Brain-inspired computing driven visual information instant analysis method and system

Abstract

The application relates to the field of artificial intelligence, and discloses a brain-computer computing driven visual information instant analysis method and system, which comprises the following steps: acquiring real-time sensing data and neural pulse signals, analyzing a signal sensing curve to construct a brain-computer computing network, identifying visual information features through the network, analyzing a preset model computing architecture, optimizing a neuron mapping relationship to obtain an optimized mapping network, analyzing a path based on the analysis information, determining a target data time sequence distribution state, calculating an analysis efficiency value, determining an analysis optimization direction, generating an instant analysis strategy, calculating an analysis attenuation rate, determining an analysis state based on the attenuation rate, analyzing an analysis dimension, and finally constructing an optimized analysis scheme. The application can improve the intelligent analysis efficiency of visual information.

Description

Technical Field

[0001] The present invention relates to a method and system for real-time analysis of visual information driven by brain-like computing, and belongs to the field of artificial intelligence. Background Art

[0002] In today's digital age, visual information processing plays a crucial role in numerous fields, such as security monitoring, autonomous driving, medical image analysis, and intelligent robotics. The rapid development of information technology has placed higher demands on the speed and accuracy of visual information processing.

[0003] Traditional visual information analysis methods are usually based on fixed algorithm models and limited computing resources, which makes them show obvious limitations when facing complex and changing visual scenes. On the one hand, these methods often find it difficult to process massive amounts of visual data in real time and efficiently, resulting in large delays in information analysis and unable to meet application scenarios with extremely high requirements for immediacy. For example, in autonomous driving, vehicles need to make accurate judgments on various visual information on the road ahead in a very short time, and the delay can easily cause serious safety problems; on the other hand, traditional methods lack deep understanding and intelligent analysis capabilities of visual information and cannot adaptively respond to the uncertainty and ambiguity in the scene. For example, in complex security monitoring environments, facing light changes, occlusions and the diversity of target objects, it is difficult to accurately identify and analyze targets, resulting in a high misjudgment rate. Therefore, there is a need for a real-time visual information analysis method driven by brain-like computing that can draw on the brain's efficient mechanism for processing visual information to improve the efficiency of intelligent analysis of visual information. Summary of the Invention

[0004] The present invention provides a method and system for real-time analysis of visual information driven by brain-like computing, the main purpose of which is to improve the efficiency of intelligent analysis of visual information.

[0005] To achieve the above objectives, the present invention provides a method for real-time visual information analysis driven by brain-like computing, comprising:

[0006] Acquiring real-time perception data corresponding to target visual information, collecting neural pulse signals corresponding to the real-time perception data, analyzing signal perception curves corresponding to the neural pulse signals, and constructing a brain-like computing network corresponding to the target visual information based on the signal perception curves;

[0007] Based on the brain-inspired computing network, visual information features in the target visual information are identified; based on the visual information features, a computing architecture in a preset brain-inspired computing model is analyzed; a neuron mapping relationship in the computing architecture is queried; and the neuron mapping relationship is optimized to obtain an optimized mapping network;

[0008] Analyzing an information parsing path corresponding to the target visual information based on the optimized mapping network, determining a temporal distribution state of target data in the target visual information based on the information parsing path, and calculating a parsing efficiency value corresponding to the target visual information based on the temporal distribution state;

[0009] Determining a parsing optimization direction corresponding to the target visual information based on the parsing efficiency value, generating a real-time parsing strategy corresponding to the target visual information based on the parsing optimization direction, extracting key execution parameters from the parsing strategy, and calculating a parsing attenuation rate corresponding to the target visual information based on the key execution parameters;

[0010] Based on the analytical attenuation rate, the analytical state corresponding to the target visual information is determined, the analytical dimension of the target visual information corresponding to the analytical state is analyzed, and based on the analytical dimension, an optimized analytical solution corresponding to the target visual information is constructed.

[0011] Optionally, constructing a brain-inspired computing network corresponding to the target visual information according to the signal perception curve includes:

[0012] Extracting perception feature parameters corresponding to the signal perception curve;

[0013] Determining a feature extraction mode corresponding to the signal perception curve based on the perception feature parameter;

[0014] Analyzing the feature distribution pattern corresponding to the target visual information based on the feature extraction mode;

[0015] Formulate a neural node connection rule corresponding to the target visual information according to the feature distribution rule;

[0016] Based on the neural node connection rules, a brain-like computing network corresponding to the target visual information is constructed.

[0017] Optionally, identifying visual information features in the target visual information based on the brain-inspired computing network includes:

[0018] Querying the activation state of neurons in the brain-inspired computing network;

[0019] collecting a state response feature set corresponding to the neuron activation state;

[0020] Extracting characteristic high-frequency signals from the state response feature set;

[0021] Generating a visual signal label corresponding to the characteristic high-frequency signal;

[0022] Based on the visual signal label, visual information features in the target visual information are identified.

[0023] Optionally, optimizing the neuron mapping relationship to obtain an optimized mapping network includes:

[0024] Querying mapping feature parameters in the neuron mapping relationship;

[0025] Determining an optimization target corresponding to the neuron mapping relationship based on the mapping feature parameters;

[0026] Based on the optimization goal, analyzing the optimization constraints corresponding to the neuron mapping relationship;

[0027] Formulate an optimization strategy corresponding to the neuron mapping relationship according to the optimization constraints;

[0028] Based on the optimization strategy, the neuron mapping relationship is optimized to obtain an optimized mapping network.

[0029] Optionally, analyzing the information parsing path corresponding to the target visual information based on the optimized mapping network includes:

[0030] Querying the node connection relationship in the optimized mapping network;

[0031] Determining an information parsing level corresponding to the target visual information based on the node connection relationship;

[0032] Locating key parsing nodes in the information parsing hierarchy;

[0033] Collecting a node parsing sequence corresponding to the key parsing node;

[0034] According to the node parsing sequence, an information parsing path corresponding to the target visual information is analyzed.

[0035] Optionally, calculating the parsing efficiency value corresponding to the target visual information according to the temporal distribution state includes:

[0036] The analytical efficiency value corresponding to the target visual information is calculated using the following formula:

[0037] ;

[0038] in, represents the analytical efficiency value corresponding to the target visual information, represents the total number of time intervals divided in the time series distribution state, The index of the number of time intervals, Indicates in The effective visual information successfully parsed within the time interval, Indicates the end time of parsing corresponding to the target visual information, Indicates the parsing start time corresponding to the target visual information, represents the adjustment coefficient, represents the average complexity corresponding to the target visual information, Indicates the reference complexity corresponding to the target visual information.

[0039] Optionally, determining the parsing optimization direction corresponding to the target visual information based on the parsing efficiency value includes:

[0040] Performing data normalization processing on the analytical efficiency value to obtain a normalized efficiency value;

[0041] Based on the normalized efficiency value, constructing an efficiency distribution map corresponding to the target visual information;

[0042] Setting an efficiency threshold corresponding to the efficiency distribution graph;

[0043] Based on the efficiency threshold, locating an inefficient parsing area in the target visual information;

[0044] According to the inefficient parsing area, a parsing optimization direction corresponding to the target visual information is determined.

[0045] Optionally, the calculating, based on the key execution parameter, a resolution attenuation rate corresponding to the target visual information includes:

[0046] The analytical attenuation rate corresponding to the target visual information is calculated using the following formula:

[0047] ;

[0048] in, represents the analytical attenuation rate corresponding to the target visual information, Indicates the total number of parameters corresponding to the key execution parameters, Indicates the quantity index corresponding to the key execution parameter, Indicates the The actual value of the parameter corresponding to each key execution parameter, Indicates the The ideal value of the key execution parameters corresponding to the parameters, Indicates the The reference standard values ​​corresponding to the key execution parameters, and Respectively represent the start time and end time corresponding to the time function, Represents the time function corresponding to the target visual information.

[0049] Optionally, determining the resolution state corresponding to the target visual information based on the resolution attenuation rate includes:

[0050] Determining a decay time axis corresponding to the analytical decay rate;

[0051] Extracting key decay nodes in the decay timeline;

[0052] Based on the key attenuation nodes, constructing an attenuation distribution matrix corresponding to the analytical attenuation rate;

[0053] querying correlation features between different attenuation points in the attenuation distribution matrix;

[0054] Based on the associated features, a parsing state corresponding to the target visual information is determined.

[0055] In order to solve the above problems, the present invention further provides a visual information instant analysis system driven by brain-like computing, the system comprising:

[0056] a network construction module for acquiring real-time perception data corresponding to target visual information, collecting neural pulse signals corresponding to the real-time perception data, analyzing signal perception curves corresponding to the neural pulse signals, and constructing a brain-like computing network corresponding to the target visual information based on the signal perception curves;

[0057] a relationship optimization module for identifying visual information features in the target visual information based on the brain-inspired computing network, analyzing a computing architecture in a preset brain-inspired computing model based on the visual information features, querying neuron mapping relationships in the computing architecture, and performing relationship optimization on the neuron mapping relationships to obtain an optimized mapping network;

[0058] an efficiency value calculation module, configured to analyze, based on the optimized mapping network, an information parsing path corresponding to the target visual information, determine, based on the information parsing path, a temporal distribution state of target data in the target visual information, and calculate, based on the temporal distribution state, a parsing efficiency value corresponding to the target visual information;

[0059] an attenuation value calculation module, configured to determine, based on the parsing efficiency value, a parsing optimization direction corresponding to the target visual information, generate, based on the parsing optimization direction, a real-time parsing strategy corresponding to the target visual information, extract key execution parameters from the parsing strategy, and calculate, based on the key execution parameters, a parsing attenuation rate corresponding to the target visual information;

[0060] A solution construction module is used to determine the analysis state corresponding to the target visual information based on the analysis attenuation rate, analyze the analysis dimension of the target visual information corresponding to the analysis state, and construct an optimized analysis solution corresponding to the target visual information based on the analysis dimension.

[0061] Compared with the problems described in the background technology, the present invention obtains real-time perception data corresponding to the target visual information and collects neural pulse signals corresponding to the real-time perception data, which can help to accurately identify the characteristics of visual information and optimize the neuron mapping relationship in the computing architecture. At the same time, it provides source data for a series of key operations such as subsequent analysis of information parsing paths and determination of parsing efficiency values, thereby greatly improving the accuracy of instant parsing of visual information. The present invention is based on the brain-like computing network to identify the visual information characteristics in the target visual information, effectively deal with complex situations such as light changes and occlusions, and accurately identify the contours, textures, dynamics and other features of the target object. This recognition method greatly improves the depth and accuracy of the understanding of visual information, and provides a solid foundation for subsequent analysis and decision-making of visual information. Furthermore, the present invention analyzes the information parsing path corresponding to the target visual information based on the optimized mapping network, which is It helps to accurately locate the key nodes and processes of information processing, quickly identify possible analysis bottlenecks, further optimize the network structure and parameter settings, significantly improve the overall efficiency and accuracy of visual information analysis, and enable the system to more efficiently cope with complex visual scenes. Furthermore, the present invention determines the analysis optimization direction corresponding to the target visual information based on the analysis efficiency value, accurately finds the key factors affecting efficiency, effectively adjusts the brain-like computing network architecture, optimizes the algorithm process, and thus efficiently improves the visual information analysis capability to better meet the needs of practical application scenarios such as security and autonomous driving. Finally, the present invention determines the analysis state corresponding to the target visual information based on the analysis attenuation rate, which helps to optimize the analysis strategy in a targeted manner, improve the processing capability of complex visual information, ensure the accuracy and immediacy of visual information processing in fields such as security monitoring and autonomous driving, and reduce the risk of misjudgment. Therefore, the brain-like computing-driven visual information instant analysis method and system provided in the embodiments of the present invention can improve the efficiency of intelligent analysis of visual information. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 A schematic flow chart of a method for instant visual information analysis driven by brain-inspired computing provided by one embodiment of the present invention;

[0063] Figure 2 A schematic diagram of the modules for implementing the brain-inspired computing-driven real-time visual information analysis system provided in one embodiment of the present invention.

[0064] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0065] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0066] The embodiment of the present application provides a method for instant analysis of visual information driven by brain-like computing. The execution subject of the method for instant analysis of visual information driven by brain-like computing includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided by the embodiment of the present application. In other words, the method for instant analysis of visual information driven by brain-like computing can be executed by software or hardware installed on a terminal device or a server device. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0067] Example 1:

[0068] Reference Figure 1 FIG2 is a flow chart of a method for real-time visual information analysis driven by brain-like computing according to an embodiment of the present invention. In this embodiment, the method for real-time visual information analysis driven by brain-like computing includes:

[0069] S1. Acquire real-time perception data corresponding to target visual information, collect neural pulse signals corresponding to the real-time perception data, analyze the signal perception curve corresponding to the neural pulse signal, and construct a brain-like computing network corresponding to the target visual information based on the signal perception curve.

[0070] By acquiring real-time perception data corresponding to target visual information and collecting neural pulse signals corresponding to the real-time perception data, the present invention can help accurately identify visual information features and optimize the neuron mapping relationship in the computing architecture. At the same time, it provides source data for a series of key operations such as subsequent analysis of information parsing paths and determination of parsing efficiency values, greatly improving the accuracy of instant parsing of visual information.

[0071] Among them, the target visual information refers to the specific visual content that needs to be processed and analyzed in specific application scenarios (such as the behavior of suspicious persons in security monitoring, road conditions in autonomous driving, human organ images in medical image analysis, etc.). It covers various visual elements such as the shape, position relationship, and dynamic changes of objects in the scene, and is the core processing object of the entire visual information instant analysis process; the real-time perception data refers to the raw data related to the target visual information obtained at the current moment through various sensors (such as cameras, image acquisition equipment, etc.). These data exist in the form of digital signals, etc., and faithfully record the physical characteristics of the target visual scene at a certain moment, such as light intensity, color distribution, and object contours; the neural pulse signal refers to the target visual information that is perceived in a simulated brain visual processing mechanism. During the process, the electrical signals generated when biological neurons transmit information and collected by relevant equipment are in the form of pulses. Their frequency, interval and other characteristics carry the key information of the target visual information after preliminary processing and are adapted to the brain processing mode. Optionally, the acquisition of real-time perception data corresponding to the target visual information can be achieved through sensor acquisition methods, such as: through visual sensors deployed in the target scene (such as cameras, lidars, infrared sensors, etc.) to capture visual information in real time, and finally obtain real-time perception data containing the target scene; the acquisition of neural pulse signals corresponding to the real-time perception data can be achieved through brain-like sensors, such as: using brain-like sensors (such as neuromorphic cameras, event cameras, etc.) to simulate the pulse emission mechanism of biological neurons, and convert real-time perception data into neural pulse signals.

[0072] Furthermore, the present invention can intuitively present the dynamic change characteristics of the neural pulse signal by analyzing the signal perception curve corresponding to the neural pulse signal, which helps to deeply understand the transmission mode of visual information under the simulated brain mechanism and provide an accurate basis for building a brain-like computing network that fits the brain processing logic.

[0073] Among them, the signal perception curve refers to a continuously changing curve drawn with time as the horizontal axis and the amplitude, frequency or other key features of the neural pulse signal as the vertical axis. It intuitively and dynamically reflects the ups and downs of the neural pulse signal in the process of receiving and processing target visual information. For example, the crest of the curve can correspond to the strong neural pulse response caused by the high contrast area or the significantly changed part in the visual scene, and the trough represents the signal state of the relatively stable or low stimulation area. Optionally, the analysis of the signal perception curve corresponding to the neural pulse signal can be achieved through a signal processing algorithm, such as: using a signal processing algorithm (such as Fourier transform, wavelet transform, etc.) to perform frequency domain or time domain analysis on the neural pulse signal, extract signal features, and finally generate a signal perception curve.

[0074] Furthermore, the present invention constructs a brain-like computing network corresponding to the target visual information based on the signal perception curve, which can simulate the way the brain processes visual information, greatly improve processing efficiency, enhance the system's adaptability to the uncertainty and ambiguity of visual information, reduce analysis delays and misjudgments, and meet application scenarios with high immediacy and accuracy requirements such as security and autonomous driving.

[0075] Among them, the brain-like computing network refers to a computing network constructed by simulating the brain's neural structure and information processing methods. It is composed of a large number of neuron-like neural nodes and connections established according to the neural node connection rules. By simulating the brain's perception, processing and cognitive process of visual information, the brain-like computing network can more effectively handle complex and changing visual scenes.

[0076] As an embodiment of the present invention, constructing a brain-like computing network corresponding to the target visual information based on the signal perception curve includes: extracting perception feature parameters corresponding to the signal perception curve; determining a feature extraction mode corresponding to the signal perception curve based on the perception feature parameters; analyzing a feature distribution law corresponding to the target visual information based on the feature extraction mode; formulating a neural node connection rule corresponding to the target visual information based on the feature distribution rule; and constructing a brain-like computing network corresponding to the target visual information based on the neural node connection rule.

[0077] Among them, the perceptual feature parameters refer to the key values ​​extracted from the signal perception curve that can quantitatively describe the curve characteristics. These parameters may include the peak value, valley value, period, slope change rate, etc. of the curve, which reflect the characteristics of the neural pulse signal in terms of intensity, frequency and change trend; the feature extraction mode refers to the specific method for extracting valuable features from the target visual information determined based on the perceptual feature parameters. For example, according to the characteristics of the perceptual feature parameters, a mode based on frequency domain analysis, time domain analysis or a combination of the two can be used to extract edge features, texture features or motion features in the visual information; the feature distribution law refers to the target visual information analyzed by the feature extraction mode. After processing, the distribution of visual features in spatial or temporal dimensions is presented. For example, in an image, edge features of different objects can be concentrated in certain specific areas, and texture features can be distributed according to a certain pattern; in a video, the motion features of an object can present a specific trajectory and speed change pattern over time; the neural node connection rule refers to a criterion formulated based on the feature distribution law to determine how the neural nodes in the brain-like computing network are connected. For example, if certain visual features are found to be closely related in space in the feature distribution law, then the corresponding neural nodes will establish stronger connections according to certain rules in order to better transmit and process these related features in the network.

[0078] Furthermore, the extraction of perceptual feature parameters corresponding to the signal perception curve can be achieved through signal processing methods, such as: extracting signal frequency domain and time domain features through Fourier transform, wavelet transform and other algorithms, thereby obtaining perceptual feature parameters; the determination of the feature extraction pattern corresponding to the signal perception curve can be achieved through machine learning methods, such as: classifying and selecting features through support vector machine (SVM), random forest and other algorithms, thereby obtaining a feature extraction pattern; the analysis of the feature distribution law corresponding to the target visual information can be achieved through statistical analysis tools, such as: performing distribution analysis of features through K-means clustering, principal component analysis (PCA) and other algorithms, thereby obtaining a feature distribution law; the formulation of neural node connection rules corresponding to the target visual information can be achieved through neural network design methods, such as: designing the connection weights between neural nodes through Hebbian learning rules, back propagation algorithms, etc., thereby obtaining neural node connection rules; the construction of a brain-like computing network corresponding to the target visual information can be achieved through a deep learning framework, such as: constructing and training a neural network model through frameworks such as TensorFlow and PyTorch, thereby obtaining a brain-like computing network.

[0079] S2. Based on the brain-like computing network, identify the visual information features in the target visual information, analyze the computing architecture in the preset brain-like computing model based on the visual information features, query the neuron mapping relationship in the computing architecture, optimize the neuron mapping relationship, and obtain an optimized mapping network.

[0080] Based on the brain-like computing network, the present invention identifies the visual information features in the target visual information, effectively copes with complex situations such as light changes and occlusions, and accurately identifies the contours, textures, dynamics and other features of the target object. This recognition method greatly improves the depth and accuracy of understanding visual information, and provides a solid foundation for subsequent visual information analysis and decision-making.

[0081] Among them, the visual signal label refers to a symbol or code generated based on a characteristic high-frequency signal and used to identify and classify specific visual information. For example, a visual signal label can represent specific visual scenes such as "pedestrians running fast" and "vehicles turning suddenly"; the visual information feature refers to the representative and recognizable attributes or characteristics in the target visual information. These features are a high degree of generalization and abstraction of the visual scene. By identifying and analyzing their features, the system can understand and interpret the observed visual content, and is the core element for realizing visual information processing, analysis and decision-making.

[0082] As an embodiment of the present application, the identifying the visual information feature in the target visual information based on the brain-like computing network comprises: querying the neuron activation state in the brain-like computing network; collecting the state response feature set corresponding to the neuron activation state; extracting the feature high-frequency signal in the state response feature set; generating the visual signal label corresponding to the feature high-frequency signal; and identifying the visual information feature in the target visual information based on the visual signal label.

[0083] The neuron activation state refers to a working state of a neuron in a brain-like computing network after receiving an input signal, which can be reflected by indicators such as membrane potential change and nerve impulse frequency, and reflects the response degree of the neuron to the target visual information. For example, when processing an image, the neurons corresponding to the edge region of the image can present a higher activation state. The state response feature set refers to a set of a series of related features caused by the neuron activation state, which covers various signal characteristics generated by the neuron activation, including time sequence of activation, duration of activation, and change of activation intensity, and each dimension of the features reflects the response mode of the neuron to the target visual information from different angles. The feature high-frequency signal refers to a signal component with relatively high frequency in the state response feature set. These high-frequency signals usually correspond to the parts of the target visual information that change dramatically and are more prominent, such as rapid movement of objects in the image and sharp changes in edges.

[0084] Further, the querying the neuron activation state in the brain-like computing network can be realized by a neural network monitoring tool, such as TensorBoard, Netron, etc., to monitor the activation state of the neuron in real time, thereby obtaining the neuron activation state. The collecting the state response feature set corresponding to the neuron activation state can be realized by a data collection algorithm, such as sliding window sampling, time series sampling, etc., to collect feature data in the neuron activation state, thereby obtaining the state response feature set. The extracting the feature high-frequency signal in the state response feature set can be realized by a signal processing algorithm, such as fast Fourier transform (FFT), wavelet transform, etc., to extract high-frequency components in the feature set, thereby obtaining the feature high-frequency signal. The generating the visual signal label corresponding to the feature high-frequency signal can be realized by a label generation tool, such as K-means clustering, Gaussian mixture model, etc., to classify the high-frequency signal and generate a label, thereby obtaining the visual signal label. The identifying the visual information feature in the target visual information can be realized by a pattern recognition algorithm, such as convolutional neural network (CNN), support vector machine (SVM), etc., to perform feature matching and identification on the visual signal label, thereby obtaining the visual information feature.

[0085] Based on the visual information features, the present invention analyzes the computing architecture in the preset brain-like computing model and queries the neuron mapping relationship in the computing architecture, which can improve the model's processing efficiency of complex visual information. At the same time, clarifying the neuron mapping relationship can enable more accurate transmission and integration of features when processing visual information, greatly enhancing the accuracy and efficiency of visual information analysis.

[0086] Among them, the preset brain-like computing model refers to a pre-built computing model that simulates the brain's neural processing mechanism. This type of model is usually based on the research results of brain neuroscience and uses mathematical models and algorithms to realize functions such as perception, learning, memory and decision-making of information. For example, some brain-like computing models built based on spiking neural networks (SNNs) can simulate the pulse emission behavior of biological neurons to process various types of information such as vision and hearing, providing new computing approaches for solving complex practical problems; the computing architecture refers to the organization and mutual relationship of the various components in the preset brain-like computing model, which determines how the model processes data and transmits information. For example, a computing architecture with a hierarchical structure can extract information through the underlying neurons when processing visual information such as images. Basic features such as edges are gradually integrated and abstracted by high-level neurons to realize the recognition of complex objects; and the fully connected computing architecture is, in some cases, more conducive to the extensive interaction of information between neurons and global information processing; the neuron mapping relationship refers to a set of relationships that describe the corresponding connections between different neurons in a preset brain-like computing model. The neuron mapping relationship reflects how neurons work together when the model processes specific tasks. For example, in visual information processing, some neurons responsible for perceiving local features of an image will be connected to neurons responsible for integrating these local features to recognize the overall shape of the object through a specific mapping relationship. Optionally, the analysis of the computing architecture in the preset brain-like computing model can be achieved through model analysis tools, such as: parsing and visualizing the computing architecture through tools such as Neural Network Libraries (NNL) and DeepLearning4J to obtain the computing architecture; the querying of the neuron mapping relationship in the computing architecture can be achieved through a graph database query method, such as: querying the connection relationship and weight distribution between neurons through graph database tools such as Neo4j and TigerGraph to obtain the neuron mapping relationship.

[0087] The present invention optimizes the neuron mapping relationship to obtain an optimized mapping network, which can eliminate redundant connections, make information transmission between neurons more efficient, reduce the waste of computing resources, and accurately capture complex visual information features, thereby greatly improving the accuracy and efficiency of visual information analysis, and enhancing the application effect of brain-like computing in security, autonomous driving and other fields.

[0088] Among them, the optimized mapping network refers to a new network structure obtained after adjusting the neuron mapping relationship according to the optimization strategy. In visual information processing, it can more effectively integrate visual information features and improve the model's processing capabilities for complex and changing visual scenes. For example, it can more accurately identify target objects in complex lighting environments, providing strong support for the realization of efficient brain-like computing-driven visual information analysis.

[0089] As an embodiment of the present invention, the relationship optimization of the neuron mapping relationship to obtain an optimized mapping network includes: querying the mapping feature parameters in the neuron mapping relationship; determining the optimization target corresponding to the neuron mapping relationship based on the mapping feature parameters; analyzing the optimization constraint conditions corresponding to the neuron mapping relationship based on the optimization target; formulating the optimization strategy corresponding to the neuron mapping relationship according to the optimization constraint conditions; and performing relationship optimization on the neuron mapping relationship based on the optimization strategy to obtain an optimized mapping network.

[0090] Among them, the mapping characteristic parameters refer to quantitative indicators that describe the characteristics of neuron mapping relationships. These parameters include connection strength, which determines the strength of signal transmission between neurons. For example, strong connections can transmit information more efficiently; connection delay, that is, the time required for a signal to be transmitted from a neuron to the neuron connected to it. Different delay settings will affect the timeliness of information transmission; there are also parameters related to the connection topology structure, such as the density of neuron connections, whether there is a specific connection pattern, etc.; the optimization goal refers to the expected result set for improving the neuron mapping relationship based on actual application needs. In the visual information processing scenario, it can be to improve the accuracy of identifying specific target objects, such as more accurately identifying suspicious persons in security monitoring; it can also be to increase the speed of the model processing visual information to meet the requirements of rapid response to real-time road conditions in autonomous driving; or to reduce the computing resources required for model operation to achieve a more efficient energy consumption ratio; the optimization constraints are Refers to the restrictive rules that must be followed when optimizing the neuron mapping relationship, including hardware-level restrictions, such as the limited memory capacity of computing devices, which limits the number of neuron connections and the scale of parameter storage; algorithm-level constraints, such as some optimization algorithms require that the adjustment of the mapping relationship must be carried out within a certain mathematical range to ensure the convergence and stability of the algorithm; there are also constraints in actual application scenarios, such as in medical image analysis, strict requirements on the accuracy of diagnostic results, which do not allow over-simplification of neuron mapping relationships at the expense of accuracy; the optimization strategy refers to the specific operating methods and steps formulated according to the optimization objectives and constraints. A gradient descent-based optimization strategy can be adopted to continuously adjust the parameter values ​​​​to move towards the optimization goal by calculating the gradient of the objective function with respect to the mapping feature parameters; or a heuristic algorithm, such as a genetic algorithm, can be used to simulate the biological evolution process, and through operations such as selection, crossover and mutation, find the optimal neuron mapping relationship under the premise of satisfying the constraints.

[0091] Furthermore, the querying of mapping feature parameters in the neuron mapping relationship can be achieved through feature extraction tools, such as: extracting key features in the mapping relationship through algorithms such as principal component analysis (PCA) and independent component analysis (ICA), thereby obtaining mapping feature parameters; the determination of the optimization target corresponding to the neuron mapping relationship can be achieved through a target optimization algorithm, such as: determining the objective function of mapping relationship optimization through algorithms such as gradient descent method and genetic algorithm, thereby obtaining the optimization target; the analysis of the optimization constraints corresponding to the neuron mapping relationship can be achieved through a constraint analysis tool, such as: analyzing the constraints in the mapping relationship through algorithms such as linear programming (LP) and nonlinear programming (NLP), thereby obtaining the optimization constraints; the formulation of the optimization strategy corresponding to the neuron mapping relationship can be achieved through a strategy optimization method, such as: designing an optimization strategy through algorithms such as reinforcement learning (RL) and Bayesian optimization, thereby obtaining the optimization strategy; the relationship optimization of the neuron mapping relationship can be achieved through a network optimization tool, such as: iteratively optimizing the mapping relationship through algorithms such as Adam optimizer and L-BFGS, thereby obtaining an optimized mapping network.

[0092] S3. Based on the optimized mapping network, analyze the information parsing path corresponding to the target visual information, determine the temporal distribution state of the target data in the target visual information based on the information parsing path, and calculate the parsing efficiency value corresponding to the target visual information according to the temporal distribution state.

[0093] Based on the optimized mapping network, the present invention analyzes the information parsing path corresponding to the target visual information, which helps to accurately locate the key nodes and processes of information processing, quickly identify possible parsing bottlenecks, and further optimize the network structure and parameter settings, significantly improving the overall efficiency and accuracy of visual information parsing, so that the system can more efficiently cope with complex visual scenes.

[0094] Among them, the information parsing path refers to the complete process that the target visual information goes through from input to final parsing result output in the optimization mapping network, which covers the information transmission path between each node, the parsing level passed through, and the processing process of key parsing nodes, and is jointly determined by the node connection relationship, information parsing level, key parsing nodes and their node parsing sequence.

[0095] As an embodiment of the present invention, the information parsing path corresponding to the target visual information is analyzed based on the optimized mapping network, including: querying the node connection relationship in the optimized mapping network; determining the information parsing level corresponding to the target visual information based on the node connection relationship; locating the key parsing nodes in the information parsing level; collecting the node parsing sequence corresponding to the key parsing node; and analyzing the information parsing path corresponding to the target visual information according to the node parsing sequence.

[0096] Among them, the node connection relationship refers to the way and property description of the mutual connection between each node (similar to neurons) in the optimized mapping network. For example, in the network simulating visual processing, the node responsible for perceiving color can be connected to the node responsible for shape recognition with a specific strength. This connection relationship determines the direction and efficiency of information transmission between nodes, and is the basic architecture for understanding the flow of information in the entire network; the information parsing level refers to the different levels of the target visual information processing process divided according to the node connection relationship. For example, the bottom layer can be the parsing level of basic elements such as image edges and textures, the middle layer is the parsing level that combines these basic elements into local object structures, and the high layer is the parsing level for understanding the meaning of the entire scene and object. Each level works together through node connection relationships to gradually complete the comprehensive analysis of visual information; the key analysis node refers to the node that plays a core role in the analysis of target visual information in the information analysis level. For example, in the visual task of identifying a vehicle, a node is specifically responsible for integrating the vehicle outline and motion feature information, which is the key analysis node; the node analysis sequence refers to the order in which the key analysis nodes process the target visual information and the record of specific operation steps. For example, a key analysis node first filters the input visual signal to remove noise, then extracts features, and then matches the extracted features with pre-stored templates. The ordered record of these operation steps is the node analysis sequence.

[0097] Furthermore, the query of the node connection relationship in the optimized mapping network can be achieved through a graph query tool, such as: querying the connection relationship and weight distribution between nodes through graph database tools such as Neo4j and TigerGraph, thereby obtaining the node connection relationship; the determination of the information parsing level corresponding to the target visual information can be achieved through a hierarchical analysis algorithm, such as: hierarchical division of the parsing path through hierarchical clustering, decision tree and other algorithms, thereby obtaining the information parsing level; the positioning of the key parsing nodes in the information parsing level can be achieved through a key node detection algorithm, such as: identifying the key nodes in the level through PageRank, centrality analysis and other algorithms, thereby obtaining the key parsing nodes; the collection of the node parsing sequence corresponding to the key parsing node can be achieved through a sequence generation tool, such as: generating a node parsing sequence through time series analysis, Markov chain and other algorithms, thereby obtaining a node parsing sequence; the analysis of the information parsing path corresponding to the target visual information can be achieved through a path analysis algorithm, such as: analyzing the optimal path of the parsing path through the shortest path algorithm, dynamic programming and other algorithms, thereby obtaining the information parsing path.

[0098] Based on the information parsing path, the present invention determines the temporal distribution state of the target data in the target visual information, can intuitively present the sequence and time interval of data processing, can accurately grasp the key changes of the target visual information at different moments, and greatly improve the fluency and real-time performance of visual information parsing.

[0099] Among them, the temporal distribution state refers to the distribution and change pattern of target data in the target visual information in the time dimension, which covers the time sequence of target data generation, transmission, processing and use, the specific position of each data on the time axis, and the time interval between data and other information. For example, in the target visual information of a video, the target data can be the detection data of the vehicle position at different times, and its temporal distribution state reflects the position change of the vehicle at different time points, whether the detection time interval is uniform, etc. Optionally, the determination of the temporal distribution state of the target data in the target visual information can be achieved through a temporal analysis tool, such as: performing temporal modeling and distribution analysis on the target data through algorithms such as time series decomposition and autoregressive integral moving average model, so as to obtain the temporal distribution state.

[0100] Furthermore, the present invention calculates the analysis efficiency value corresponding to the target visual information based on the temporal distribution state, which can intuitively judge the fluency and timeliness of the information analysis process, quickly locate the links with slow processing speed, and help to optimize the visual information analysis system in a targeted manner, reasonably allocate computing resources, and greatly improve the efficiency and performance of the overall system in processing visual information.

[0101] Among them, the parsing efficiency value refers to a comprehensive consideration of factors such as the time span in the target visual information parsing process, the amount of valid visual information successfully parsed in different time intervals, and the complexity of the target visual information itself. It is used to measure the efficiency of parsing the target visual information. The higher the value, the better the parsing efficiency of the target visual information under the same conditions.

[0102] As an embodiment of the present invention, calculating the parsing efficiency value corresponding to the target visual information according to the temporal distribution state includes:

[0103] The analytical efficiency value corresponding to the target visual information is calculated using the following formula:

[0104] ;

[0105] in, represents the analytical efficiency value corresponding to the target visual information, represents the total number of time intervals divided in the time series distribution state, The index of the number of time intervals, Indicates in The effective visual information successfully parsed within the time interval, Indicates the end time of parsing corresponding to the target visual information, Indicates the parsing start time corresponding to the target visual information, represents the adjustment coefficient, represents the average complexity corresponding to the target visual information, Indicates the reference complexity corresponding to the target visual information.

[0106] Specifically, the effective visual information refers to the The effective visual information in each time interval correctly recognized, processed and meeting specific requirements and standards by the analysis system, that is, the visual content actually valuable for subsequent analysis, decision and other operations, such as the feature information of the target object accurately recognized in image recognition; the adjustment coefficient is a parameter artificially set according to different visual information processing scenes and system performance and the like, which mainly functions to correct the calculation result of the analysis efficiency value, so as to more accurately reflect the actual analysis efficiency; the average complexity is an average measurement of the complexity of visual information in the whole target visual information analysis process, which comprehensively considers various characteristics of visual information in different time intervals, such as the number of objects in the picture, the relationship between objects, the richness of color and texture and the like, and is calculated through a certain algorithm, reflecting the overall complexity of the target visual information; the reference complexity is a reference value about the complexity of the target visual information determined according to historical experience, standard scene or pre-set rules.

[0107] Further, is the summation of the number of effective visual information successfully analyzed in each time interval in the whole analysis process, which reflects the actual effective results obtained in the analysis process, and the more the effective visual information, the higher the output of the analysis work and the greater the contribution to the analysis efficiency; represents the square root of the time interval between the analysis start time and the analysis end time , and the shorter the time under the same effective information analysis amount, the higher the analysis efficiency; the influence of the complexity of visual information on the analysis efficiency is considered, is an adjustment coefficient, which can adjust the influence degree of the complexity factor on the efficiency value according to the actual situation, represents the absolute value of the difference between the average complexity and the reference complexity , reflecting the deviation degree of the actual visual information complexity from the reference value, and when the actual complexity and the reference complexity differ greatly, the influence on the analysis efficiency will also be greater, so as to be reflected in the corresponding change of the analysis efficiency value in the formula.

[0108] S4, based on the analysis efficiency value, determining the analysis optimization direction corresponding to the target visual information, generating the real-time analysis strategy corresponding to the target visual information based on the analysis optimization direction, extracting the key execution parameters in the analysis strategy, and calculating the analysis decay rate corresponding to the target visual information based on the key execution parameters.

[0109] Based on the analytical efficiency value, the present invention determines the analytical optimization direction corresponding to the target visual information, accurately finds the key factors affecting efficiency, effectively adjusts the brain-like computing network architecture, and optimizes the algorithm process, thereby efficiently improving the visual information analysis capability to better meet the needs of actual application scenarios such as security and autonomous driving.

[0110] Among them, the analysis optimization direction refers to determining the specific direction of improving and optimizing the target visual information analysis system based on the located inefficient analysis area. For example, if the inefficient analysis area is caused by the low efficiency of the algorithm in processing complex image textures, then the analysis optimization direction can be to improve or replace the algorithm for processing textures; if the low efficiency is caused by insufficient hardware resources within a certain period of time, the optimization direction can be to increase hardware resources or optimize resource allocation strategies.

[0111] As an embodiment of the present invention, determining the analytical optimization direction corresponding to the target visual information based on the analytical efficiency value includes: performing data normalization processing on the analytical efficiency value to obtain a normalized efficiency value; constructing an efficiency distribution map corresponding to the target visual information based on the normalized efficiency value; setting an efficiency threshold corresponding to the efficiency distribution map; locating an inefficient analytical area in the target visual information based on the efficiency threshold; and determining the analytical optimization direction corresponding to the target visual information based on the inefficient analytical area.

[0112] The normalized efficiency value refers to the result obtained after data normalization processing of the original parsing efficiency value. It maps parsing efficiency values ​​of different ranges to a specific interval (such as [0, 1]) according to a certain mathematical method, facilitating comparison and analysis of parsing efficiency values ​​under different scenarios and conditions. The efficiency distribution graph refers to a graph constructed based on the normalized efficiency value, visually presenting the parsing efficiency of the target visual information in different parts or different time stages. For example, the horizontal axis represents time or different areas of visual information, and the vertical axis represents the normalized efficiency value, and the efficiency distribution is displayed in the form of a graph. The efficiency threshold refers to one or more set reference values ​​used as a standard for judging the parsing efficiency. When the normalized efficiency value is lower than the threshold, the corresponding area or time period is considered to have low parsing efficiency; when it is higher than the threshold, the parsing efficiency is considered to be at an acceptable or high level. The inefficient parsing area refers to the part of the target visual information where the parsing efficiency is lower than the threshold according to the set efficiency threshold. These areas can be specific time periods or certain local areas in the visual information. In these areas, the working efficiency of the parsing system does not meet the expected standards.

[0113] Further, the data normalization processing of the analysis efficiency value can be realized by a data standardization method, such as: mapping the analysis efficiency value to a unified range by minimum-maximum normalization, Z-score standardization and the like algorithms, so as to obtain a normalized efficiency value; the construction of the efficiency distribution graph corresponding to the target visual information can be realized by a data visualization tool, such as: drawing the distribution heat map or histogram of the efficiency value by Matplotlib, Seaborn and the like tools, so as to obtain the efficiency distribution graph; the setting of the efficiency threshold value corresponding to the efficiency distribution graph can be realized by a threshold analysis method, such as: automatically determining the optimal threshold value in the efficiency distribution graph by Otsu algorithm, K-means clustering and the like algorithms, so as to obtain the efficiency threshold value; the positioning of the low-efficiency analysis area in the target visual information can be realized by a region detection algorithm, such as: identifying the low-efficiency area in the efficiency distribution graph by edge detection, region growing algorithm and the like algorithms, so as to obtain the low-efficiency analysis area; the determination of the analysis optimization direction corresponding to the target visual information can be realized by an optimization algorithm, such as: analyzing the improvement direction of the low-efficiency area by gradient descent method, genetic algorithm and the like algorithms, so as to obtain the analysis optimization direction.

[0114] Based on the analysis optimization direction, the instant analysis strategy corresponding to the target visual information is generated, and the key execution parameters in the analysis strategy are extracted, which can quickly respond to the problems in visual information analysis, can improve the analysis efficiency and accuracy; the key execution parameters provide clear guidance for strategy implementation, facilitate accurate resource allocation, ensure efficient landing of the strategy, and help to process visual information in real time and high quality in security monitoring, automatic driving and the like scenes.

[0115] The instant analysis strategy refers to the specific steps in accordance with the analysis optimization direction, which can quickly and effectively solve the problems existing in the current analysis process, and covers a series of operation processes from data preprocessing, algorithm selection to result output, so as to realize efficient analysis of the target visual information; the key execution parameter refers to the parameter that plays a decisive role in the instant analysis strategy, which will significantly affect the execution effect of the analysis strategy, such as threshold setting in the algorithm, quantity ratio of resource allocation, time interval of data processing and the like, optionally, the generation of the instant analysis strategy corresponding to the target visual information can be realized by a strategy optimization algorithm, such as: generating an analysis strategy suitable for the target visual information by reinforcement learning, dynamic programming and the like algorithms, so as to obtain the instant analysis strategy; the extraction of the key execution parameter in the analysis strategy can be realized by a parameter extraction method, such as: identifying the core parameter in the strategy by principal component analysis, linear regression and the like methods, so as to obtain the key execution parameter.

[0116] Furthermore, the present invention calculates the resolution attenuation rate corresponding to the target visual information based on the key execution parameters, and can intuitively discover the downward trend of resolution efficiency with the passage of time or the increase of information processing volume. This helps to predict the performance bottleneck of the resolution system in advance, adjust the key execution parameters in time, maintain the efficient and stable resolution of the target visual information, and ensure the reliable operation of related applications (such as intelligent security monitoring and autonomous driving visual perception).

[0117] Among them, the analysis attenuation rate refers to the proportion of the degree to which the analysis effect decreases over time or other conditions due to the influence of various factors in the process of analyzing the target visual information based on key execution parameters. The larger the value, the more obvious the decline in the analysis effect of the target visual information.

[0118] As an embodiment of the present invention, the calculating, based on the key execution parameter, the analytical attenuation rate corresponding to the target visual information includes:

[0119] The analytical attenuation rate corresponding to the target visual information is calculated using the following formula:

[0120] ;

[0121] in, represents the analytical attenuation rate corresponding to the target visual information, Indicates the total number of parameters corresponding to the key execution parameters, Indicates the quantity index corresponding to the key execution parameter, Indicates the The actual value of the parameter corresponding to each key execution parameter, Indicates the The ideal value of the key execution parameters corresponding to the parameters, Indicates the The reference standard values ​​corresponding to the key execution parameters, and Respectively represent the start time and end time corresponding to the time function, Represents the time function corresponding to the target visual information.

[0122] In detail, the actual value of the parameter refers to the numerical value of the key execution parameter actually obtained when parsing the target visual information, which reflects the true status of the relevant factors in the current parsing process; the ideal value of the parameter refers to the pre-set numerical value that the key execution parameter is expected to reach under ideal circumstances, which is used as a reference standard for evaluating whether the actual value meets the expectation; the reference standard value refers to a numerical value with measurement significance determined for each key execution parameter, which is used to measure the difference between the actual value and the ideal value. By dividing the difference between the actual value and the ideal value by the reference standard value, it can more reasonably reflect the degree of deviation of the key execution parameter; the time function refers to a function that describes how one or some factors that affect the parsing effect change over time during the parsing process of the target visual information, and the function is in the time interval The integral within reflects the comprehensive effect of these influencing factors on the analysis process during this period.

[0123] S5. Based on the analytical attenuation rate, determine the analytical state corresponding to the target visual information, analyze the analytical dimension of the target visual information corresponding to the analytical state, and construct an optimized analytical solution corresponding to the target visual information based on the analytical dimension.

[0124] The present invention determines the parsing state corresponding to the target visual information based on the parsing attenuation rate, which helps to optimize the parsing strategy in a targeted manner, improve the processing capability of complex visual information, ensure the accuracy and immediacy of visual information processing in fields such as security monitoring and autonomous driving, and reduce the risk of misjudgment.

[0125] Among them, the parsing state refers to a comprehensive description of the current parsing effect of the target visual information, which can be obtained based on the parsing attenuation rate and its related analysis, including the status evaluation of the accuracy, stability and other aspects of the parsing, which can be used to guide subsequent parsing strategy adjustments.

[0126] As an embodiment of the present invention, determining the analytical state corresponding to the target visual information based on the analytical attenuation rate includes: determining the attenuation time axis corresponding to the analytical attenuation rate; extracting key attenuation nodes in the attenuation time axis; constructing an attenuation distribution matrix corresponding to the analytical attenuation rate based on the key attenuation nodes; querying the correlation features between different attenuation points in the attenuation distribution matrix; and determining the analytical state corresponding to the target visual information based on the correlation features.

[0127] Among them, the attenuation time axis refers to a coordinate axis with time as the dimension, which describes the process of the analysis attenuation rate changing with time. It records the dynamic changes of the analysis attenuation rate from the beginning of processing the target visual information to the current moment; the key attenuation nodes refer to those time points on the attenuation time axis that have special significance or can significantly affect the analysis process, such as the turning point where the analysis attenuation rate suddenly increases or decreases, or the time point when a specific threshold is reached. These nodes are crucial for analyzing the changes in the analysis effect; the attenuation distribution matrix refers to a matrix structure that arranges the key attenuation nodes and their corresponding analysis attenuation rates and other related information in an orderly manner. Through this matrix, the distribution of the analysis attenuation conditions at different time points can be intuitively displayed, which is convenient for data analysis and feature extraction; the correlation features refer to the internal connections and laws between different attenuation points in the attenuation distribution matrix, such as the change trend between attenuation points, the correlation of the change amplitude, and the relationship between the attenuation conditions in different dimensions. These features can reflect the characteristics of the analysis process.

[0128] Furthermore, the determination of the attenuation time axis corresponding to the analytical attenuation rate can be achieved through a time series analysis method, such as: constructing the trend of the analytical attenuation rate over time through algorithms such as the autoregressive integral moving average model and the long short-term memory network, thereby obtaining the attenuation time axis; the extraction of the key attenuation nodes in the attenuation time axis can be achieved through a key point detection algorithm, such as: identifying the key nodes in the attenuation time axis through algorithms such as peak detection and mutation point detection, thereby obtaining the key attenuation nodes; the construction of the attenuation distribution matrix corresponding to the analytical attenuation rate can be achieved through a matrix construction tool, such as: expressing the distribution of attenuation nodes in matrix form through tools such as NumPy and Pandas, thereby obtaining the attenuation distribution matrix; the query of the correlation features between different attenuation points in the attenuation distribution matrix can be achieved through an association analysis algorithm, such as: analyzing the correlation between attenuation points through algorithms such as the Pearson correlation coefficient and mutual information, thereby obtaining the correlation features; the determination of the analytical state corresponding to the target visual information can be achieved through a state evaluation algorithm, such as: performing a comprehensive evaluation of the analytical attenuation rate and the correlation features through algorithms such as support vector machines and random forests, thereby obtaining the analytical state.

[0129] By analyzing the analytical dimensions of the target visual information under the analytical state, the present invention helps to optimize brain-inspired computing networks and analytical strategies, improve adaptability to complex and changing visual scenes, enhance the accuracy and efficiency of visual information analysis, and provide more reliable support for applications such as security and autonomous driving.

[0130] Among them, the parsing dimension refers to the multiple aspects and angles involved in parsing the target visual information, covering different levels and characteristics of information processing, which may include but are not limited to spatial dimensions, such as the analysis of spatial features such as the position, size, and shape of objects in the visual scene; time dimension, that is, analyzing the changes in visual information over time, such as the motion trajectory and speed of the target object; it also includes semantic dimensions, which are used to understand the meaning conveyed by visual information, such as the category of objects in the image, behavioral intentions, etc.; and dimensions such as the accuracy, stability, and completeness of information to measure the parsing effect. Optionally, the analysis of the parsing dimension corresponding to the target visual information in the parsing state can be achieved through multidimensional data analysis methods, such as: reducing the dimension and visualizing the features in the parsing state through principal component analysis, t-distribution neighborhood embedding algorithm, etc., thereby obtaining the parsing dimension.

[0131] Furthermore, based on the analytical dimensions, the present invention constructs an optimized analytical solution corresponding to the target visual information, utilizes insights from each dimension, accurately adjusts the parameters of the brain-like computing network, and improves adaptability to complex scenarios. At the same time, the optimized solution can significantly enhance analytical efficiency and accuracy, providing strong support for visual information processing in key areas such as security and medical imaging.

[0132] Among them, the optimized analysis scheme refers to a set of comprehensive improvement strategies designed by comprehensively considering the target visual information analysis dimensions and utilizing the advantages of brain-like computing. For example, in order to solve the problem of fuzzy object edge recognition in the spatial dimension, the neuron connection weights of the brain-like computing network are optimized to enhance the sensitivity to shape features. In the time dimension, if the tracking of the moving target is delayed, the data processing speed is accelerated by adjusting the information analysis path. At the same time, the visual information is deeply mined in combination with semantic understanding to reduce misjudgment. Optionally, the construction of the optimized analysis scheme corresponding to the target visual information can be implemented through scheme construction tools, such as TensorFlow, PyTorch and other tools.

[0133] Compared with the problems described in the background technology, the present invention obtains real-time perception data corresponding to the target visual information and collects neural pulse signals corresponding to the real-time perception data, which can help to accurately identify the characteristics of visual information and optimize the neuron mapping relationship in the computing architecture. At the same time, it provides source data for a series of key operations such as subsequent analysis of information parsing paths and determination of parsing efficiency values, thereby greatly improving the accuracy of instant parsing of visual information. The present invention is based on the brain-like computing network to identify the visual information characteristics in the target visual information, effectively deal with complex situations such as light changes and occlusions, and accurately identify the contours, textures, dynamics and other features of the target object. This recognition method greatly improves the depth and accuracy of the understanding of visual information, and provides a solid foundation for subsequent analysis and decision-making of visual information. Furthermore, the present invention analyzes the information parsing path corresponding to the target visual information based on the optimized mapping network, which is It helps to accurately locate the key nodes and processes of information processing, quickly identify possible analysis bottlenecks, further optimize the network structure and parameter settings, significantly improve the overall efficiency and accuracy of visual information analysis, and enable the system to more efficiently cope with complex visual scenes. Furthermore, the present invention determines the analysis optimization direction corresponding to the target visual information based on the analysis efficiency value, accurately finds the key factors affecting efficiency, effectively adjusts the brain-like computing network architecture, optimizes the algorithm process, and thus efficiently improves the visual information analysis capability to better meet the needs of practical application scenarios such as security and autonomous driving. Finally, the present invention determines the analysis state corresponding to the target visual information based on the analysis attenuation rate, which helps to optimize the analysis strategy in a targeted manner, improve the processing capability of complex visual information, ensure the accuracy and immediacy of visual information processing in fields such as security monitoring and autonomous driving, and reduce the risk of misjudgment. Therefore, the brain-like computing-driven visual information instant analysis method and system provided in the embodiments of the present invention can improve the efficiency of intelligent analysis of visual information.

[0134] Example 2:

[0135] like Figure 2 The figure shows a functional module diagram of a visual information instant analysis system driven by brain-like computing according to the present invention.

[0136] The brain-inspired computing-driven instant visual information analysis system 200 described in the present invention can be installed in an electronic device. Depending on the functionality implemented, the brain-inspired computing-driven instant visual information analysis system can include a network construction module 201, a relationship optimization module 202, an efficiency value calculation module 203, an attenuation value calculation module 204, and a solution construction module 205. The modules described in the present invention, also referred to as units, are a series of computer program segments that can be executed by an electronic device processor and perform a fixed function, and are stored in the electronic device's memory.

[0137] In the embodiment of the present invention, the functions of each module / unit are as follows:

[0138] The network construction module 201 is used to obtain real-time perception data corresponding to target visual information, collect neural pulse signals corresponding to the real-time perception data, analyze signal perception curves corresponding to the neural pulse signals, and construct a brain-like computing network corresponding to the target visual information based on the signal perception curves;

[0139] The relationship optimization module 202 is configured to identify visual information features in the target visual information based on the brain-inspired computing network, analyze a computing architecture in a preset brain-inspired computing model based on the visual information features, query neuron mapping relationships in the computing architecture, and optimize the neuron mapping relationships to obtain an optimized mapping network;

[0140] The efficiency value calculation module 203 is used to analyze the information parsing path corresponding to the target visual information based on the optimized mapping network, determine the temporal distribution state of the target data in the target visual information based on the information parsing path, and calculate the parsing efficiency value corresponding to the target visual information according to the temporal distribution state;

[0141] The attenuation value calculation module 204 is configured to determine, based on the parsing efficiency value, a parsing optimization direction corresponding to the target visual information, generate an immediate parsing strategy corresponding to the target visual information based on the parsing optimization direction, extract key execution parameters from the parsing strategy, and calculate a parsing attenuation rate corresponding to the target visual information based on the key execution parameters;

[0142] The solution construction module 205 is used to determine the analysis state corresponding to the target visual information based on the analysis attenuation rate, analyze the analysis dimension of the target visual information corresponding to the analysis state, and construct an optimized analysis solution corresponding to the target visual information based on the analysis dimension.

[0143] In detail, each module in the brain-like computing driven visual information instant analysis system 200 in the embodiment of the present invention adopts the same method as above when in use. Figure 1 The same technical means as the brain-like computing-driven real-time analysis method of visual information described in the previous section can produce the same technical effects, so I will not go into details here.

[0144] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0145] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for real-time visual information analysis driven by brain-like computing, characterized in that: The method comprises: Acquiring real-time perception data corresponding to target visual information, collecting neural pulse signals corresponding to the real-time perception data, analyzing signal perception curves corresponding to the neural pulse signals, and constructing a brain-like computing network corresponding to the target visual information based on the signal perception curves; Based on the brain-inspired computing network, visual information features in the target visual information are identified; based on the visual information features, a computing architecture in a preset brain-inspired computing model is analyzed; a neuron mapping relationship in the computing architecture is queried; and the neuron mapping relationship is optimized to obtain an optimized mapping network; Analyzing an information parsing path corresponding to the target visual information based on the optimized mapping network, determining a temporal distribution state of target data in the target visual information based on the information parsing path, and calculating a parsing efficiency value corresponding to the target visual information based on the temporal distribution state; Determining a parsing optimization direction corresponding to the target visual information based on the parsing efficiency value, generating a real-time parsing strategy corresponding to the target visual information based on the parsing optimization direction, extracting key execution parameters from the parsing strategy, and calculating a parsing attenuation rate corresponding to the target visual information based on the key execution parameters; Based on the analytical attenuation rate, the analytical state corresponding to the target visual information is determined, the analytical dimension of the target visual information corresponding to the analytical state is analyzed, and based on the analytical dimension, an optimized analytical solution corresponding to the target visual information is constructed.

2. The method for instant visual information analysis driven by brain-inspired computing as claimed in claim 1, characterized in that: The step of constructing a brain-like computing network corresponding to the target visual information according to the signal perception curve includes: Extracting perception feature parameters corresponding to the signal perception curve; Determining a feature extraction mode corresponding to the signal perception curve based on the perception feature parameter; Analyzing the feature distribution pattern corresponding to the target visual information based on the feature extraction mode; Formulate a neural node connection rule corresponding to the target visual information according to the feature distribution rule; Based on the neural node connection rules, a brain-like computing network corresponding to the target visual information is constructed.

3. The method for instant visual information analysis driven by brain-inspired computing as claimed in claim 1, characterized in that: The identifying visual information features in the target visual information based on the brain-inspired computing network includes: Querying the activation state of neurons in the brain-inspired computing network; collecting a state response feature set corresponding to the neuron activation state; Extracting characteristic high-frequency signals from the state response feature set; Generating a visual signal label corresponding to the characteristic high-frequency signal; Based on the visual signal label, visual information features in the target visual information are identified.

4. The method for instant visual information analysis driven by brain-inspired computing as claimed in claim 1, characterized in that: Optimizing the neuron mapping relationship to obtain an optimized mapping network includes: Querying mapping feature parameters in the neuron mapping relationship; Determining an optimization target corresponding to the neuron mapping relationship based on the mapping feature parameters; Based on the optimization goal, analyzing the optimization constraints corresponding to the neuron mapping relationship; Formulate an optimization strategy corresponding to the neuron mapping relationship according to the optimization constraints; Based on the optimization strategy, the neuron mapping relationship is optimized to obtain an optimized mapping network.

5. The method for instant visual information analysis driven by brain-inspired computing as claimed in claim 1, characterized in that: The analyzing the information parsing path corresponding to the target visual information based on the optimized mapping network includes: Querying the node connection relationship in the optimized mapping network; Determining an information parsing level corresponding to the target visual information based on the node connection relationship; Locating key parsing nodes in the information parsing hierarchy; Collecting a node parsing sequence corresponding to the key parsing node; According to the node parsing sequence, an information parsing path corresponding to the target visual information is analyzed.

6. The method for instant visual information analysis driven by brain-inspired computing as claimed in claim 1, characterized in that: Calculating the parsing efficiency value corresponding to the target visual information according to the temporal distribution state includes: The analytical efficiency value corresponding to the target visual information is calculated using the following formula: ； in, represents the analytical efficiency value corresponding to the target visual information, represents the total number of time intervals divided in the time series distribution state, The index of the number of time intervals, Indicates in The effective visual information successfully parsed within the time interval, Indicates the end time of parsing corresponding to the target visual information, Indicates the parsing start time corresponding to the target visual information, represents the adjustment coefficient, represents the average complexity corresponding to the target visual information, Indicates the reference complexity corresponding to the target visual information.

7. The method for instant visual information analysis driven by brain-inspired computing as claimed in claim 1, characterized in that: The determining, based on the parsing efficiency value, a parsing optimization direction corresponding to the target visual information includes: Performing data normalization processing on the analytical efficiency value to obtain a normalized efficiency value; Based on the normalized efficiency value, constructing an efficiency distribution map corresponding to the target visual information; Setting an efficiency threshold corresponding to the efficiency distribution graph; Based on the efficiency threshold, locating an inefficient parsing area in the target visual information; According to the inefficient parsing area, a parsing optimization direction corresponding to the target visual information is determined.

8. The method for instant visual information analysis driven by brain-inspired computing as claimed in claim 1, characterized in that: The calculating, based on the key execution parameter, the analytical attenuation rate corresponding to the target visual information includes: The analytical attenuation rate corresponding to the target visual information is calculated using the following formula: ； in, represents the analytical attenuation rate corresponding to the target visual information, Indicates the total number of parameters corresponding to the key execution parameters, Indicates the quantity index corresponding to the key execution parameter, Indicates the The actual value of the parameter corresponding to each key execution parameter, Indicates the The ideal value of the key execution parameters corresponding to the parameters, Indicates the The reference standard values ​​corresponding to the key execution parameters, and Respectively represent the start time and end time corresponding to the time function, Represents the time function corresponding to the target visual information.

9. The method for instant visual information analysis driven by brain-inspired computing as claimed in claim 1, characterized in that: The determining, based on the resolution attenuation rate, a resolution state corresponding to the target visual information includes: Determining a decay time axis corresponding to the analytical decay rate; Extracting key decay nodes in the decay timeline; Based on the key attenuation nodes, constructing an attenuation distribution matrix corresponding to the analytical attenuation rate; Querying correlation features between different attenuation points in the attenuation distribution matrix; Based on the associated features, a parsing state corresponding to the target visual information is determined.

10. A visual information real-time analysis system driven by brain-like computing, characterized in that: The system comprises: a network construction module for acquiring real-time perception data corresponding to target visual information, collecting neural pulse signals corresponding to the real-time perception data, analyzing signal perception curves corresponding to the neural pulse signals, and constructing a brain-like computing network corresponding to the target visual information based on the signal perception curves; a relationship optimization module for identifying visual information features in the target visual information based on the brain-inspired computing network, analyzing a computing architecture in a preset brain-inspired computing model based on the visual information features, querying neuron mapping relationships in the computing architecture, and performing relationship optimization on the neuron mapping relationships to obtain an optimized mapping network; an efficiency value calculation module, configured to analyze, based on the optimized mapping network, an information parsing path corresponding to the target visual information, determine, based on the information parsing path, a temporal distribution state of target data in the target visual information, and calculate, based on the temporal distribution state, a parsing efficiency value corresponding to the target visual information; an attenuation value calculation module, configured to determine, based on the parsing efficiency value, a parsing optimization direction corresponding to the target visual information, generate, based on the parsing optimization direction, a real-time parsing strategy corresponding to the target visual information, extract key execution parameters from the parsing strategy, and calculate, based on the key execution parameters, a parsing attenuation rate corresponding to the target visual information; A solution construction module is used to determine the analysis state corresponding to the target visual information based on the analysis attenuation rate, analyze the analysis dimension of the target visual information corresponding to the analysis state, and construct an optimized analysis solution corresponding to the target visual information based on the analysis dimension.

Images