Video stream based log generator design

Abstract

The invention discloses a log generator design based on video streaming in the technical field of log generation, comprising an end side camera device, an end side data acquisition system, a cloud large model, an end side log recording and displaying system, a data transmission system and a cloud storage module, the end side data acquisition system comprises a coprocessor and an event generator; the end-side camera equipment sends video stream data to the end-side data acquisition system, and the video stream data is processed by the coprocessor and the event generator to generate a LOG file; the cloud big model carries out processing and analysis through a deep learning algorithm, a natural language processing algorithm, a reinforcement learning algorithm and the like, and inputs an analysis result into the end side log recording and displaying system, the log generator based on the video stream is reasonable in design and structural design, can realize log recording with low storage cost and high retrieval efficiency, and is suitable for large-scale popularization and application. And the later operation and maintenance requirements of various intelligent video applications are met.

Description

Technical Field

[0001] The present invention relates to the technical field of log generation, and specifically to the design of a log generator based on video streams. Background Technique

[0002] With the rapid development of computer technology and big data, deep learning has made progress in various fields. In order to further improve the performance of computing models, large models with tens of millions or even hundreds of millions of parameters have received extensive attention in the industry. Large models receive data, learn to improve their own capabilities, and execute various tasks to generate useful results. It is speculated that in the next few years, the market scale of Chinese AI large models will reach tens of billions of US dollars, and artificial intelligence will enter a critical period of large-scale implementation and application.

[0003] However, as the computing volume of AI applications gradually increases, whether on the edge side or the cloud side, the data storage cost also increases accordingly. Completing accurate data retrieval in massive data is challenging, especially for video data with a large storage space. Traditional video information storage and retrieval mainly record and store the complete video stream in the cloud and then search one by one according to the scope of information, which has the dilemmas of unclear search scope, high labor and time costs. However, with the wide application of more and more intelligent videos, how to reduce the storage cost of video streams and improve the retrieval efficiency has become the focus of current research. For this reason, we propose the design of a log generator based on video streams. Summary of the Invention

[0004] The purpose of the present invention is to provide the design of a log generator based on video streams to solve the problems raised in the above background technique.

[0005] To achieve the above purpose, the present invention provides the following technical solution: The design of a log generator based on video streams includes an edge-side camera device, an edge-side data acquisition system, a cloud large model, an edge-side log recording and display system, a data transmission system, and a cloud storage module. The edge-side data acquisition system includes a coprocessor and an event generator;

[0006] The edge-side camera device sends video stream data to the edge-side data acquisition system, and processes it through the coprocessor and the event generator to generate a LOG file;

[0007] The cloud large model processes and analyzes through deep learning algorithms, natural language processing algorithms, reinforcement learning algorithms, etc., and inputs the analysis results into the edge-side log recording and display system;

[0008] The edge-side log recording and display system records the analysis results and determines whether they are key information. If they are key information, it outputs data to the data transmission system; if they are not key information, it performs edge-side filtering;

[0009] The described data transmission system uploads the output key information to the cloud storage module.

[0010] Preferably, the edge - side data acquisition system is configured with one or more cameras, and an artificial intelligence chip is built into the device to achieve the ability of local intelligent data screening, extract key information from the video stream data, complete the edge - side video data filtering task, with the edge - side device resolution reaching 1080P and the speed reaching 10fps.

[0011] Preferably, based on the data obtained by the acquisition system, the general cloud - side large model is used for recognition and analysis to achieve video - to - text conversion, generate a relatively coarse - grained text record, record it in sequence according to time, and at the same time mark the key data at key nodes, and record it in the edge - side log record and display system.

[0012] Preferably, the key data is uploaded to the cloud storage module through the data transmission system, and the remaining large amount of non - key data is filtered out by the edge - side.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: The design of the log generator based on the video stream, through the innovative edge - side and cloud - side collaborative video log generator, realizes the deployment and application of the large model, optimizes the efficiency of video backtracking; the edge - side device intelligently filters data, thus greatly reducing the storage cost of the cloud - side. At the same time, it can capture user needs in real - time, play a role in timely deviation correction and warning, solve the operation and maintenance problems of intelligent video applications, and pioneeringly realize the intelligent application of the large model. It is widely applicable to various intelligent video application scenarios with high requirements for storage space and operation and maintenance efficiency, and can achieve log records with low storage cost and high retrieval efficiency, meeting the later operation and maintenance requirements of various intelligent video applications. Brief Description of the Drawings

[0014] Figure 1 It is the system flow chart of the log generator design based on the video stream provided by the embodiment of the present invention. Detailed Embodiments

[0015] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0016] Embodiment 1:

[0017] Please refer to Figure 1, the present invention provides a technical solution: the design of a log generator based on video streams, including an end-side camera device, an end-side data acquisition system, a cloud large model, an end-side log recording and display system, a data transmission system, and a cloud storage module. The end-side data acquisition system includes a coprocessor and an event generator;

[0018] The end-side camera device sends video stream data to the end-side data acquisition system, and processes it through the coprocessor and the event generator to generate a LOG file;

[0019] The cloud large model processes and analyzes through deep learning algorithms, natural language processing algorithms, reinforcement learning algorithms, etc., and inputs the analysis results into the end-side log recording and display system;

[0020] The end-side log recording and display system records the analysis results and determines whether they are key information. If they are key information, it outputs the data to the data transmission system; if they are not key information, it performs end-side filtering;

[0021] The data transmission system uploads the output key information to the cloud storage module.

[0022] Through the innovative video log generator with end-side and cloud-side collaboration, the deployment and application of the large model are realized, optimizing the efficiency of video backtracking; the end-side device intelligently filters data, thus greatly reducing the storage cost on the cloud side, and at the same time can capture user needs in real time, playing a role in timely correction and warning.

[0023] Embodiment 2:

[0024] Please refer to Figure 1 , the present invention provides a technical solution: the design of a log generator based on video streams, including an end-side camera device, an end-side data acquisition system, a cloud large model, an end-side log recording and display system, a data transmission system, and a cloud storage module. The end-side data acquisition system includes a coprocessor and an event generator;

[0025] The end-side camera device sends video stream data to the end-side data acquisition system, and processes it through the coprocessor and the event generator to generate a LOG file;

[0026] The cloud large model processes and analyzes through deep learning algorithms, natural language processing algorithms, reinforcement learning algorithms, etc., and inputs the analysis results into the end-side log recording and display system;

[0027] The end-side log recording and display system records the analysis results and determines whether they are key information. If they are key information, it outputs the data to the data transmission system; if they are not key information, it performs end-side filtering;

[0028] The data transmission system uploads the output key information to the cloud storage module.

[0029] The end - side data acquisition system is configured with one or more cameras, and an artificial intelligence chip is built into the device to achieve the ability of local intelligent data screening, extract key information from the video stream data, complete the end - side video data filtering task, with the resolution of the end - side device reaching 1080P and the speed reaching 10fps.

[0030] Based on the data obtained by the acquisition system, the general cloud large - model conducts identification and analysis to achieve video - to - text conversion, generate a relatively coarse - grained text record, record it sequentially according to time, and at the same time mark the key data at key nodes, and record it in the end - side log recording and display system.

[0031] The key data is uploaded to the cloud storage module through the data transmission system, and the remaining large amount of non - key data is filtered out by the end - side.

[0032] The cloud side completes the storage of the key video stream data, and through identification and analysis, can quickly complete the query of a specific event.

[0033] Solve the operation and maintenance problems of intelligent video applications, pioneeringly realize the intelligent application of the large - model, widely applicable to various intelligent video application scenarios with high requirements for storage space and operation and maintenance efficiency, can achieve log records with low storage costs and high retrieval efficiency, and meet the requirements of the later operation and maintenance of various intelligent video applications.

[0034] Embodiment 3:

[0035] Please refer to Figure 1 , the present invention provides a technical solution: the design of a log generator based on video streams, including an end - side camera device, an end - side data acquisition system, a cloud large - model, an end - side log recording and display system, a data transmission system, and a cloud storage module. The end - side data acquisition system includes a coprocessor and an event generator;

[0036] The end - side camera device sends the video stream data to the end - side data acquisition system, and processes it through the coprocessor and the event generator to generate a LOG file;

[0037] The cloud large - model processes and analyzes through deep - learning algorithms, natural - language processing algorithms, reinforcement - learning algorithms, etc., and inputs the analysis results into the end - side log recording and display system;

[0038] The end - side log recording and display system records the analysis results and determines whether it is key information. If it is key information, it outputs data to the data transmission system; if it is not key information, it conducts end - side filtering;

[0039] The described data transmission system uploads the output key information to the cloud storage module.

[0040] The edge - side data acquisition system is configured with one or more cameras. An artificial intelligence chip is built into the device to achieve the ability of local intelligent data screening, extract key information from the video stream data, complete the edge - side video data filtering task. The resolution of the edge - side device reaches 1080P and the speed reaches 10fps.

[0041] Based on the data obtained by the acquisition system, the general cloud - side large model performs recognition and analysis to achieve video - to - text conversion, generate a relatively coarse - grained text record, record it in sequence according to time, and at the same time mark the key data at key nodes, and record it in the edge - side log record and display system.

[0042] The key data is uploaded to the cloud storage module through the data transmission system, and the remaining large amount of non - key data is filtered out by the edge - side.

[0043] The cloud - side completes the storage of the key video stream data, and through recognition and analysis, can quickly complete the query of a specific event.

[0044] By combining the edge - side data acquisition system and the cloud - side large model, coordinating the intelligent data processing capabilities of the edge - side and the cloud - side, generating text log records of the key data of the video stream of the intelligent application, selectively uploading to the cloud, while greatly saving storage costs and significantly improving the video retrieval efficiency. The present invention aims to solve the problem of high cost of "log processing + storage + query analysis" of video data in intelligent applications through the application of the large model in intelligent terminals, thereby further improving and optimizing the operation and maintenance capabilities of intelligent applications, and can be applied to intelligent application scenarios that require large - scale video data acquisition at the edge - side.

[0045] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non - exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or elements inherent to such process, method, article or device.

[0046] Although the present invention has been described above with reference to the embodiments, various modifications thereof can be made and components thereof can be replaced with equivalents without departing from the scope of the present invention. In particular, as long as there is no structural conflict, the features of the embodiments disclosed in the present invention can be combined with each other in any way, and the exhaustive description of these combinations is not given in this specification only for the consideration of saving space and resources. Therefore, the present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. Design of a log generator based on a video stream, characterized in that: It includes an edge-side camera device, an edge-side data acquisition system, a cloud-based large model, an edge-side log recording and display system, a data transmission system, and a cloud storage module. The edge-side data acquisition system includes a coprocessor and an event generator; The edge-side camera device sends video stream data to the edge-side data acquisition system, and processes it through the coprocessor and the event generator to generate a LOG file; The cloud-based large model processes and analyzes through deep learning algorithms, natural language processing algorithms, reinforcement learning algorithms, etc., and inputs the analysis results into the edge-side log recording and display system; The edge-side log recording and display system records the analysis results and determines whether they are critical information. If they are critical information, it outputs data to the data transmission system. If they are not critical information, it performs edge-side filtering; The data transmission system uploads the output critical information to the cloud storage module.

2. The design of the log generator based on video stream according to claim 1, characterized in that: The edge-side data acquisition system is configured with one or more cameras, and the device is built with an artificial intelligence chip to achieve the ability of local intelligent data screening, extract key information from video stream data, complete the edge-side video data filtering task, and the edge-side device has a resolution of 1080P and a speed of 10fps.

3. The design of the log generator based on the video stream according to claim 1, characterized in that: According to the data obtained by the acquisition system, the general cloud-based large model performs recognition and analysis to achieve video-to-text conversion, generates a relatively coarse-grained text record, records it in sequence according to time, and at the same time marks the key data at key nodes and records it in the edge-side log recording and display system.

4. The design of the log generator based on video stream according to claim 1, characterized in that: The key data is uploaded to the cloud storage module through the data transmission system, and the remaining large amount of non-critical data is filtered out on the edge side.

5. The design of the log generator based on video stream according to claim 1, wherein: The cloud side completes the storage of key video stream data, and through recognition and analysis, it can quickly complete the query of a specific event.

Images