Intelligent identification algorithm model edge-end fusion type deployment method

Abstract

The invention relates to the field of edge computing, and discloses an intelligent recognition algorithm model edge-end fusion type deployment method, which comprises the following steps: collecting original data from different sensors, acquiring data from various sensors in real time by edge equipment, and carrying out primary processing and screening on the data; the edge device operates a lightweight intelligent recognition algorithm, processes the data in real time and makes a preliminary judgment; an intelligent task scheduling system is designed, and the processing requirement of each task is evaluated in real time; when the edge equipment cannot independently complete a calculation task, data is transmitted to the cloud, and the cloud processes a complex intelligent identification task by using powerful calculation resources to generate a high-level reasoning result; and the cloud transmits the optimized model or processing result back to the edge device, and the edge device performs adaptive adjustment according to the feedback of the cloud to optimize the local processing capability. The method has the advantages of improving the response speed and the computing power.

Description

Technical Field

[0001] The present invention relates to the field of edge computing, and specifically to an edge fusion deployment method for an intelligent recognition algorithm model. Background Art

[0002] In current intelligent recognition applications, with the advancement of deep learning and big data analysis technologies, more and more industries are beginning to rely on computer vision and artificial intelligence for real-time monitoring and data analysis. The traditional cloud-based centralized computing model has problems such as large transmission delay, high bandwidth consumption, and high computing pressure when processing real-time video surveillance and dynamic sensor data, especially in high-frequency data collection and analysis scenarios. Edge computing has emerged as the times require. It solves the problems of high data transmission delay and bandwidth bottleneck by distributing data processing tasks to edge devices closer to the data source, and can provide higher response speed and computing efficiency in scenarios with high real-time requirements. However, the computing power of edge computing is limited and cannot meet the needs of all complex computing tasks. Therefore, it is necessary to design an edge-end fusion deployment method for intelligent recognition algorithm models that improves response speed and computing power. Summary of the invention

[0003] 1. Technical issues to be resolved

[0004] In view of the shortcomings of the prior art, the present invention provides an edge fusion deployment method for an intelligent recognition algorithm model, which has the advantages of improving response speed and computing power, and solves the problems in the above-mentioned background technology.

[0005] (II) Technical solution

[0006] In order to achieve the above-mentioned purpose of improving response speed and computing power, the present invention provides the following technical solution: a method for edge fusion deployment of an intelligent recognition algorithm model, comprising the following steps:

[0007] S1: Collect raw data from different sensors. Edge devices obtain data from various sensors in real time and perform preliminary processing and screening on the data.

[0008] Preferably, S1 further includes collecting raw data from different sensors, and the edge device obtains data from various sensors in real time, including video surveillance, temperature and humidity sensors, accelerometers, and ambient light sensors, and performs preliminary processing and screening on these data.

[0009] S2: The edge device runs a lightweight intelligent recognition algorithm to process data in real time and make preliminary judgments;

[0010] Preferably, S2 further includes applying a lightweight pattern recognition algorithm to perform real-time classification or prediction on the preprocessed data. The edge device classifies and predicts the data in real time, sets rules and thresholds, and determines whether to trigger a specific event or alarm based on the feature data. For data from multiple sensors, data fusion is performed to integrate information from different sensors. Through the fusion of multimodal information, the edge device can make judgments based on data from different sensors. Based on the recognition results, the edge device makes decisions and takes responses locally.

[0011] S3: Design an intelligent task scheduling system to evaluate the processing requirements of each task in real time;

[0012] Preferably, the S3 further includes dynamically evaluating the priority of each task according to the urgency of the task, resource consumption and predetermined time limit, predicting the computing resources and execution time required for the task according to the task type and historical data of the current task, predicting the task requirements using machine learning algorithms, analyzing the dependencies between tasks, analyzing through topological sorting, data flow graphs or task dependency graphs, dynamically adjusting resource allocation according to the task priority, resource requirements, dependencies and the current resource status of the system, achieving load balancing among multiple computing nodes or devices, and designing fault-tolerant mechanisms.

[0013] S4: When edge devices cannot complete computing tasks alone, data will be transmitted to the cloud, which uses powerful computing resources to process complex intelligent recognition tasks and generate high-level reasoning results;

[0014] Preferably, the S4 further includes receiving and storing data from edge devices in the cloud, decompressing and decrypting the data, and storing the data in a distributed storage system. In the cloud, complex intelligent recognition tasks will be further decomposed into multiple subtasks and scheduled according to resource requirements and priorities. The scheduling system will analyze the current load of cloud computing resources and allocate corresponding computing resources to each subtask. The cloud uses intelligent algorithms to perform in-depth analysis and reasoning on the data and process data analysis tasks.

[0015] S5: The cloud transmits the optimized model or processing results back to the edge device, and the edge device makes adaptive adjustments based on the feedback from the cloud to optimize local processing capabilities.

[0016] Preferably, S5 further includes the edge device receiving the optimization model or processing results from the cloud, the device obtaining the inference results and updated model parameters from the cloud through the data transmission protocol, and the edge device updating the local intelligent recognition model according to the cloud feedback, including adjusting the parameters of the algorithm, introducing new features, and optimizing the inference method. According to the cloud feedback, the edge device adaptively adjusts the local processing algorithm.

[0017] (III) Beneficial effects

[0018] Compared with the prior art, the present invention provides an edge-end fusion deployment method for an intelligent recognition algorithm model, which has the following beneficial effects:

[0019] The present invention runs a lightweight intelligent recognition algorithm on the edge device to process the data in real time and make preliminary judgments; designs an intelligent task scheduling system to evaluate the processing requirements of each task in real time; when the edge device cannot complete the computing task alone, the data will be transmitted to the cloud, and the cloud will use powerful computing resources to process complex intelligent recognition tasks and generate high-level reasoning results; the preliminary data processing and intelligent recognition tasks are assigned to the edge device for real-time response, and the complex computing tasks are deeply processed in the cloud, thereby achieving efficient data transmission, reducing latency, optimizing computing power, and improving the overall performance of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 Schematic diagram of the method of the present invention. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] Example 1

[0023] The present invention provides a technical solution: an edge fusion deployment method of an intelligent recognition algorithm model, comprising the following steps:

[0024] S1: Collect raw data from different sensors. Edge devices obtain data from various sensors in real time and perform preliminary processing and screening on the data.

[0025] Collecting raw data from different sensors, edge devices obtain data from various sensors in real time, including video surveillance, temperature and humidity sensors, accelerometers, and ambient light sensors, and then perform preliminary processing and screening on these data. This process further includes data synchronization and timestamp marking to ensure the temporal consistency of multi-source data; data cleaning to improve data quality by removing invalid, redundant, erroneous or abnormal data points; data formatting to unify data from different sensors into a standardized format for subsequent processing; and data compression to reduce data transmission volume and bandwidth consumption. At the same time, edge devices will also perform preliminary feature extraction and denoising processing to extract valuable information from the raw data, eliminate interference from environmental noise, and ensure the efficiency and accuracy of subsequent intelligent recognition tasks. All of these processes are completed locally on the edge device to ensure efficient data processing and real-time response.

[0026] S2: The edge device runs a lightweight intelligent recognition algorithm to process data in real time and make preliminary judgments;

[0027] Applying lightweight pattern recognition algorithms to classify or predict preprocessed data in real time, edge devices can classify and predict data in real time, such as identifying abnormal behavior in videos, determining whether sensor data exceeds standards, etc. This step ensures that the system can maintain real-time performance and respond quickly when processing high-frequency data.

[0028] Set rules and thresholds to determine whether to trigger specific events or alarms based on feature data. For example, if sensor data exceeds a certain threshold, an alarm is triggered, or the behavior is determined according to preset rules to determine whether it meets expectations. By setting rules and thresholds, it is ensured that when data meets certain standards, edge devices can automatically make judgments and take immediate actions, such as starting an alarm or triggering a system response. This step enhances the automation capability of real-time decision-making.

[0029] Data fusion is performed on data from multiple sensors to integrate information from different sensors to improve the accuracy and reliability of judgments. Through the fusion of multimodal information, edge devices can make more comprehensive judgments based on data from different sensors. For example, combining video images and temperature and humidity sensor data can more accurately identify fires, equipment failures, or other emergencies.

[0030] Based on the recognition results, the edge device makes decisions locally and takes appropriate responses. For example, for abnormal detection results, the edge device can immediately start the alarm system or control the equipment for repair. Through local decision-making and response, the edge device can respond instantly when the data is processed locally, reducing data transmission and system response time, improving system efficiency, and providing instant feedback at critical moments.

[0031] S3: Design an intelligent task scheduling system to evaluate the processing requirements of each task in real time;

[0032] Dynamically evaluate the priority of each task based on its urgency, resource consumption, and scheduled deadline. Priority can be determined based on factors such as the time sensitivity of the task, computational complexity, and business importance. Ensure that high-priority tasks are processed in a timely manner to prevent low-priority tasks from affecting the system's response time. Through accurate priority evaluation, the real-time performance and efficiency of the system are improved to ensure that critical tasks are not delayed.

[0033] Predict the computing resources and execution time required for tasks based on the task type and historical data of the current task. Use machine learning algorithms to predict task requirements. By accurately predicting the resource requirements of each task, avoid resource waste and over-allocation, ensure that tasks can be executed under the optimal resource configuration, improve overall resource utilization, and reduce delays and bottlenecks.

[0034] Analyze the dependencies between tasks to determine which tasks can be executed in parallel and which need to wait for other tasks to complete. This can be done through topological sorting, data flow graphs, or task dependency graphs. Effectively identify the dependencies between tasks to ensure that the correct execution order is followed during processing. By optimizing the task scheduling process, the task parallelism can be improved and the total execution time can be shortened.

[0035] Dynamically adjust resource allocation based on task priority, resource requirements, dependencies, and the current resource status of the system. For example, use scheduling algorithms such as FIFO, Round Robin, and Shortest Job First to schedule tasks. Through dynamic resource allocation, the system can respond to changing task requirements in real time, improve task processing efficiency, reduce idle resources and bottlenecks, and ensure system load balance.

[0036] Load balancing is achieved between multiple computing nodes or devices to avoid overloading of a single node. At the same time, a fault-tolerant mechanism is designed to ensure that tasks can continue to be executed on other nodes when a node fails. Load balancing improves the processing power of the system and avoids performance degradation caused by resource overload. The fault-tolerant mechanism ensures high availability of the system, improves the stability and reliability of the system, and ensures the smooth completion of tasks.

[0037] S4: When edge devices cannot complete computing tasks alone, data will be transmitted to the cloud, which uses powerful computing resources to process complex intelligent recognition tasks and generate high-level reasoning results;

[0038] The cloud receives and stores data from edge devices, and decompresses and decrypts the data. The data is stored in a distributed storage system for quick access and processing. The distributed storage system ensures that data can be accessed quickly, has high availability and redundant storage, and prevents data loss.

[0039] In the cloud, complex intelligent recognition tasks are further decomposed into multiple subtasks and scheduled according to resource requirements and priorities. The scheduling system analyzes the current load of cloud computing resources and allocates corresponding computing resources to each subtask. Through task decomposition and scheduling, the cloud can flexibly assign tasks to multiple computing nodes for parallel processing, improving the parallelism and efficiency of task processing. At the same time, it optimizes resource utilization and ensures efficient load balancing.

[0040] The cloud uses deep learning, neural networks or other intelligent algorithms to perform in-depth analysis and reasoning on data, processing tasks such as image recognition, speech recognition or complex data analysis. The powerful computing resources of the cloud can process more complex algorithms and massive data. The cloud has powerful computing capabilities and can handle tasks that require a large amount of computing resources and deep learning, and generate high-level reasoning results. This allows edge devices to rely on the cloud for complex decision-making and advanced recognition tasks without leaving all the computing burden to the edge devices.

[0041] In the cloud, the inference results of intelligent recognition will be further optimized and enhanced, for example, by updating the model, tuning parameters or multimodal fusion technology to improve the accuracy and robustness of the results. Through the inference optimization of the cloud computing power, the recognition results are more accurate and reliable, and can continuously adapt to new environments and new data, improving the intelligence and adaptability of the system.

[0042] S5: The cloud transmits the optimized model or processing results back to the edge device, and the edge device makes adaptive adjustments based on the feedback from the cloud to optimize local processing capabilities.

[0043] The edge device receives the optimization model or processing results from the cloud. The device obtains the inference results and updated model parameters from the cloud through the data transmission protocol. This ensures that the edge device can obtain the latest cloud optimization information, thereby avoiding lagging or outdated processing strategies that affect the performance of the device and enhancing the collaboration between the device and the cloud.

[0044] The edge device updates the local intelligent recognition model based on cloud feedback, including adjusting algorithm parameters, introducing new features, and optimizing reasoning methods. The edge device will load the new model or algorithm locally and merge or replace it with the existing model. Through local model updates, the edge device can continuously learn and adapt to new data patterns or recognition tasks, thereby improving its local processing capabilities and recognition accuracy. This step improves the adaptability and intelligence of the device and reduces its dependence on the cloud.

[0045] Based on feedback from the cloud, edge devices adaptively adjust local processing algorithms. For example, the device may adjust the complexity of the algorithm to adapt to different hardware resources, or select a lighter algorithm for reasoning when computing resources are insufficient. Adaptive algorithm adjustment enables edge devices to optimize algorithm operation in different computing environments, ensuring that good performance can be maintained even when resources are limited, thereby improving local processing efficiency and response speed.

[0046] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0047] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for edge fusion deployment of an intelligent recognition algorithm model, characterized in that: The following steps are involved: S1: Collect raw data from different sensors. Edge devices obtain data from various sensors in real time and perform preliminary processing and screening on the data. S2: The edge device runs a lightweight intelligent recognition algorithm to process data in real time and make preliminary judgments; S3: Design an intelligent task scheduling system to evaluate the processing requirements of each task in real time; S4: When edge devices cannot complete computing tasks alone, data will be transmitted to the cloud, which uses powerful computing resources to process complex intelligent recognition tasks and generate high-level reasoning results; S5: The cloud transmits the optimized model or processing results back to the edge device, and the edge device makes adaptive adjustments based on the feedback from the cloud to optimize local processing capabilities.

2. According to claim 1, the edge fusion deployment method of the intelligent recognition algorithm model is characterized in that: The S1 further includes collecting raw data from different sensors. The edge device obtains data from various sensors in real time, including video surveillance, temperature and humidity sensors, accelerometers, and ambient light sensors, and performs preliminary processing and screening on these data.

3. According to claim 1, the edge fusion deployment method of the intelligent recognition algorithm model is characterized in that: The S2 further includes applying a lightweight pattern recognition algorithm to classify or predict the preprocessed data in real time. The edge device classifies and predicts the data in real time, sets rules and thresholds, and determines whether to trigger a specific event or alarm based on the feature data. For data from multiple sensors, data fusion is performed to integrate information from different sensors. Through the fusion of multimodal information, the edge device can make judgments based on data from different sensors. Based on the recognition results, the edge device makes decisions and takes responses locally.

4. According to claim 1, the edge fusion deployment method of the intelligent recognition algorithm model is characterized in that: The S3 further includes dynamically evaluating the priority of each task according to the urgency of the task, resource consumption and scheduled time limit, predicting the computing resources and execution time required for the task according to the task type and historical data of the current task, using machine learning algorithms to predict task requirements, analyzing the dependencies between tasks, analyzing through topological sorting, data flow graphs or task dependency graphs, dynamically adjusting resource allocation according to the task priority, resource requirements, dependencies and the current resource status of the system, achieving load balancing among multiple computing nodes or devices, and designing fault-tolerant mechanisms.

5. According to claim 1, the edge fusion deployment method of the intelligent recognition algorithm model is characterized in that: The S4 further includes receiving and storing data from edge devices in the cloud, decompressing and decrypting the data, and storing the data in a distributed storage system. In the cloud, complex intelligent recognition tasks will be further decomposed into multiple subtasks and scheduled according to resource requirements and priorities. The scheduling system will analyze the current load of cloud computing resources and allocate corresponding computing resources to each subtask. The cloud uses intelligent algorithms to perform in-depth analysis and reasoning on the data and process data analysis tasks.

6. According to claim 1, the edge fusion deployment method of the intelligent recognition algorithm model is characterized in that: The S5 further includes the edge device receiving the optimization model or processing result from the cloud, the device obtaining the inference result and updated model parameters from the cloud through the data transmission protocol, and the edge device updating the local intelligent recognition model according to the cloud feedback, including adjusting the algorithm parameters, introducing new features, and optimizing the inference method. According to the cloud feedback, the edge device adaptively adjusts the local processing algorithm.

Images