External access compliance verification processing method and system based on intelligent hardware

Abstract

The invention provides an external access compliance verification processing method and system based on intelligent hardware, and belongs to the technical field of data processing, and the system specifically comprises an intelligent hardware data acquisition module which is responsible for realizing real-time acquisition and data processing of multi-modal data, a multi-modal sensing module which is responsible for realizing real-time sensing and understanding of the environment, and a data processing module which is responsible for realizing real-time verification of the external access compliance. The semantic analysis module is responsible for realizing semantic analysis and understanding of voice data, visual scenes and behavior actions in an external visiting process, and the compliance verification module is responsible for realizing compliance check and risk assessment of external visiting behaviors based on a predefined compliance rule and a machine learning algorithm. The cloud platform is responsible for data storage of the intelligent hardware and analysis and processing of violation behaviors and transmitting the violation behaviors back to the intelligent hardware, early warning and assistance are carried out on the external visiting process, and the reliability of the external compliance process is improved.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology, and in particular relates to a method and system for verifying compliance of external visits based on smart hardware. Background Technology

[0002] Post-loan management is a crucial aspect of risk control for financial institutions, and field visits, as a key component of post-loan management, undertake important functions such as debt collection, asset verification, and risk assessment. However, traditional field visit models present numerous problems and challenges.

[0003] To address the aforementioned technical issues, existing technical solutions have provided numerous methods for identifying and processing collection behaviors based on smart devices. However, these solutions all neglect the construction of target battery ranges for identifying and analyzing violations using multimodal data on cloud platforms. Consequently, when the remaining battery power beyond the target range is low, it becomes impossible to determine a targeted solution for identifying and processing violations, making it difficult to meet the needs of on-site data recording and analysis.

[0004] Specifically, this application provides a smart hardware field perception, semantic analysis, and smart hardware-based field visit compliance verification and processing system and method. Specifically, it utilizes smart hardware to analyze and process compliance status using multimodal data within a target power range. In order to determine the identification, analysis, and processing strategy for subsequent violations when the remaining power decreases after the target power range, it greatly improves the reliability of on-site data recording, analysis, and processing. Summary of the Invention

[0005] To achieve the objectives of this invention, the following technical solution is adopted: Specifically, this application provides an external visit compliance verification and processing system based on smart hardware, which includes: The smart hardware data acquisition module is responsible for real-time acquisition and processing of multimodal data; the multimodal perception module is responsible for environmental perception and understanding; the semantic analysis module is responsible for semantic analysis and understanding of voice data, visual scenes, and behavioral actions during outbound visits; the compliance verification module is responsible for compliance checks and risk assessments of outbound visit behavior based on predefined compliance rules and machine learning algorithms; and the cloud platform is responsible for storing smart hardware data, analyzing and processing violations, transmitting the data back to the smart hardware, and providing early warnings and assistance during outbound visits.

[0006] Furthermore, the multimodal data includes audio, text, video, and images that the smart hardware can simultaneously collect.

[0007] Furthermore, the cloud platform, based on the analysis results of multi-dimensional data using a cloud-based multimodal big data model, determines the behaviors and multilingual scripts that foreign visitors should take, and transmits them to the smart hardware terminal in real time to provide real-time early warning and assistance during the foreign visit process.

[0008] Furthermore, the multilingual scripts are a strategy that supports real-time generation of Chinese and English scripts.

[0009] Furthermore, the aforementioned outbound visit process is monitored and early warning is issued, specifically including: Based on the dialogue data between debt collectors and respondents, we analyze the respondents' repayment intentions and emotions in real time. Based on images of the respondents' facial expressions and actions, we analyzed whether the respondents exhibited any abnormal behavior.

[0010] Furthermore, it also includes a security module: including data encryption, identity authentication, access control, and audit log functions, responsible for data security management of smart hardware and cloud platforms.

[0011] Secondly, this application provides a method for verifying compliance with external access based on smart hardware, applied to the aforementioned system for verifying compliance with external access for smart hardware, specifically including: S1 uses the access data of smart hardware on the cloud platform as a basis to determine the power consumption data of the smart hardware using the fusion strategy for data transmission processing, and uses the power consumption data as a basis to determine the remaining power range for compliance identification processing based on the fusion strategy, and uses it as the target power range. S2 determines, based on the interval between the remaining power of the smart hardware and the target power range, that the overlap risk of smart hardware in the target power range does not meet the requirements, and determines the overlapping associated smart hardware in the target power range based on the remaining power of the smart hardware. Based on the association between the smart hardware and the target power range, and the data of overlapping smart hardware, S3 determines the compliance verification strategy of the smart hardware on the cloud platform based on the fusion strategy, and uses the processing results of the compliance verification of overlapping smart hardware under each compliance verification strategy to determine whether the target power range of the smart hardware needs to be adjusted.

[0012] Furthermore, the fusion strategy involves transmitting both environmental images and voice data to the cloud platform.

[0013] Furthermore, the method for determining the target power range is as follows: Based on the power consumption data, the power consumption of the smart hardware is determined; Based on the power consumption data of transmitting voice data to the cloud platform separately, determine the power consumption data of transmitting voice data to the cloud platform separately, and take it as the voice power consumption. By utilizing the data transmission duration and power consumption of the smart hardware under voice power consumption in different remaining power ranges, a target power range is determined within the remaining power range.

[0014] Furthermore, determining whether the target power range adjustment process for the smart hardware needs to be performed specifically includes: Based on the processing results of compliance verification of overlapping smart hardware under various compliance verification strategies, identify overlapping smart hardware with violations and classify them as violating smart hardware. Based on the different compliance verification strategies for overlapping and related smart hardware, determine the overlapping and related smart hardware data that adopt different compliance verification strategies; Based on overlapping and non-compliant smart hardware data using different compliance verification strategies, it is determined whether the target power range of the smart hardware needs to be adjusted.

[0015] The beneficial effects of this invention are as follows: Based on power consumption data, the remaining power range for compliance identification and processing is determined using a fusion strategy. This enables the screening of the remaining power range for compliance identification and processing. Before the power drops to a lower range, the compliance status is accurately identified, which in turn ensures the determination of the identification strategy for violations when the power drops to a lower range, thus improving the reliability of violation identification and processing.

[0016] Reduced compliance risks: Through real-time compliance verification and risk warnings, compliance risk events during external visits were reduced by 75%. Customer complaint rates decreased from 8% to 2%, and legal disputes decreased by 80%. This improvement not only reduced the organization's legal risks but also enhanced its social image.

[0017] Data quality improvement: Through automated data collection and intelligent analysis, the completeness of field visit data increased from 70% to 95%, and the accuracy increased from 75% to 92%. The proportion of structured data increased from 30% to 85%, greatly improving the usability and value of the data.

[0018] Cost-effectiveness optimization: Although the initial investment in the system was substantial, the return on investment exceeded 300% due to benefits such as improved efficiency, reduced risk, and enhanced quality. The cost per person for field visits decreased by 35%, and overall operating costs decreased by 28%.

[0019] Other features and advantages will be set forth in the following description, and the objects and other advantages of the invention are realized and obtained through the structures particularly pointed out in the description and the drawings.

[0020] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0021] The above and other features and advantages of the present invention will become more apparent from a detailed description of exemplary embodiments thereof with reference to the accompanying drawings.

[0022] Figure 1 A framework diagram of an external visit compliance verification and processing system based on smart hardware. Figure 2 This is a flowchart of an external visit compliance verification process based on smart hardware; Figure 3 A flowchart illustrating the method for determining the target power range; Figure 4 A flowchart of a method for determining the overlap risk of smart hardware within a target power range that does not meet requirements; Figure 5 A flowchart illustrating the method for determining overlapping smart hardware within a target power range. Detailed Implementation

[0023] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0024] Example 1 like Figure 1 As shown, Figure 1 As shown, this application provides an external visit compliance verification and processing system based on smart hardware, specifically including: The smart hardware data acquisition module is responsible for real-time acquisition and processing of multimodal data; the multimodal perception module is responsible for environmental perception and understanding; the semantic analysis module is responsible for semantic analysis and understanding of voice data, visual scenes, and behavioral actions during outbound visits; the compliance verification module is responsible for compliance checks and risk assessments of outbound visit behavior based on predefined compliance rules and machine learning algorithms; and the cloud platform is responsible for storing smart hardware data, analyzing and processing violations, transmitting the data to the smart hardware, and providing early warnings and assistance during outbound visits.

[0025] Furthermore, the multimodal data includes audio, text, video, and images that the smart hardware can simultaneously collect.

[0026] Furthermore, the cloud platform, based on the analysis results of multi-dimensional data using a cloud-based multimodal big data model, determines the behaviors and multilingual scripts that foreign visitors should take, and transmits them to the smart hardware terminal in real time to provide real-time early warning and assistance during the foreign visit process.

[0027] Furthermore, the multilingual scripts are a strategy that supports real-time generation of Chinese and English scripts.

[0028] Furthermore, the aforementioned outbound visit process is monitored and early warning is issued, specifically including: Based on the dialogue data between debt collectors and respondents, we analyze the respondents' repayment intentions and emotions in real time. Based on images of the respondents' facial expressions and actions, we analyzed whether the respondents exhibited any abnormal behavior.

[0029] Furthermore, it also includes a security module: including data encryption, identity authentication, access control, and audit log functions, responsible for data security management of smart hardware and cloud platforms.

[0030] Furthermore, it also includes an intelligent decision support module that combines business knowledge base and historical data analysis to provide real-time decision suggestions and operational guidance to field visitors. Furthermore, it also includes a security module: including data encryption, identity authentication, access control, and audit log functions, responsible for data security management of smart hardware and cloud platforms.

[0031] The intelligent hardware data acquisition module is the core of the system. Its design must fully consider the special needs and operating environment of field visits. The optical display system adopts advanced micro-display technology and optical design to achieve high-definition, low-power visual display. The display screen supports 1080P resolution and a brightness of up to 3000 nits, enabling clear display even in strong light environments. The optical system uses a freeform surface design with a field of view of 40 degrees, providing users with an immersive visual experience.

[0032] The image acquisition system integrates dual-lens high-definition cameras, supporting 4K video recording and high-speed image acquisition. The cameras utilize Sony IMX series sensors, boasting excellent low-light performance and color reproduction capabilities. They support advanced features such as optical image stabilization, autofocus, and HDR, ensuring high-quality image data is obtained under various environmental conditions.

[0033] The audio processing system employs a microphone array design, supporting 360-degree omnidirectional recording and intelligent noise reduction. The microphone array consists of four high-performance MEMS microphones, supporting audio processing technologies such as beamforming, echo cancellation, and noise suppression. It also integrates bone conduction speakers to provide users with private audio feedback.

[0034] The sensor system integrates multiple sensors, including a nine-axis inertial sensor, an ambient light sensor, and a proximity sensor, to achieve comprehensive perception of user status and environmental conditions. The inertial sensor supports high-precision attitude detection and motion tracking, the ambient light sensor supports automatic brightness adjustment, and the proximity sensor supports intelligent wake-up and energy-saving control.

[0035] The positioning and navigation system integrates multi-mode positioning technologies such as GPS, BeiDou, and GLONASS, supporting centimeter-level high-precision positioning. It also integrates auxiliary positioning technologies such as WiFi and Bluetooth, providing accurate location information even indoors.

[0036] The communication system supports multiple communication technologies such as 5G, WiFi 6, and Bluetooth 5.0, ensuring stable data transmission in various network environments. The system employs a multi-antenna design and supports MIMO technology to improve communication quality and transmission speed.

[0037] The computing system employs a high-performance ARM processor and a dedicated AI chip, supporting complex edge computing and AI inference. The processor is manufactured using a 7nm process, boasting powerful computing capabilities and excellent energy efficiency. The AI ​​chip is specifically optimized for deep learning inference, enabling efficient execution of various AI algorithms.

[0038] The power management system employs a high-energy-density lithium battery and intelligent power management technology to ensure stable operation of the device over extended periods. The battery capacity reaches 3000mAh, supporting over 8 hours of continuous operation. The power management system supports functions such as fast charging, wireless charging, and low-power mode.

[0039] The multimodal perception module is the core technical component of the system, responsible for real-time, comprehensive, and accurate perception and understanding of the external environment.

[0040] The visual perception module, based on deep convolutional neural network technology, enables intelligent understanding of visual scenes. This module integrates multiple visual algorithms, including object detection, scene segmentation, face recognition, and behavior analysis. The object detection algorithm uses the YOLOv8 architecture and can detect and identify various objects in the field of view in real time, including people, vehicles, buildings, and signs. The algorithm has undergone specialized training and optimization, achieving a detection accuracy of over 95% in outdoor scenarios.

[0041] The scene segmentation algorithm, employing the DeepLab v3+ architecture, enables pixel-level semantic segmentation of visual scenes, accurately identifying semantic information in different regions. This provides crucial foundational data for subsequent scene understanding and analysis. The face recognition algorithm, based on the ArcFace architecture, supports high-precision face detection, feature extraction, and identity recognition. The algorithm achieves a 99.8% recognition accuracy on the LFW dataset, demonstrating its ability to accurately identify individuals in complex environments.

[0042] The behavior analysis algorithm, based on temporal convolutional networks, can identify and analyze various human behaviors, such as walking, standing, talking, and arguing. This provides crucial support for behavioral monitoring and risk warning during field visits. The audio perception module, based on deep learning and signal processing technologies, enables intelligent analysis and understanding of audio information.

[0043] The speech recognition algorithm employs a Transformer architecture, supporting real-time speech recognition in multiple languages ​​and dialects. The algorithm achieves a 98% accuracy rate in quiet environments and maintains over 90% accuracy in noisy environments. The emotion recognition algorithm, based on the fusion analysis of acoustic and semantic features, can accurately identify the speaker's emotional state, such as anger, anxiety, cooperation, and resistance. This provides crucial information for situation assessment and risk warning during field visits.

[0044] The speaker recognition algorithm employs x-vector technology, accurately identifying different speakers and supporting the analysis of multi-person dialogue scenarios. This ensures the integrity and accuracy of visit records. The sound event detection algorithm can identify various sound events in the environment, such as doorbells, telephone rings, arguments, and cries for help. This supports environmental monitoring and security alerts during visits.

[0045] The environmental perception module, based on multi-sensor fusion technology, achieves comprehensive perception of the external environment, including location perception and motion perception. Specifically, the location perception algorithm integrates multiple technologies such as GPS, inertial navigation, and visual positioning to achieve high-precision location positioning and trajectory tracking. Positioning accuracy reaches within 1 meter in open environments and within 3 meters in complex environments. The motion perception algorithm, based on inertial sensor data, can accurately identify the user's movement state and behavior patterns, such as walking, driving, standing still, and rapid movement. This provides support for status monitoring and anomaly detection during the visit. The environmental perception algorithm can identify various characteristics of the external environment, such as indoor and outdoor environments, lighting conditions, noise levels, and personnel density. This provides a basis for the system's adaptive adjustment and optimization.

[0046] The semantic analysis module is the intelligent core of the system, responsible for deep semantic understanding and analysis of multimodal information. The multimodal large-scale model architecture adopts the latest Transformer architecture and multimodal fusion technology, capable of simultaneously processing information from multiple modalities such as text, images, and audio. With a model parameter scale reaching the 10 billion level, it possesses powerful understanding and reasoning capabilities.

[0047] The model's visual encoder, based on the Vision Transformer architecture, transforms image information into high-dimensional feature representations. Pre-trained on large-scale image data, the encoder possesses excellent visual understanding capabilities. The model's audio encoder, based on the Wav2Vec architecture, transforms audio information into semantic feature representations. The encoder supports multilingual audio processing and exhibits powerful speech understanding capabilities.

[0048] The model's text encoder is based on the BERT architecture, enabling it to handle complex textual information and semantic reasoning. The encoder is pre-trained on a large-scale text corpus, exhibiting excellent language understanding capabilities. The multimodal fusion layer employs attention mechanisms and cross-modal learning techniques, effectively integrating information from different modalities to achieve a more comprehensive and accurate understanding.

[0049] Domain-adaptive optimization is implemented for specific scenarios related to post-loan management field visits. The large model undergoes specialized domain-adaptive training and optimization. The training dataset contains a large amount of field visit scenario data, covering various typical field visit situations and dialogue content. The dataset has undergone professional annotation and cleaning to ensure data quality and annotation accuracy.

[0050] The training process employs a multi-task learning strategy, simultaneously optimizing multiple related tasks such as scene understanding, dialogue analysis, sentiment recognition, and risk assessment. This approach improves the model's performance in specific domains. Model optimization utilizes techniques such as knowledge distillation and model compression to reduce computational complexity while maintaining performance, enabling it to run efficiently on edge devices.

[0051] The real-time inference engine is designed to be highly efficient, supporting real-time analysis and processing of multimodal information. Employing a streaming architecture, the engine can perform real-time analysis of continuous multimodal data streams. It supports incremental inference and caching optimization, significantly improving inference efficiency.

[0052] The inference engine integrates acceleration technologies such as model quantization, operator fusion, and memory optimization, significantly improving inference speed while maintaining accuracy. Inference latency on edge devices is controlled to within 100 milliseconds. The inference engine supports dynamic load balancing and resource scheduling, dynamically adjusting the allocation of computing resources according to actual needs to ensure system stability and reliability.

[0053] The compliance verification module is a crucial component for ensuring the compliance and security of field visits. Through intelligent rule checks and risk assessments, it utilizes environmental images or voice data for compliance verification, ensuring that the field visit process complies with relevant laws, regulations, and internal standards.

[0054] The compliance rules engine has built a comprehensive compliance rules system covering all aspects of foreign visit behavior. The behavioral norms define the standards of conduct that foreign visitors should follow, including polite language, professional image, reasonable timing, and appropriate methods. The rules employ a formalized descriptive approach, supporting automated checking and verification.

[0055] The legal and regulatory rules integrate relevant legal and regulatory requirements, including personal information protection, debt collection standards, and consumer rights protection. The rule base is regularly updated to ensure consistency with the latest laws and regulations. Internal rules, combined with the financial institution's internal management systems, define specific operational procedures and process requirements. The rules support personalized configuration to adapt to the management needs of different institutions.

[0056] The intelligent behavior analysis algorithm, based on deep learning technology, enables intelligent analysis and evaluation of outbound visit behavior. The behavior recognition algorithm can accurately identify various actions during the outbound visit process, such as knocking, conversing, showing identification, and recording information. The algorithm employs a temporal convolutional network architecture to model and analyze behavioral sequences.

[0057] The cloud platform, based on the analysis results of multi-dimensional data by the cloud multimodal big model, determines the behavior and multilingual scripts that foreign visitors should take, and transmits them to the smart hardware in real time to provide real-time warnings and assistance during the foreign visit process. The multilingual scripts are a strategy that supports the real-time generation of Chinese and English scripts.

[0058] The cloud platform serves as the unified management and service center for the system, providing robust backend support and service guarantees for the entire external visit management system. The distributed storage system employs advanced distributed storage technology to ensure the secure storage and efficient access to massive amounts of external visit data. The data storage architecture is based on the Hadoop Distributed File System and object storage technology, supporting the storage and management of petabyte-level data. The storage system utilizes a multi-replica mechanism to ensure data reliability and availability.

[0059] Data tiered management categorizes data into three levels—hot data, warm data, and cold data—based on access frequency and importance. Hot data is stored on high-performance SSDs, warm data on hard disk drives (HDDs), and cold data in tape libraries, achieving an optimal balance between cost and performance. Data compression and deduplication technologies significantly reduce storage space usage. Video data uses the H.265 encoding standard, achieving a compression rate more than 50% higher than H.264. Audio data uses AAC encoding, greatly reducing file size while maintaining sound quality.

[0060] The big data analytics engine, based on big data processing frameworks such as Spark and Flink, enables in-depth analysis and mining of field visit data. The batch processing analysis engine supports offline analysis of large-scale historical data, revealing patterns and trends in field visit operations. Analysis includes field visit success rate statistics, time efficiency analysis, geographical distribution characteristics, and risk factor identification.

[0061] The stream processing analytics engine supports real-time online data analysis, enabling timely detection of anomalies and risk signals. Stream processing latency is controlled within seconds, ensuring real-time performance. The machine learning platform integrates mainstream machine learning frameworks such as TensorFlow and PyTorch, supporting the training and deployment of various AI models. The platform provides an automated model training pipeline, supporting functions such as hyperparameter optimization, model selection, and performance evaluation.

[0062] The business management system provides complete management functions for field visits, supporting full-process management of field visit work. The case management module supports the entire lifecycle management of field visit cases, including creation, allocation, tracking, and closure. The module provides a flexible case classification and tagging system, supporting multi-dimensional case queries and statistics.

[0063] The personnel management module supports functions such as information management, access control, performance evaluation, and training records for field visitors. The module provides complete personnel files and work records, and supports personnel capability assessment and optimized configuration. The scheduling management module supports intelligent scheduling and route optimization for field visit tasks. The module considers factors such as personnel capabilities, geographical location, traffic conditions, and case priority to achieve optimal task allocation.

[0064] The quality management module supports the monitoring and evaluation of the quality of external visits. The module provides a multi-dimensional quality indicator system, utilizes a semantic analysis module to achieve cloud-based identification and processing of violations, and offers rich data reporting and analysis functions to provide data support for management decisions.

[0065] The real-time monitoring dashboard displays the real-time status and key indicators of field visits, including the number of online personnel, ongoing field visit tasks, risk warning information, and system operation status. The statistical reporting system supports data statistics and analysis across multiple dimensions, including time, region, personnel, and case dimensions. Reports support customizable configurations to meet the needs of different users.

[0066] The trend analysis function can identify the development trends and changing patterns of field visits, providing a basis for business optimization and strategic planning. The analysis includes success rate trends, efficiency changes, and risk evolution. The predictive analysis function, based on historical data and machine learning models, can predict future field visit needs and resource allocation. The prediction results provide a reference for personnel deployment and resource planning.

[0067] The security module is a crucial component of the system, ensuring the security, reliability, and compliance of external access data and system operation. Data security protection employs a multi-layered mechanism to safeguard sensitive data. Data encryption combines AES-256 symmetric encryption and RSA-2048 asymmetric encryption. Data during transmission is encrypted using the TLS 1.3 protocol, and stored data utilizes transparent data encryption technology.

[0068] The key management system employs a Hardware Security Module (HSM) for key generation, storage, and management. A hierarchical key management strategy is used, with different keys protecting different levels of data. Data anonymization technology automatically de-identifies sensitive personal information, ensuring data availability while protecting personal privacy. Anonymization methods include data masking, data replacement, and data scrambling.

[0069] The system establishes a comprehensive identity authentication and access control framework. Multi-factor authentication requires users to verify their identity through various methods such as passwords, SMS verification codes, and biometrics. The system supports biometric authentication methods such as fingerprint recognition, facial recognition, and voiceprint recognition.

[0070] Role-based access control (RBAC) assigns appropriate system permissions based on user roles and responsibilities. Access control is precise down to the functional module and data field level, ensuring users can only access authorized resources. Single sign-on (SSO) technology allows users to access multiple related systems with a single authentication, improving user experience while ensuring security.

[0071] The session management system monitors and manages user sessions throughout the entire process, and supports security functions such as session timeout, remote login detection, and abnormal behavior identification. The network security protection system has built a multi-layered network security protection system.

[0072] The firewall system employs next-generation firewall technology, supporting application-layer detection, intrusion prevention, and malware detection. Firewall rules are finely configured based on business needs to ensure network access security. The intrusion detection system (IDS) monitors network traffic and system activity in real time, promptly detecting and reporting security threats. The system utilizes a dual detection mechanism based on both signature and behavior.

[0073] The DDoS protection system effectively defends against various distributed denial-of-service attacks, ensuring system availability. The system supports multiple protection strategies such as traffic scrubbing, blacklisting / whitelisting, and rate limiting. Network isolation technology physically or logically isolates networks of different security levels to prevent the lateral propagation of security threats.

[0074] The audit and compliance management system has established a complete audit and compliance management mechanism. The operation audit system records all user operations and system events, forming a complete audit log. The audit log includes detailed information such as operation time, user, operation content, and operation result.

[0075] Data access auditing tracks all data access activities, including data queries, downloads, modifications, and deletions. The auditing system can identify abnormal data access patterns and promptly detect potential security threats. The compliance check system automatically checks the system's security configuration and operational behavior based on relevant laws, regulations, and industry standards. The check results generate a compliance report, providing a basis for compliance management.

[0076] The security incident response mechanism defines the processes for classifying, reporting, handling, and recovering from security incidents. The mechanism supports a combination of automated response and manual intervention to ensure timely handling of security incidents.

[0077] Specifically, the warning system for the aforementioned outbound visit process includes: Based on the dialogue data between debt collectors and respondents, we analyze the respondents' repayment intentions and emotions in real time. Based on images of the respondents' facial expressions and actions, we analyzed whether the respondents exhibited any abnormal behavior.

[0078] Specifically, the abnormal behavior detection algorithm can identify potentially risky abnormal behaviors, such as heated arguments, threats, intimidation, and violent tendencies. Employing unsupervised learning methods, the algorithm can discover abnormal patterns not previously seen in the training data. The compliance assessment algorithm comprehensively considers multiple dimensions, including behavioral norms, dialogue content, and environmental factors, to evaluate the compliance of outbound visits. The algorithm outputs a compliance score and risk level, providing a basis for management decisions.

[0079] The real-time risk warning mechanism establishes a multi-layered risk warning system, capable of timely detection and response to various risk situations. The risk identification model, based on machine learning technology, can identify potential risk signals from multimodal data. The model considers information from multiple dimensions, including voice emotion, visual scene, behavioral patterns, and environmental factors. Risk level assessment employs the analytic hierarchy process (AHP) and fuzzy evaluation methods, classifying risks into four levels: low, medium, high, and urgent. Different risk levels trigger different warning and response mechanisms.

[0080] The early warning notification system supports multiple notification methods, including device alerts, SMS notifications, telephone alarms, and email reports. The system can select the appropriate notification method based on risk level and user preferences. The emergency response mechanism defines emergency handling procedures for different risk scenarios, including on-site guidance, remote support, emergency evacuation, and safety protection. The mechanism supports a combination of automated triggering and manual intervention.

[0081] The intelligent decision support module provides real-time professional guidance and decision-making suggestions to field visitors, enhancing the professionalism and effectiveness of their work.

[0082] The knowledge graph construction establishes a professional knowledge graph covering the field of post-loan management, providing a rich knowledge foundation for decision support. The business knowledge graph includes professional knowledge such as debt collection, risk assessment, laws and regulations, and operational procedures. The knowledge graph adopts ontology modeling methods to define semantic structures such as entities, relationships, and attributes.

[0083] The case knowledge base collects a large number of field visit cases and handling experiences, including success stories, lessons learned from failures, and handling of special situations. The case knowledge base supports similarity retrieval and analogical reasoning, providing references for handling new situations. The expert experience base integrates the experience and knowledge of senior field visitors and management experts, forming structured rules of experience and decision-making templates. The experience base supports dynamic updates and continuous learning.

[0084] The intelligent reasoning engine combines symbolic reasoning and neural network reasoning to achieve complex decision-making and suggestion generation. The symbolic reasoning engine, based on logical rules and knowledge graphs, enables precise logical reasoning and knowledge derivation. The engine supports multiple reasoning modes, including forward reasoning, backward reasoning, and hypothetical reasoning.

[0085] The neural reasoning engine, based on deep learning technology, is capable of handling reasoning problems with high uncertainty and ambiguity. Employing a graph neural network architecture, the engine enables end-to-end reasoning learning on a knowledge graph. Its hybrid reasoning framework organically combines symbolic and neural reasoning, fully leveraging the advantages of both methods. The framework supports advanced reasoning functions such as multi-step reasoning, counterfactual reasoning, and explanatory reasoning.

[0086] Personalized suggestion generation generates personalized decision-making suggestions and operational guidance based on specific field visit circumstances and user characteristics. The situation analysis module conducts a comprehensive analysis of the current field visit situation, including debtor characteristics, environmental conditions, historical records, and risk factors. The analysis results provide basic information for suggestion generation. The strategy matching algorithm selects the most suitable handling strategy from the strategy library for the current situation. The algorithm considers factors such as the applicability of the strategy, historical effects, and risk level, and performs a comprehensive evaluation and ranking. The suggestion generator generates specific operational suggestions and communication guidance based on the matched strategy and the current situation. The generator supports multiple suggestion types, including communication strategies, negotiation skills, risk avoidance, and evidence collection. The suggestion optimization module continuously optimizes the quality and accuracy of suggestions based on user feedback and implementation results. The module uses reinforcement learning technology, enabling it to learn and improve from experience.

[0087] Example 2 Secondly, such as Figure 2As shown, this application provides a method for external access compliance verification processing based on smart hardware, applied to the aforementioned smart hardware external access compliance verification processing system, specifically including: S1 uses the access data of smart hardware on the cloud platform as a basis to determine the power consumption data of the smart hardware using the fusion strategy for data transmission processing, and uses the power consumption data as a basis to determine the remaining power range for compliance identification processing based on the fusion strategy, and uses it as the target power range. Furthermore, the fusion strategy involves transmitting both environmental images and voice data to the cloud platform.

[0088] Furthermore, the power consumption data is determined based on the average total power consumption of the smart hardware within a unit of time when the smart hardware uses a fusion strategy for data transmission processing in history. In other words, when using the fusion strategy for data transmission processing, the power consumption of the smart hardware for data transmission processing, the power consumption for collecting environmental images and voice data, and the power consumption for local data analysis and processing by the smart hardware are all summed up to obtain the total power consumption within a unit of time.

[0089] In one possible embodiment, the unit duration is in the range of 1 minute.

[0090] Specifically, such as Figure 3 As shown, the method for determining the target power range is as follows: Based on the power consumption data, the power consumption of the smart hardware is determined; Based on the power consumption data of transmitting voice data to the cloud platform separately, determine the power consumption data of transmitting voice data to the cloud platform separately, and take it as the voice power consumption. By utilizing the data transmission duration and power consumption of the smart hardware under voice power consumption in different remaining power ranges, a target power range is determined within the remaining power range.

[0091] It should be noted that the data transmission duration is determined based on the ratio of the remaining battery range to the voice power consumption.

[0092] It is understood that, by utilizing the data transmission duration and power consumption of the smart hardware under voice power consumption within different remaining power ranges, the target power range within the remaining power range is determined, specifically including: Based on the data transmission duration under voice power consumption, determine the remaining power range where the data transmission duration exceeds a preset duration threshold. Based on the power consumption, determine the power consumption required to achieve the target transmission duration under the given power consumption. The maximum value of the endpoints of the remaining power range where the data transmission duration is greater than a preset duration threshold is taken as the minimum value of the endpoints of the target power range, and the sum of the minimum value of the endpoints of the target power range and the power consumption is taken as the maximum value of the endpoints of the target power range.

[0093] Specifically, the system analyzes the battery discharge curve and finds that when the remaining power is above 10%, the device can guarantee data transmission at a stable rate for more than 2 hours. Once the power is below 10%, the system may enter a low-power mode, the network transmission rate will decrease or become unstable, and the continuous transmission capability for 2 hours cannot be guaranteed.

[0094] Therefore, the "remaining battery range for data transmission duration greater than 2 hours" is [10%, 100%]. The maximum value at the end of this range is 100%, but we need the minimum value of 10% because it represents the threshold battery level for safe transmission. Therefore, the lower limit of the target battery range is set to 10%.

[0095] Step 2: Calculate the power required to complete the "target transmission duration". Target transmission duration: The system sets a reasonable duration for voice data that needs to be transmitted, which is 30 minutes in this example.

[0096] According to the power consumption model, the power consumption for transmitting 30 minutes of voice is 3%, the upper limit is the lower limit plus the power consumption required to complete the target transmission, and the upper limit is 10% + 3% = 13%. Finally, the calculated "target power consumption range" is [10%, 13%].

[0097] Monitoring and Triggering: During daily use, the device continuously monitors the remaining battery level. When the battery level reaches the upper limit of 13%, the system will proactively perform compliance verification to determine the compliance identification and processing method for the battery level range below 10%.

[0098] In summary, this strategy is essentially a "safe brake before the battery drains to a critical point." Through accurate battery forecasting and resource pre-allocation, it identifies and addresses violations by debt collectors before the battery level drops to a dangerous threshold, thereby significantly improving the reliability of field data collection and the system's intelligence.

[0099] Optionally, the method for determining the target power range is as follows: Based on the power consumption data, the power consumption of the smart hardware is determined; The data transmission duration is determined based on the power consumption within different remaining power ranges. By utilizing the data transmission duration of the smart hardware in different remaining power ranges, the target power range within the remaining power range is determined.

[0100] In one possible embodiment, the maximum value of the endpoint of the target power range is the maximum value of the endpoint of the remaining power range when the data transmission duration is greater than a preset duration threshold, and the minimum value of the endpoint of the target power range is the difference between the maximum value of the endpoint and the preset value. In one possible specific embodiment, the minimum value of the endpoint is the difference between the maximum value of the endpoint and 3%.

[0101] S2 determines, based on the interval between the remaining power of the smart hardware and the target power range, that the overlap risk of smart hardware in the target power range does not meet the requirements, and determines the overlapping associated smart hardware in the target power range based on the remaining power of the smart hardware. Specifically, such as Figure 4 As shown, the overlap risk of smart hardware within the target power range does not meet the requirements, specifically including: Based on the interval between the remaining battery power of the smart hardware and the target battery power range, the difference between the remaining battery power of the smart hardware and the maximum value of the endpoint of the target battery power range is determined and used as the battery power deviation value. Based on the remaining power of the smart hardware and the endpoints of the target power range, identify the overlapping associated smart hardware within the target power range. Using the overlapping smart hardware data and the power deviation value, it is determined whether the overlap risk of smart hardware in the target power range meets the requirements.

[0102] Specifically, such as Figure 5 As shown, the method for determining the overlapping associated smart hardware within the target power range is as follows: The endpoint difference is determined by the difference between the maximum and minimum values ​​at the endpoints of the target power range. The smart hardware with remaining power between the difference between the remaining power of the smart hardware and the endpoint difference, and the sum of the remaining power and the endpoint difference, is regarded as the overlapping associated smart hardware in the target power range.

[0103] In one possible specific embodiment, if the remaining power is 30% and the endpoint difference is 3%, then the smart hardware with a remaining power between 27% and 33% is considered as the overlapping associated smart hardware.

[0104] Specifically, using the overlapping smart hardware data and power deviation values, it is determined whether the overlap risk of smart hardware within the target power range meets the requirements, specifically including: Based on the power deviation value, the overlapping and associated smart hardware data of smart hardware in different power deviation value ranges are determined; Based on the overlapping smart hardware data within different power deviation ranges, determine whether the overlap risk of smart hardware within the target power range meets the requirements.

[0105] It is understandable that, based on the overlapping and correlated smart hardware data within different power deviation ranges, it is determined whether the overlap risk of smart hardware within the target power range meets the requirements, specifically including: The average number of overlapping smart hardware devices within different power deviation ranges is used to determine the power deviation range where the average number of overlapping smart hardware devices exceeds a preset threshold for the number of overlapping smart hardware devices, and this range is taken as the overlap risk range. When there are overlapping risk zones with overlapping endpoints, the number of smart devices involved is large. Therefore, the number of smart devices entering the target power range that require compliance verification is also large. In other words, if the endpoints of the overlapping risk zones also overlap, for example, if the maximum value of one overlapping risk zone coincides with the minimum value of another overlapping risk zone, then there will be a large number of smart devices requiring compliance verification within consecutive overlapping risk zones in the target power range. Therefore, it is necessary to restrict the compliance verification of smart devices that are not in the target power range to improve the efficiency and reliability of compliance verification for smart devices that are in the target power range.

[0106] To illustrate specifically, the core objective of this application is to prevent too many smart devices from entering the "target power range" simultaneously, thereby overloading the cloud-based compliance verification service. By identifying "overlapping risks," the system can preemptively restrict the verification of non-critical devices, ensuring a smooth and reliable verification process for high-priority devices (i.e., those devices currently in the target power range and urgently needing to transmit data).

[0107] Target battery range: As mentioned above, let's assume it's [10%, 13%]. This is the critical window where the device needs to prioritize voice data upload. Battery deviation value: This refers to the difference between the current remaining battery power of a smart device and the upper limit of the target battery range. The formula is: Battery deviation value = Remaining battery power - 13%.

[0108] Positive value: indicates that the battery level is above the target range, and there is still some time before the pre-upload operation needs to be performed. Negative value: indicates that the battery level has entered or fallen below the target range. Overlapping related smart hardware: refers to other smart hardware with similar battery levels to the target smart hardware. They are "potential competitors" that may enter the target range at the same time, thus creating resource competition.

[0109] Target battery level range: [10%, 13%]; Endpoint difference: 13% - 10% = 3%; Preset threshold for the number of overlapping smart devices: 100 units; Battery deviation range division: divided in 5% increments, i.e., [0%, 5%], [5%, 10%], ... We are monitoring a group of 1,000 smart devices. We are focusing on a specific smart device A, whose current remaining battery is 30%, and its battery deviation is 30% - 13% = +17%.

[0110] III. Strategy Execution Process Step 1: Identify the "overlapping associated smart hardware" for device A. Calculation rules: According to the method given in the embodiment, the overlapping associated smart hardware of device A consists of other smart hardware whose remaining battery power is within the range of [30% - 3%, 30% + 3%], i.e., within [27%, 33%]. Statistical results: After system query, there are a total of 50 smart hardware devices (including device A itself) with current battery power within the range of [27%, 33%]. These 50 devices constitute the "overlapping associated smart hardware group" of device A. Their current battery states are very similar.

[0111] Step 2: Identification of "overlapping risk zones" within the system. The system doesn't calculate for just one device; instead, it performs a global analysis of all 1000 devices. It traverses all possible power deviation ranges, calculating the average number of overlapping smart devices within each range. A key finding is that the system detects a high risk between two adjacent power deviation ranges. Risk Zone 1: Deviation range [-1%, 0%] (i.e., battery level [31%, 30%]). Devices within this range have an average of 120 overlapping smart hardware units. Risk Zone 2: Deviation range [0%, 2%] (i.e., battery level [28%, 30%]). Devices within this range have an average of 110 overlapping smart hardware units.

[0112] Risk assessment: Since both 120 and 110 are greater than the preset threshold of 100, these two intervals are marked as "overlapping risk intervals".

[0113] Risk overlap: More importantly, these two risk ranges coincide at their endpoints (the former's maximum value of 0% coincides with the latter's minimum value of 0%), forming a continuous high-risk zone of [-1%, 2%]. This means that devices with battery levels in the relatively large range of [28%, 31%] have a very high group density, making it extremely easy for them to trigger compliance verification simultaneously. Therefore, it is determined that the overlap risk of the smart eye in the target battery range does not meet the requirements.

[0114] Based on the association between the smart hardware and the target power range, and the data of overlapping smart hardware, S3 determines the compliance verification strategy of the smart hardware on the cloud platform based on the fusion strategy, and uses the processing results of the compliance verification of overlapping smart hardware under each compliance verification strategy to determine whether the target power range of the smart hardware needs to be adjusted.

[0115] Specifically, the method for determining the compliance verification strategy of the smart hardware on the cloud platform based on the fusion strategy is as follows: Based on the association between the smart hardware and the target power range, the difference between the remaining power of the smart hardware and the maximum value of the endpoint of the target power range is determined and used as the power deviation value. Based on the power deviation value, the average number of overlapping smart hardware in the power deviation value range where the power deviation value of the smart hardware is located is determined. Based on the overlapping and associated smart hardware data of the smart hardware, the number of overlapping and associated smart hardware is determined. By using the average number of overlapping and associated smart hardware devices within the power deviation value range of the smart hardware device, and the total number of overlapping and associated smart hardware devices, the compliance verification strategy of the smart hardware on the cloud platform based on the fusion strategy is determined.

[0116] Specifically, when the average number of overlapping smart hardware devices within the power deviation value range of the smart hardware is determined based on the power deviation value range of the smart hardware, and the power deviation value range of the smart hardware is determined to be within the overlap risk range, then the compliance verification strategy of the smart hardware on the cloud platform based on the fusion strategy is determined according to the number of smart hardware devices within the target power range and the benchmark verification strategy.

[0117] Additionally, it can be understood that when the average number of overlapping smart hardware devices within the power deviation range of the smart hardware's power deviation value is determined to be outside the overlap risk range, and when the power deviation value is greater than a preset deviation threshold, the remaining power of the smart hardware is too far from the target power range. Therefore, based on the number of smart hardware devices within the target power range and the preset verification strategy, a compliance verification strategy for the smart hardware on the cloud platform based on the fusion strategy is determined.

[0118] Furthermore, when the power deviation value is not greater than a preset deviation threshold, the compliance verification strategy of the smart hardware on the cloud platform based on the fusion strategy is determined according to the number of smart hardware devices within the target power range and the benchmark verification strategy.

[0119] Specifically, the method for determining the compliance verification strategy of the smart hardware on the cloud platform based on the fusion strategy is as follows: Based on the correlation between the smart hardware and the target power range, the difference between the remaining power of the smart hardware and the maximum value of the endpoint of the target power range is determined and used as the power deviation value. Based on the overlapping and associated smart hardware data of the smart hardware, the number of overlapping and associated smart hardware is determined. Using the power deviation value and the number of overlapping smart hardware devices, the compliance verification strategy of the smart hardware on the cloud platform based on the fusion strategy is determined.

[0120] It should be noted that when the power deviation value is greater than the preset deviation threshold, since the remaining power of the smart hardware is too far from the target power range, the compliance verification strategy of the smart hardware on the cloud platform based on the fusion strategy is determined according to the number of smart hardware in the target power range and the preset verification strategy.

[0121] In one possible embodiment, if the power deviation value is greater than 20%, the compliance verification strategy of the smart hardware on the cloud platform based on the fusion strategy is determined according to the number of smart hardware within the target power range and the preset verification strategy.

[0122] Additionally, it should be noted that when the power deviation value is not greater than the preset deviation threshold, it is also necessary to determine whether the power deviation value is within the overlapping risk range. If the power deviation value of the smart hardware is within the overlapping risk range, the compliance verification strategy of the smart hardware on the cloud platform based on the fusion strategy is determined according to the number of smart hardware in the target power range and the benchmark verification strategy.

[0123] Furthermore, when the power deviation value of the smart hardware is not within the overlapping risk range, it is also necessary to determine the number of overlapping associated smart hardware. When the number of overlapping associated smart hardware does not meet the requirements, that is, when the number of overlapping associated smart hardware is large, the compliance verification strategy of the smart hardware on the cloud platform based on the fusion strategy is determined according to the number of smart hardware within the target power range and the benchmark verification strategy.

[0124] In one possible embodiment, if the number of overlapping smart hardware devices is greater than 100, the compliance verification strategy of the smart hardware on the cloud platform based on the fusion strategy is determined according to the number of smart hardware devices within the target power range and the benchmark verification strategy.

[0125] Additionally, it is understood that when the number of overlapping associated smart hardware devices meets the requirements, it is necessary to further determine the number of overlapping associated smart hardware devices falling within the target power range. If the proportion of overlapping associated smart hardware devices falling within the target power range is greater than a preset proportion threshold, then it is determined that the smart hardware does not need to undergo compliance verification processing on the cloud platform. However, if the proportion of overlapping associated smart hardware devices falling within the target power range is not greater than the preset proportion threshold, then the compliance verification strategy for the smart hardware on the cloud platform based on the fusion strategy is determined according to the number of smart hardware devices within the target power range and the preset verification strategy.

[0126] In one possible specific embodiment, when the ratio of the number of overlapping associated smart hardware devices falling within the target power range to the total number of overlapping associated smart hardware devices is greater than one-half, then in order to ensure the efficiency of compliance verification processing for overlapping associated smart hardware devices, compliance verification processing does not need to be performed on the cloud platform.

[0127] It should be noted that, based on the number of smart devices within the target power range and the preset verification strategy, the compliance verification strategy for the smart devices on the cloud platform based on the fusion strategy is determined, specifically including: When the number of smart devices within the target power range is less than a first preset threshold, the smart devices are subjected to compliance verification on the cloud platform based on the fusion strategy.

[0128] It should be noted that, based on the number of smart devices within the target power range and the benchmark verification strategy, the compliance verification strategy for the smart devices on the cloud platform based on the fusion strategy is determined, specifically including: When the number of smart devices within the target power range is less than the second preset threshold, the smart devices are subjected to compliance verification on the cloud platform based on the fusion strategy.

[0129] It is understood that the first preset quantity threshold is less than the second preset quantity threshold. In one possible embodiment, the first preset quantity threshold is 30 and the second preset quantity threshold is 60.

[0130] This system analyzes the power status and device distribution characteristics of smart hardware from multiple dimensions, and establishes a dynamic cloud-based compliance verification strategy to optimize cloud resource allocation and ensure the efficiency of compliance verification for critical devices.

[0131] System parameter configuration Target power range: [10%, 13%], preset deviation threshold: 20%, threshold for the number of overlapping associated smart hardware devices: 100 units, threshold for the percentage of overlapping associated smart hardware devices falling into the target range: 50%, first preset quantity threshold: 30 units, second preset quantity threshold: 60 units; Decision-making process implementation Scenario 1: Delayed Verification Strategy for High-Power Devices Device A currently has 38% remaining battery power. Its battery deviation is calculated as 38% - 13% = 25%, which is greater than the preset deviation threshold of 20%. The system determines that this device is far from the target battery power range and belongs to the low-priority device category. At this time, the system finds that there are 25 devices currently within the target battery power range [10%, 13%], which is less than the first preset number threshold of 30. According to the preset verification strategy, the system performs compliance verification processing on the smart hardware of device A on the cloud platform based on the fusion strategy.

[0132] Scenario 2: Priority Verification Strategy for Devices in High-Risk Zones Device B has 17% remaining battery power, with a battery deviation of 4% (17%-13%), which is not greater than the preset deviation threshold. The system detects that this deviation value falls within its corresponding overlap risk range [1%, 5%]. At this time, the system queries the number of devices within the target battery power range, which is 45, less than the second preset threshold of 60. According to the benchmark verification strategy, the system initiates the "priority verification" process for device B, immediately performing a complete compliance verification on the cloud platform to ensure that the device completes verification before its battery power decreases further.

[0133] Scenario 3: Collaborative Verification Strategy for Densely Populated Devices Device C has 28% battery remaining, with a battery deviation of 15%, which is within the preset deviation threshold and not in the overlap risk range. The system detected 120 overlapping smart hardware devices (devices with battery levels in the range of [25%, 31%]) associated with Device C, exceeding the threshold requirement of 100 devices. Although Device C itself has sufficient battery, due to the high device density in its battery range, the system performs compliance verification on it in advance according to the baseline verification strategy to avoid potential verification congestion in the future. When the number of devices in the target battery range is 50, which is less than the second preset threshold of 60, cloud verification is immediately initiated.

[0134] Scenario 4: Verification-free strategy in group collaboration Device D has 14% battery remaining, with a battery deviation of 1%. The system detected 80 overlapping smart devices, meeting the requirement (≤100 devices). Further analysis revealed that 50 of these 80 overlapping smart devices have fallen into the target battery range, accounting for 62.5%, exceeding the preset threshold of 50%. This indicates that the device group to which Device D belongs has largely entered the critical battery stage. To ensure verification resources for these critical devices, the system decided to adopt a "group collaborative verification exemption" strategy for Device D, temporarily exempting it from cloud compliance verification and concentrating verification resources on those more urgently needed associated devices.

[0135] Through this refined hierarchical processing, the system can intelligently balance cloud load while ensuring compliance, ensuring that every device that needs verification can receive timely and reliable service response during critical moments of field operations.

[0136] Specifically, determining whether the target power range adjustment process for the smart hardware needs to be performed includes: Based on the processing results of compliance verification of overlapping smart hardware under various compliance verification strategies, identify overlapping smart hardware with violations and classify them as violating smart hardware. Based on different compliance verification strategies for overlapping and related smart hardware, determine the proportion of overlapping and related smart hardware that adopts the benchmark verification strategy for compliance verification. Based on the aforementioned quantity ratio and the data on illegally associated smart hardware, it is determined whether the target power range of the smart hardware needs to be adjusted.

[0137] It should be noted that when the compliance verification of the smart hardware results in a violation, it is necessary to improve the reliability of identifying the violation as much as possible. Therefore, based on this, it is determined that there is no need to adjust the target power range of the smart hardware.

[0138] Additionally, it is understandable that when the compliance verification result of the smart hardware does not show any violations, but the number of violating smart hardware does not meet the requirements, the large number of violating smart hardware may cause even non-target power range smart hardware to undergo compliance verification on the cloud platform based on the fusion strategy due to the severity of the violations. Therefore, the large number of smart hardware requiring compliance verification on the cloud platform based on the fusion strategy necessitates adjusting the target power range of the smart hardware.

[0139] Furthermore, when the number of illegally associated smart hardware devices meets the requirements, if the proportion of overlapping associated smart hardware devices that undergo compliance verification using the benchmark verification strategy is greater than a preset proportion threshold, then it is determined that the target power range of the smart hardware needs to be adjusted. If the proportion of overlapping associated smart hardware devices that undergo compliance verification using the benchmark verification strategy is not greater than the preset proportion threshold, then it is determined that the target power range of the smart hardware does not need to be adjusted.

[0140] System parameter configuration Threshold for the number of illegally associated smart hardware devices: 10 units; Threshold for the percentage of devices using preset verification strategies: 60%; Initial value for target power range: [10%, 13%] Scenario Analysis and Decision-Making Process Scenario 1: The equipment itself has violated regulations. After device A completes compliance verification, the system detects a data encryption violation. According to the rules, when a smart hardware device's own compliance verification result shows a violation, the system must prioritize ensuring the accuracy of violation identification. In this case, the system maintains the original target battery range [10%, 13%]. This ensures that device A completes full data transmission and in-depth compliance testing within the optimal battery range, avoiding any impact on the accurate identification and handling of violations due to adjustments in the battery range.

[0141] Scenario 2: Exceeding the limit for the number of illegally associated smart hardware devices Device B passed its own compliance verification, but a significant number of non-compliant devices were found in its overlapping smart hardware group. Statistics showed that among Device B's 80 overlapping smart hardware devices, 15 were non-compliant, exceeding the threshold of 10. This indicates a high risk of non-compliance within Device B's device group. The system determined that even if these devices were not within the target power range, they still required strict compliance verification, which would significantly increase the cloud verification load. To balance system pressure, the system decided to adjust Device B's target power range from [10%, 13%] to [12%, 15%] for pre-verification, thereby reducing system processing load.

[0142] Scenario 3: Preset verification strategy is used too frequently. Device C passed its own verification, with only 8 illegally associated smart hardware devices (meeting the requirement of ≤10 devices). However, analysis showed that among its overlapping smart hardware devices, 75% used the baseline verification strategy (i.e., the priority verification strategy), exceeding the 60% threshold. This indicates a high dependence of this group on priority verification resources. To avoid excessive concentration of verification resources, the system decided to adjust the target battery range for device C from [10%, 13%] to [12%, 15%] for early verification.

[0143] Scenario 4: Maintaining the status quo and achieving stability Device D passed its self-verification, with only 5 illegally associated smart hardware devices, and 45% of the devices using the benchmark verification strategy met the requirements (number of illegal devices ≤ 10, percentage of devices using the preset strategy ≤ 60%). The system determined that the current state was stable and that there was no need to adjust the target power range, maintaining the original [10%, 13%] range setting.

[0144] This dynamic adjustment mechanism based on compliance verification results enables the system to intelligently optimize resource allocation according to the actual security situation, ensuring the rigor of compliance verification while improving the overall operational efficiency of the system. By continuously learning from the adjustment results, the system will also establish more precise adjustment strategies, forming a closed-loop management system for continuous optimization.

[0145] Optionally, determining whether to adjust the target power range of the smart hardware specifically includes: Based on the processing results of compliance verification of overlapping smart hardware under various compliance verification strategies, identify overlapping smart hardware with violations and classify them as violating smart hardware. Based on the different compliance verification strategies for overlapping and related smart hardware, determine the overlapping and related smart hardware data that adopt different compliance verification strategies; Based on overlapping and non-compliant smart hardware data using different compliance verification strategies, it is determined whether the target power range of the smart hardware needs to be adjusted.

[0146] It should be noted that when the compliance verification of the smart hardware results in a violation, it is necessary to improve the reliability of identifying the violation as much as possible. Therefore, based on this, it is determined that there is no need to adjust the target power range of the smart hardware.

[0147] Additionally, it is understandable that when the compliance verification result of the smart hardware does not show any violations, it is necessary to determine the number of times the smart hardware has undergone compliance verification on the cloud platform based on the fusion strategy. When the number of times the smart hardware has undergone compliance verification on the cloud platform based on the fusion strategy does not meet the requirements, it is determined that no adjustment of the target power range of the smart hardware is required because the number of compliance verifications on the cloud platform is not high.

[0148] In one possible specific embodiment, if the number of compliance verification processes performed by the smart hardware on the cloud platform based on the fusion strategy is less than 50, it is determined that no adjustment process for the target power range of the smart hardware is required.

[0149] Furthermore, when the number of compliance verification processes for the smart hardware on the cloud platform based on the fusion strategy meets the requirements, it is also necessary to identify the smart hardware associated with violations. If the number of illegally associated smart hardware does not meet the requirements, due to the large number of illegally associated smart hardware, even if the illegally associated smart hardware is not in the target power range, it may still need to undergo compliance verification processing on the cloud platform based on the fusion strategy because the violation is relatively serious. Therefore, since the number of smart hardware that needs to undergo compliance verification processing on the cloud platform based on the fusion strategy is large, it is determined that the target power range of the smart hardware needs to be adjusted.

[0150] In one possible specific embodiment, when the number of illegally associated smart hardware devices exceeds 20, it is determined that the target power range of the smart hardware needs to be adjusted. If the number of violations detected by the smart hardware on the cloud platform exceeds 5, compliance verification processing based on the fusion strategy is still required on the cloud platform even if the device is not within the target power range. However, if the number of violations detected on the cloud platform is no more than 5, and the device is not within the target power range, compliance verification processing based on the fusion strategy is not required on the cloud platform. Instead, compliance verification processing using voice data on the cloud platform is only required according to a preset period. In one possible embodiment, the preset period is 1 minute.

[0151] Specifically, when it is necessary to adjust the target power range of the smart hardware, the maximum value of the endpoint of the target power range of the smart hardware is adjusted to the middle value of the original target power range. That is, the average of the maximum and minimum values ​​of the original endpoints is used as the maximum value of the endpoint of the adjusted target power range, while the minimum value of the endpoint remains the minimum value of the endpoint of the original target power range.

[0152] Additionally, it should be noted that when the number of illegally associated smart hardware devices meets the requirements, it is necessary to further determine the average number of verifications for overlapping associated smart hardware devices. When the average number of verifications for overlapping associated smart hardware devices does not meet the requirements, it is necessary to adjust the target power range of the smart hardware.

[0153] In one possible specific embodiment, when the average number of compliance verification processes performed on the cloud platform based on the fusion strategy for the overlapping smart hardware is less than 30, it is determined that the target power range adjustment process for the smart hardware needs to be performed.

[0154] Furthermore, it is understood that when the average number of verifications meets the requirements, it is necessary to further determine the compliance verification strategy for overlapping smart hardware. If the proportion of overlapping smart hardware that undergoes compliance verification using the benchmark verification strategy is greater than a preset proportion threshold, it is determined that the target power range of the smart hardware needs to be adjusted. If the proportion of overlapping smart hardware that undergoes compliance verification using the benchmark verification strategy is not greater than the preset proportion threshold, it is determined that the target power range of the smart hardware does not need to be adjusted.

[0155] In one possible specific embodiment, when the proportion of overlapping associated smart hardware that undergoes compliance verification using a benchmark verification strategy is greater than two-thirds, it is determined that the target power range of the smart hardware needs to be adjusted.

[0156] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and non-volatile computer storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0157] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0158] The above description is merely one or more embodiments of this specification and is not intended to limit this specification. Various modifications and variations can be made to the one or more embodiments of this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of one or more embodiments of this specification should be included within the scope of the claims of this specification.

Claims

1. An intelligent hardware based out-visit compliance verification processing system, characterized in that, Specifically comprising: The intelligent hardware data acquisition module is responsible for realizing real-time acquisition and analysis processing of multi-modal data, the multi-modal perception module is responsible for realizing perception and understanding of the environment, the semantic analysis module is responsible for realizing semantic analysis and understanding of voice data, visual scene and behavior action in the external visit process, the compliance verification module is responsible for realizing compliance inspection and risk assessment of the external visit behavior based on compliance rules through machine learning algorithm, and the cloud platform is responsible for data storage of the intelligent hardware, analysis and processing of illegal behavior, and transmission back to the intelligent hardware, early warning and assistance to the external visit process.

2. The smart hardware based out-visit compliance verification processing system of claim 1, wherein, The multi-modal data includes audio, text, video and image collected by the intelligent hardware at the same time.

3. The smart hardware-based field visit real-time compliance early warning processing method and system according to claim 1, characterized in that, The cloud platform determines the behavior and multi-lingual rhetoric that the external visit personnel should take based on the analysis result of multi-dimensional data of the multi-modal large model, and transmits back to the intelligent hardware in real time, and performs real-time early warning and assistance to the external visit process.

4. The smart hardware-based field visit real-time compliance early warning processing method and system according to claim 3, characterized in that, The multi-lingual rhetoric is a strategy supporting real-time rhetoric generation in Chinese and English.

5. The smart hardware-based field visit real-time compliance early warning processing method and system according to claim 1, characterized in that, The early warning to the external visit process specifically includes: Based on the dialogue data of the collector and the visited person, the repayment intention and emotion of the visited person are analyzed in real time; Based on the facial expression and action picture of the visited person, whether the visited person has abnormal behavior is analyzed.

6. The smart hardware based out-visit compliance verification processing system of claim 1, wherein, It also includes a security module: including data encryption, identity authentication, access control, audit log function, responsible for data security management of the intelligent hardware and the cloud platform. 7.A method for smart hardware-based compliance verification of an external visit, applied to the smart hardware-based compliance verification system of any one of claims 1-6, characterized in that, Specifically comprising: Based on the data accessed by the cloud platform of the intelligent hardware, the power consumption data of the intelligent hardware using the fusion strategy for data transmission processing is determined, and the residual power interval for compliance recognition processing based on the fusion strategy is determined based on the power consumption data, and the residual power interval is used as the target power interval; When the overlap risk of the target power interval of the intelligent hardware does not meet the requirements, the overlap associated hardware of the intelligent hardware in the target power interval is determined according to the residual power of the intelligent hardware based on the interval of the residual power of the intelligent hardware and the target power interval; Based on the association between the intelligent hardware and the target power interval and the associated hardware data of the intelligent hardware, the compliance verification strategy of the intelligent hardware based on the fusion strategy in the cloud platform is determined, and whether the target power interval of the intelligent hardware needs to be adjusted is determined by using the processing result of the compliance verification of each compliance verification strategy. 8.The smart hardware-based out-visit compliance verification processing method of claim 7, wherein, The fusion strategy is a strategy of transmitting environmental image and voice data to the cloud platform. 9.The smart hardware-based out-visit compliance verification processing method of claim 7, wherein, The method for determining the target power interval of the residual power interval is: Based on the power consumption data, the power consumption of the intelligent hardware is determined; Based on the power consumption, the data transmission time length in different residual power intervals is determined; Using the data transmission time length of the intelligent hardware in different residual power intervals, the target power interval in the residual power interval is determined. 10.The smart hardware-based out-visit compliance verification processing method of claim 7, wherein, The method for the cloud platform to generate compliance verification strategy based on the fusion strategy of the intelligent hardware is: The smart hardware is used to determine the transmission priority of the compliance verification strategy generated by the cloud platform based on the fusion strategy, and the strategy with higher priority is transmitted to the smart hardware end based on the target power interval in the remaining power interval.

Images