the state of each hunting unmanned aerial vehicle and the environment state are obtained respectively; sub-targets are generated around the escaper unmanned aerial vehicle through a Fibonacci ball algorithm, and the sub-targets are dynamically adjusted; based on the dynamically adjusted sub-targets and the state of each trapper unmanned aerial vehicle, optimal sub-target distribution is carried out on each trapper unmanned aerial vehicle through an improved market auction algorithm, wherein the improved market auction algorithm considers path cost and angle cost in a cost function; constructing a dynamic target control obstacle constraint based on the environment state, the state of the escaper unmanned aerial vehicle and the state of each trapper unmanned aerial vehicle; an escaper position change factor and a self-adaptive attenuation rate are introduced into the dynamic target control obstacle constraint; and constructing a distributed MPC optimization problem based on optimal sub-target distribution and a dynamic target control barrier function to perform trajectory optimization and motion control. According to the invention, the efficiency and the safety are ensured at the same time.

The invention discloses a data distribution method and device, electronic equipment and a storage medium, the data distribution method is applied to a distribution server deployed at a network side, and comprises the following steps: obtaining distribution characteristic statistical data of a target data packet to be sent by terminal equipment; inputting the distribution characteristic statistical data of the target data packet into a first data distribution decision network to obtain a first distribution confidence coefficient of the target data packet matched with each transmission path; inputting the distribution characteristic statistical data of the target data packet into a second data distribution decision network to obtain a second distribution confidence coefficient of the target data packet matched with each transmission path; and matching the first distribution confidence coefficient and the second distribution confidence coefficient of each transmission path based on the target data packet, and determining a target distribution strategy of the target data packet. By applying the technical scheme provided by the invention, the distribution strategy of the uplink data packet of the terminal equipment can be accurately and effectively determined, and the data transmission efficiency is improved.

The invention relates to the technical field of source network load storage cooperative control, in particular to an area source network load storage cooperative optimization control system based on an intelligent fusion terminal, and the system comprises an acquisition module which acquires energy characteristic parameters and cooperative response characteristic parameters of a target area in a historical period; analyzing an energy characteristic characterization value based on the energy characteristic parameter through an analysis module, and calculating a collaborative response characteristic characterization value based on the collaborative response characteristic parameter; determining a corresponding control strategy based on the energy feature representation value through a control module; through a verification module, whether the stability of the transformer area source network load storage meets the standard is judged based on the collaborative response characteristic representation value; and through a feedback module, based on a difference value between the energy feature representation value and a predetermined energy feature representation threshold, determining an adjustment amplitude of the cooperative response feature representation threshold, and performing collection and verification again. According to the invention, through the verification module and the feedback module, the response rate of cooperative control of each link of source network load storage is improved.

The invention discloses a fire emergency command system and method based on dynamic fire scene modeling and double-algorithm cooperation. The system comprises a fire scene intelligent map system, a fireman terminal, an unmanned fire fighting equipment terminal and external information. The method comprises the steps that the information management layer extracts the latitude and longitude of a fire point, calculates a spreading rate and a high-risk area mark and pushes the same to the display module; a commander issues a fire-fighting instruction to an AI auxiliary decision-making layer; a fire fighting task plan and a target fire point priority ranking table are generated by a target distribution module and a Hungary algorithm of the AI auxiliary decision-making layer; the AI auxiliary decision-making layer drives a fire extinguishing planning module to cooperate with an action path through an MADDPG algorithm, plans an unmanned aerial vehicle group airdrop route optimization and robot group breakthrough route, carries out data and instruction interaction to generate an end-to-end execution instruction set, and transmits the instruction set to the instruction layer; and the instruction layer sends the generated fire extinguishing instruction to each terminal to guide the firefighter and the unmanned equipment to extinguish the fire. According to the invention, full-chain optimization of fire rescue is realized through multi-dimensional technology fusion.

Disclosed are apparatuses, systems, and techniques that may use machine learning for implementing generative text-to-speech models. The techniques include identifying a mapping of speech characteristics (SC) on a target distribution of a latent variable using a non-linear transformation for at least a subset of the SC. Parameters of the non-linear transformation are determined using a neural network that approximates a statistics of the SC with a statistics predicted for the SC based on the identified mapping and the target distribution of the latent variable.

The invention provides an optimal configuration method and device for a magnetic reactance type dynamic voltage restorer, and relates to the technical field of optimal configuration, and the method specifically comprises the steps: obtaining the historical voltage sag event data of a target power distribution bus and the electrical parameters of a downstream load motor; screening a to-be-compensated event set based on the principle that the sag depth exceeds a load tolerance threshold and the duration is less than the critical stall time, and determining a compensation strategy according to a comparison result of a phase jump angle maximum value and the threshold; the maximum compensation voltage is determined according to the maximum sag depth, the rated output current is determined according to the load parameters, the theoretical energy required for compensating the longest event is calculated based on the compensation strategy, and the actually required energy is determined in combination with the energy correction coefficient of the device topology; and finally, generating an optimal configuration scheme based on the compensation strategy, the maximum compensation voltage, the rated output current and the actually required energy. According to the invention, accurate mapping from power grid measured data to device key parameters is realized.

The invention relates to a large model module level training strategy optimization method and device for a heterogeneous cluster and computer equipment, and the method comprises the steps: obtaining the calculation information of a large model module level and the communication information of an operator level for the heterogeneous cluster; the calculation information comprises calculation time information and module video memory information of modules under different data scales under different distributed strategies and on different chips; the communication information is communication time delays of communication operators under different data scales; determining an initial distributed training strategy of each assembly line based on the module video memory information; based on the calculation time information and the communication information, calculation time of different stages of the assembly line is determined; and optimizing each initial distributed training strategy according to the calculation time and a video memory threshold value for bearing the large model equipment to obtain a target distributed training strategy. Through the method and the device, the problem of relatively low resource utilization rate of a large model oriented to heterogeneous clusters is solved.

The invention discloses a semi-supervised three-dimensional target detection method and system based on dynamic three thresholds, and relates to the field of electrical digital data processing. Performing clustering calculation according to the confidence coefficient score and the consistency IoU score of the first prediction result to obtain three pairs of high, medium and low dynamic thresholds; inputting the unlabeled scene into a teacher model for reasoning to obtain a second prediction result, comparing and grading a confidence coefficient score and a consistency IoU score of the second prediction result with three pairs of dynamic thresholds to generate a high-level pseudo-label, a middle-level pseudo-label, a fuzzy-level pseudo-label and a low-level pseudo-label, and outputting the high-level pseudo-label, the middle-level pseudo-label, the fuzzy-level pseudo-label and the low-level pseudo-label. In the training process, adaptive target distribution and multi-level supervision are carried out, and different layers of pseudo labels adopt different supervision modes to supervise student model training. By implementing the method, the detection precision of the three-dimensional target detector under semi-supervised setting can be improved.

The invention discloses an automatic question setting method based on graph structure traversal and a semantic transformation model. The method comprises the following steps: establishing a normalized knowledge graph; establishing a question generation configuration parameter set and path difficulty target distribution; configuring an improved random walk algorithm control parameter set; extracting a candidate reasoning path set, and generating a path description record for each reasoning path in the candidate reasoning path set; obtaining a qualified reasoning path set, and keeping a binding relationship between the qualified reasoning path set and the corresponding path description record; and generating a difficulty level label and a knowledge point label. In the question generation stage, the system binds and stores key sampling parameters and the question stem, options and interference items of each question, and manageable question bank, controllable difficulty, extensible question types and batch high-quality generation of the automatic question setting system are realized according to multi-dimensional comprehensive scoring during quality evaluation.

The invention discloses a multi-mode emotion fusion robot voice style conversion method and equipment. The method comprises the following steps: extracting emotion feature vectors of text, audio and image input through a multi-mode emotion feature fusion extraction network; constructing a probability path taking Gaussian noise as a starting point and a target Mel spectrum as a terminal, taking the emotional characteristics as a condition guide vector, and predicting a vector field from the noise to target distribution by using a time condition U-Net; wasserstein-2 regularization constraint based on an optimal transmission theory is introduced to learn a smooth and efficient probability flow path; and finally, reconstructing a high-fidelity voice waveform with the target emotion through an acoustic decoder. According to the method, the problems of insufficient emotion expressive force and low generation efficiency in a traditional speech synthesis technology are solved, end-to-end optimization of emotion feature cross-modal mapping and speech generation is realized, and the emotion naturalness and the generation stability of synthesized speech are remarkably improved.

The invention belongs to the technical field of biological detection, and discloses a transgenic crop intelligent identification method based on high-throughput sequencing, which comprises the following steps: acquiring multi-platform high-throughput sequencing original data and sample metadata, evaluating quality by platforms, dividing according to a sample partitioning rule, and generating a quality-controlled data block set; distributing data blocks to a plurality of parallel computing nodes, synchronously executing technical detection to obtain technical fact fragments, aggregating and supplementing batch statistical information, and generating a technical fact report; matching the judgment rule set, carrying out parallel judgment, triggering expert rechecking on boundary cases, and integrating to generate a supervision judgment conclusion set; the method comprises the following steps: extracting original sequencing reads of all samples, generating a unique hash for each read, constructing a processing track chain, integrating metadata and visual evidence, generating a credible authentication report, dividing different scene versions, and performing targeted distribution; and extracting identification cases for mining and clustering, generating optimization suggestions, and feeding back the optimization suggestions to the quality control rule base and the judgment rule base.

The invention provides a diversified text sensitive data synthesis method for desensitization effect evaluation, which comprises the following steps: constructing a sensitive entity system meeting desensitization evaluation requirements, the sensitive entity system comprises a general field, a medical field and a financial field, and each field comprises a plurality of entity types; obtaining an original data set, counting entity distribution on the original data set, constructing a target distribution model, and designing a diversified strategy based on entity types and sentence patterns; guiding the large language model to generate candidate corpora according to the target distribution model and the diversification strategy, performing character-level alignment labeling and consistency verification on the candidate corpora, and generating a synthetic data set based on the candidate corpora; and performing multi-dimensional quality verification on the synthetic data set from the data layer, the entity layer and the semantic layer to obtain an evaluation result, and feeding back the evaluation result to a closed-loop controller to adjust a quota and a generation parameter so as to generate a final synthetic data set. The method can be applied to validity evaluation of a data desensitization tool, and the problems of single evaluation dimension, data sparsity and the like in text desensitization evaluation are solved.

The invention discloses a large language model liver cancer prediction method and system based on an improved sampling strategy, and relates to the technical field of artificial intelligence medical diagnosis, and the method comprises the steps: obtaining an electronic medical record of a patient, carrying out the sequential reconstruction, and constructing a structured cue word; a plurality of reasoning paths are generated in parallel by using a random decoding strategy according to a general large language model with frozen input parameters; constructing a target distribution model based on gamma distribution, and calculating the normalized weight of each path by adopting an importance sampling algorithm to suppress low-quality paths and amplify the weight conforming to medical logic paths; and finally, performing weighted aggregation on the diagnosis conclusion based on the weight to obtain a prediction result, and outputting the reasoning process with the highest weight as an interpretability report. The accuracy of liver cancer prediction and the clinical decision transparency can be remarkably improved without fine adjustment of the model.

In general, the current subject matter relates to layer 1 / layer 2 (L1 / L2) triggered mobility (LTM) for intra-base station centralized unit user plane (CU-UP) relocation. In some implementations, a target distributed unit (DU) of a base station to serve a user equipment (UE) can be determined to be served by a target CU-UP of the base station. A serving CU-UP of the base station can serve a serving DU of the base station that is currently serving the UE. Using a centralized unit control plane (CU-CP) of the base station, the target CU-UP can be prepared for layer 1 / layer 2 triggered mobility (LTM), and, using the CU-CP, the target DU can be prepared for LTM.

The invention discloses a multi-unmanned aerial vehicle cooperative search and coverage optimization method based on an improved Lloyd algorithm, and is suitable for a target search task in a dynamic environment. Firstly, target distribution is perceived and estimated in real time by using a probability hypothesis density filter, and a target probability map is dynamically updated, so that the response capability and search precision of a system to target state change are improved; secondly, a multi-graph fusion mechanism is adopted, a target probability map, an uncertainty map and a search pheromone map are subjected to dynamic weighted integration, priority sharing and updating of key area information are achieved, and the cooperation efficiency and the communication resource utilization rate among multiple unmanned aerial vehicles are improved; and finally, based on an improved Lloyd algorithm, introducing a multi-step prediction mechanism and a direction-guided centroid updating strategy, optimizing search path generation, and preventing the path from being converged to a low-value region. Through the method of the invention, the system can realize adaptive and efficient search and coverage control in a complex dynamic environment, and stability, real-time performance and overall robustness of task execution of multiple unmanned aerial vehicles are significantly improved.

The invention provides a distributed database data migration method which can be applied to the technical field of big data and the field of distributed technologies. The distributed database data migration method comprises the steps that a data migration plan is obtained, wherein the data migration plan comprises distribution information that each source data fragment in a source distributed database cluster is distributed to a target node of a target distributed database cluster and parallel migration degree representing the number of source data fragments of concurrent migration; executing the data migration plan, including: controlling m source data fragments to migrate in parallel, m being a positive integer and m not greater than the parallel migration degree; and migrating each source data fragment to the allocated target node based on the allocation information, and obtaining a target data fragment corresponding to the source data fragment at the target node after migration is completed. The invention further provides a distributed database data migration device and equipment, a storage medium and a program product.

A method may include inputting, into a draft model, a sequence of tokens conditioned on a given first context to obtain an original draft distribution of draft tokens. The method may also include applying, based on a watermark code, a reweight function to adjust probabilities of the original draft distribution to generate a watermarked draft distribution of watermarked draft tokens. The method may further include inputting, into a target model, watermarked draft tokens and a sequence of tokens conditioned on a given second context to obtain an original target distribution. In addition, the method may include applying, based on the watermark code, the reweight function to probabilities of the original target distribution to generate a watermarked target distribution of watermarked target tokens. The method may further include sampling a watermarked output from the watermarked draft distribution or the watermarked target distribution based on the adjusted probabilities.

The invention discloses a video frame sharing output processing method and a related device, and relates to the technical field of software, and the method comprises the steps: obtaining an original video frame collected by a camera; the original video frame is converted into a first target video frame according to the format configuration of the stream pushing end, the original video frame is converted into a second target video frame according to the format configuration of the virtual camera, and the stream pushing end and the virtual camera are pre-created with respective shared memories; and writing the first target video frame into a shared memory of the stream pushing end, and writing the second target video frame into a shared memory of the virtual camera. According to the method, the architecture design of single-source processing and multi-target distribution can be realized, repeated calculation is avoided, in addition, cross-process access can be supported, and the effect of isolated storage of double shared memories is achieved.

The application relates to the technical field of time synchronization, and discloses a mixed time service method based on Beidou satellite navigation and TSN, which comprises the following steps: acquiring network topology structure information and historical time service error data of each synchronization node in a target distributed system, and constructing a dynamic time delay state set for jitter identification; based on the dynamic time delay state set, a current global time reference is extracted through a time calibration algorithm, a time difference reference vector sequence is generated, and initial synchronization adjustment is carried out on a TSN master node; according to the response stability of a slave node, a clock convergence function curve between nodes is constructed, whether the synchronization network is in a low jitter interval is judged based on time offset rate and time drift trend change; in combination with the node time state after directional offset suppression, a multi-channel redundant time adjustment mechanism in a time window is applied, a weight adjustment coefficient is extracted, and a non-uniform calibration factor is introduced; and according to a regression correction result, a time service stability index is evaluated in real time. The application has the advantages of improving time service stability.

The invention discloses an optimal distribution method for electric vehicle charging stations, and the method comprises the steps: constructing a state transition matrix based on the travel time of a driver of a to-be-charged electric vehicle, the charging cost, the number of available charging piles of the charging stations, the comprehensive position of the charging stations, and the attraction of the charging stations, and generating charging demand probability distribution in real time; calculating a reachable radius according to the residual electric quantity and the energy consumption rate of the to-be-charged electric vehicle, screening a candidate charging station set, fusing a driver preference weight, and establishing a multi-target distribution cost evaluation mechanism; constructing an optimal distribution model of the electric vehicle charging stations; an initial population is generated through greedy random adaptive search, a multi-parent cross strategy or a directional exchange strategy is dynamically executed according to the remaining time of a driver, and a corresponding optimal charging station distribution scheme is output. According to the method, the optimal charging station is allocated for the electric vehicle user, and the driving distance cost, the charging currency cost and the destination walking distance cost are reduced to the maximum extent.

The invention provides a terminal redirection method and device, and the method comprises the steps: determining a target distributed network for a to-be-migrated terminal in a current distributed network under the condition that the current distributed network meets an internal terminal redirection condition; acquiring base station information, and determining a plurality of candidate service paths according to the base station information, the to-be-migrated terminal and the target distributed network; determining a target service path by taking the optimal comprehensive performance parameter of the road section as a target; and generating a service strategy based on the target service path, and issuing the service strategy to the to-be-migrated terminal, the to-be-migrated terminal being used for redirecting to the target distributed network according to the service strategy. Therefore, the automatic redirection of the terminal among the distributed networks is realized by dynamically evaluating the network state and intelligently selecting the optimal migration path, and the network load balancing efficiency and the terminal service quality can be improved while the energy consumption is reduced.

Certain aspects of the present disclosure provide techniques and apparatus for generating a response to query input in a generative artificial intelligence model. An example method generally includes receiving a plurality of token sets generated based on an input cue and a first generative artificial intelligence model, each token set of the plurality of token sets corresponding to a candidate response to the input cue; selecting a set of tokens from the plurality of sets of tokens using a second generative artificial intelligence model and a recursive adjustment to the target distribution associated with the received plurality of sets of tokens; and outputting the selected set of tokens as a response to the input cue.

The embodiment of the invention provides an unmanned aerial vehicle distribution path planning method, device and equipment, and the method comprises the steps: constructing a plurality of distribution paths according to a distribution region of an unmanned aerial vehicle and distribution points included in the distribution region; obtaining topographic data, coordinate information of a plurality of distribution points, cargo weight information of the plurality of distribution points, initial environment data and unmanned aerial vehicle initial state data; according to the coordinate information of each distribution point and the cargo weight information of each distribution point, screening an initial distribution path from the distribution paths based on a load near-point priority strategy; determining cost information of each initial distribution path according to an arrangement sequence of distribution points included in each screened initial distribution path, topographic data, coordinate information of each distribution point, cargo weight information of each distribution point, initial environment data and unmanned aerial vehicle initial state data; and determining a target distribution path according to the cost information of each initial distribution path. The optimal distribution efficiency of the unmanned aerial vehicle under the complex road condition is ensured, the energy consumption of the unmanned aerial vehicle is lowest, and the distribution is safer.

The invention discloses a heterogeneous spacecraft cluster target allocation method based on a multi-agent reinforcement learning algorithm, and mainly relates to a heterogeneous weapon system model design and a hierarchical hybrid collaborative decision framework design based on multi-agent reinforcement learning. According to the method, the problem of heterogeneous target allocation of the spacecraft cluster in the technical field of spacecraft cooperative control is solved. The spacecraft cluster heterogeneous target allocation method based on multi-agent reinforcement learning comprises the following steps: 1, establishing a heterogeneous weapon target allocation model for a spacecraft cluster heterogeneous target allocation problem; 2, constructing a'global-local-monomer 'hierarchical decision model based on a multi-agent depth deterministic policy gradient (MADDPG) algorithm framework, wherein a global task allocation layer agent adopts centralized training to learn a heterogeneous weapon task allocation and cluster hierarchical strategy; and intelligent agents of the local collaboration layer and the single execution layer adopt distributed training and local observation to carry out collaborative decision making and target distribution. And 3, designing a composite reward function fusing the target value, the intelligent agent capability matching degree and multiple constraints to guide the intelligent agent to maximize the overall task income and generate an optimal allocation strategy on the premise of meeting the complex constraints, thereby effectively improving the heterogeneous target allocation efficiency under the spacecraft large-scale cluster. The method can be applied to the aerospace field and in-orbit autonomous decision-making scenes.

The invention discloses a hierarchical scanning perception fusion super-resolution enhancement target detection method and system, and belongs to the technical field of image processing, and the method comprises the steps: carrying out the target detection of a down-sampled input image, and generating a global coarse detection result; generating a saliency map based on the input image, dividing the input image into a plurality of image blocks, calculating a normalized priority probability according to the saliency map and the position information of each image block, and generating a scanning sequence; inputting the image blocks into a sequence encoder according to the sequence of the scanning sequence, and processing the image blocks to obtain an enhanced feature map; and on the basis of the enhanced feature map, a self-adaptive threshold strategy is adopted to identify key image blocks with densely distributed targets, super-resolution reconstruction is performed on the key image blocks, then target detection is performed to obtain a local fine detection result, and the local fine detection result is fused with a global coarse detection result to output a final detection result. According to the method, super-resolution reconstruction is only performed on the key blocks which are densely distributed locally, so that the calculation overhead is reduced while the detection precision is ensured.

The invention relates to the technical field of artificial intelligence, in particular to a data prediction method and device, electronic equipment and a storage medium, which are used for improving the accuracy of contribution degree estimation. The method comprises the following steps: constructing a comprehensive feature according to respective attributes and association of an object to be predicted and target content; obtaining a first prediction result, a second prediction result and initial distribution parameters of the object under various statistical distributions according to the comprehensive features; the first prediction result represents the possibility that the object generates a resource interaction behavior on the target content in the future; the second prediction result is used for indicating whether the future resource interaction behavior of the object reaches a preset contribution standard or not; each initial distribution parameter represents the potential state of the contribution degree estimated value of the object under the corresponding statistical distribution; for each statistical distribution, adjusting the initial distribution parameter based on the second prediction result to obtain a target distribution parameter; and determining a contribution degree estimated value according to the first prediction result and each target distribution parameter.

The invention provides an air conditioning unit distribution box system diagram generation method, system and device and a storage medium, and relates to the technical field of electrical design, and the method comprises the steps: creating a power distribution attribute block, a device parameter template and a model selection database of electrical elements of each functional section device, and a system diagram framework style of a distribution box of an air conditioning unit; obtaining a device parameter table of each function section device, and performing device parameter information proofreading with the device parameter template to obtain target device parameter information; assigning values to the power distribution attribute blocks according to the target equipment parameter information and the model selection database to obtain target power distribution attribute blocks; associating the target power distribution attribute block with the power distribution box according to the target equipment parameter information to obtain associated functional section equipment; and calculating current information according to the load information of the associated function section equipment, and generating a distribution box system diagram based on the current information, the target distribution attribute block and the system diagram framework style. The design efficiency and accuracy of the air conditioning unit distribution box system can be improved.

The invention relates to the technical field of image sensing, general image recognition and feature positioning under computer vision, and discloses an acupuncture point positioning method and system based on artificial intelligence, and the method comprises the steps: obtaining a non-rigid target section image through an image collection module; the key point identification module determines topological anchor points at two ends of the section; the reference construction module establishes a structure reference vector; a deflection calculation module extracts surface texture feature points, obtains a high-frequency component and a low-frequency component through multi-scale frequency domain decomposition, and determines a spatial deflection parameter based on the gradient of the energy distribution ratio of the high-frequency component and the low-frequency component; the proportion compensation module corrects a target distribution proportion coefficient; according to the method, the pose is inversed by using the scale invariance between the frequency domain components, the decoupling of the anatomical positioning logic and the pixel space is realized, and the non-linear interference of the non-uniformity of the biological surface and the dynamic environment illumination on the positioning precision is eliminated.