Alarm method based on secondary identification screening

Abstract

The invention belongs to the technical field of intelligent security and protection, and particularly relates to an alarm method based on secondary identification screening, which comprises the following steps: S1, introducing an AI algorithm training service to a GiSIM intelligent security and protection platform built in a port, performing data training on a large number of IGV lane human body identification results generated before, and optimizing an identification model; s2, when a person appears on the IGV lane, the 3800 platform identifies a human body identification event for the first time, the GiSIM intelligent security platform authorizes the 3800 platform to report the human body identification event, and an identification result is uploaded to the GiSIM intelligent security platform; s3, introducing an AI algorithm reasoning service to the established GiSIM intelligent security platform of the port, and carrying out secondary reasoning recognition on the human body recognition event reported by the 3800 platform; and S4, the GiSIM intelligent security platform determines whether to give an alarm according to the secondary identification result and pushes the alarm to the client. The method can solve the problems that in the prior art, the human body recognition false alarm rate is high, and a port IGV lane scene customization algorithm is lacked, and has good market application prospects.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent security, and particularly relates to an alarm method based on secondary recognition and screening. Background Art

[0002] Intelligent security refers to the use of modern information technologies such as the Internet of Things, big data, cloud computing, etc. to realize the intelligence of security prevention. Currently, many port IGV (Intelligent Guided Vehicle) lanes in China have adopted cameras for human body recognition to achieve the intelligent management of security and port logistics. The camera deploys a human body recognition algorithm on the built-in chip, captures image data in the scene through a high-resolution camera, and improves the image quality through image preprocessing steps, including denoising, contrast enhancement, and size adjustment; then, the foreground and background are separated through a subtraction algorithm to highlight the human body target in the video. The deep learning model extracts key features such as edges, textures, and shapes from the preprocessed images; then, through a large amount of data training, the model can learn the complex features of the human body; finally, the extracted features are compared with the known human body models in the database to achieve accurate recognition.

[0003] Through the statistical analysis of the recognition data of the existing technology, it is found that the existing technology has the following disadvantages:

[0004] First, the false alarm rate of camera human body recognition is high. In the existing technology, although the camera has achieved remarkable results in the field of human body recognition, in the actual application process, especially in the complex environment of the port IGV lane, limited by various factors such as the environment, light, and human body posture, the false alarm rate of human body recognition is relatively high. This not only increases the workload of security personnel and reduces work efficiency, but also may cause the interruption of the IGV lane operation, affecting the normal operation of port logistics. Long-term false alarms may lead to a decline in users' trust in the security system, thus affecting the construction and development of the entire intelligent port.

[0005] Second, there is a lack of customized recognition optimization for the characteristics of the IGV lane. The current human body recognition algorithms have not been deeply optimized for the characteristics of the port IGV lane. The IGV lane has unique scenarios and requirements, and the existing mature algorithms are difficult to meet the accuracy and stability requirements of human body recognition in such scenarios.

[0006] To solve the above problems, the present application proposes a customized human body recognition optimization technology for the characteristics of the port IGV lane, aiming to solve the problems of high false alarm rate of human body recognition and lack of customized algorithms for the port IGV lane scenario in the existing technology, and providing strong support for the intelligent security of the port IGV lane.

[0007] The information disclosed in this background section is only intended to enhance the overall understanding of the present invention and should not be regarded as an admission or any form of implication that this information constitutes prior art already known to those of ordinary skill in the art. Summary of the Invention

[0008] The object of the present invention is to provide an alarm method based on secondary recognition and screening to solve the problems of high false alarm rate of human body recognition and lack of customized algorithms for port IGV lane scenarios in the prior art.

[0009] To achieve the above object, the present invention provides the following technical solutions:

[0010] An alarm method based on secondary recognition and screening, comprising the following steps:

[0011] S1. Introduce AI algorithm training services into the established GiSIM intelligent security platform in the port, perform data training on a large number of previous IGV lane human body recognition results, and optimize the recognition model;

[0012] S2. When a person appears in the IGV lane, the 3800 platform first recognizes the human body recognition event, and the GiSIM intelligent security platform authorizes the 3800 platform to report the human body recognition event, and the recognition result is uploaded to the GiSIM intelligent security platform;

[0013] S3. Introduce AI algorithm inference services into the established GiSIM intelligent security platform in the port, and perform secondary inference recognition on the human body recognition events reported by the 3800 platform; in this stage, focus on the detailed features of the recognition target, screen out false alarm events, return correct human body recognition events, and improve the recognition accuracy;

[0014] S4. The GiSIM intelligent security platform determines whether to alarm according to the secondary recognition result and pushes it to the client to reduce the management personnel from receiving false alarm reports.

[0015] Compared with the prior art, the present invention has the following beneficial effects:

[0016] (1) The alarm method based on secondary recognition and screening of the present invention has the ability of algorithm visualization application scenario arrangement, and can configure the complex visual algorithm inference process through a graphical interface, which is convenient for users to understand and operate.

[0017] (2) The alarm method based on secondary recognition and screening of the present invention has the ability of full-process closed-loop training and inference, realizing the full-process closed-loop of visual algorithm training and inference from material annotation, algorithm training, algorithm management, algorithm authorization and distribution, algorithm inference, false alarm reporting, iterative training to algorithm optimization.

[0018] (3) The alarm method based on secondary recognition and screening of the present invention has the ability to manage the entire life cycle of the training model, and can realize the management of the entire life cycle of the training model from design, production, use, maintenance to final abandonment. Description of the Drawings

[0019] Figure 1 It is a schematic diagram of the AI algorithm training service;

[0020] Figure 2 It is a schematic diagram of the AI algorithm inference service. Detailed Implementation Modes

[0021] The technical solutions of the present invention will be described clearly and completely below. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0022] The present invention provides an alarm method based on secondary recognition and screening. The first goal is to optimize the recognition rate of the existing recognition algorithm and reduce false alarms; the second is to develop a customized recognition algorithm for the characteristics of the IGV lane; the third is to strengthen the application of artificial intelligence technology in the security field and improve the accuracy and stability of human body recognition. Specifically, it includes the following steps:

[0023] Step 1: Introduce the AI algorithm training service into the already established GiSIM intelligent security platform in the port, and refer to the attached Figure 1 , and through data training on a large number of previous human body recognition results in the IGV lane, continuously optimize the recognition model to improve the recognition accuracy and stability;

[0024] Step 2: When a person appears in the IGV lane, the 3800 platform (Huawei) first recognizes the human body recognition event, and the GiSIM intelligent security platform authorizes the 3800 platform to report the human body recognition event, and the recognition result is uploaded to the GiSIM intelligent security platform;

[0025] Step 3: Introduce the AI algorithm inference service into the already established GiSIM intelligent security platform in the port, and refer to the attached Figure 2 , perform secondary inference recognition on the human body recognition events reported by the 3800 platform. In this stage, focus on the detailed features of the recognition target, screen out the false alarm events among them, and return the correct human body recognition events to improve the recognition accuracy;

[0026] Step 4: The GiSIM intelligent security platform determines whether to alarm according to the secondary recognition result and pushes it to the client to reduce the managers from receiving false alarm reports.

[0027] In the AI algorithm training service, images and videos are uploaded to the dataset and data is labeled with tags; the data annotation process can choose manual annotation or automatic annotation; after data annotation, it enters the training task, and the data is trained with various training parameters and training resources; after training is completed, an algorithm model is generated, the algorithm model is tested and uploaded to the AI supermarket for users to download and deploy the algorithm; the algorithm model automatically annotates the model in the AI supermarket and performs model conversion. Resource monitoring is carried out during the training service, data conversion and model training are carried out during the execution of the training task, and the results are saved; the training service also includes model testing, model conversion and algorithm pre-labeling.

[0028] In the AI algorithm inference service, the scene view is split into multiple inference tasks, and the image recognition task is sent down. The GiSIM intelligent security platform responds to the request through the image recognition interface; after the inference task is sent down, real-time logs and inference events are reported. After the inference events are forwarded, they are reported to the GiSIM intelligent security platform. In the inference computing service, it will interact with the local servo, including resource monitoring, data processing, model loading, algorithm scheduling, algorithm inference, result feedback, etc.

[0029] The warning method based on secondary recognition and screening of the present invention can configure the complex visual algorithm inference process through a graphical interface, which is convenient for users to understand and operate. Through scenario-based orchestration, different hardware devices and different visual algorithm inference steps can be combined and configured to form scenario-based applications, realizing flexible and variable inference processes. At the same time, the visual scenario orchestration can also improve the transparency and interpretability of the inference process, helping engineers and users better understand and master the core ideas and methods of algorithm inference.

[0030] In addition, from material annotation, algorithm training, algorithm management, algorithm authorization distribution, algorithm inference, false alarm reporting, iterative training to algorithm optimization, the present invention realizes the full-process closed-loop of visual algorithm training and inference. The rigor and systematicness of this full-process closed-loop provide a strong guarantee for the efficient operation and continuous improvement of visual algorithms in practical applications, ensuring that visual algorithms can be continuously optimized and improved, and continuously improving algorithm accuracy, user experience and work efficiency.

[0031] In the design stage, the present invention can collect, screen, and label materials for the target, and select an appropriate model structure and algorithm according to the characteristics and requirements of the problem; in the generation stage, the present invention can iteratively train multiple times by continuously adjusting parameters and optimizing algorithms to improve the performance and accuracy of the model, and support testing and evaluation of the training results; in the usage stage, the present invention can uniformly manage the trained model, including model version and details, model authorization, and model download, and support deploying the model to the intelligent inference sub-platform for use by means of offline download or remote distribution; during the operation of the model, the present invention can monitor its inference process; in the maintenance stage, the present invention can promptly detect problems and provide material feedback, and perform secondary training on the model for the material feedback to adapt to the changing business requirements and data environment; in the abandonment stage, the present invention can control the deactivation of model download. It can be seen that the present invention has the ability to manage the entire life cycle of the training model from design, production, use, maintenance to final abandonment.

[0032] In the actual application of human body recognition by cameras in the IGV lane of a port, the concurrency ability is an important indicator to measure the system performance. The maximum concurrency of secondary human body recognition for human body recognition events is 60 images per second, thus ensuring the stable operation of the system during peak hours.

[0033] The formula for calculating the human body recognition rate is as follows:

[0034] Human body recognition rate = (total number of human body recognition events - number of false alarms) / total number of human body recognition events * 100%.

[0035] The technical solution of the present invention aims to reduce the number of false alarms. Through secondary recognition technology and customized optimization combined with the characteristics of the scenario, the false alarm phenomenon is effectively reduced, thereby reducing the ratio between the number of false alarms and the total number of human body recognition events and improving the overall recognition effect. For the requirements of human body recognition time consumption and recognition rate for AI services, the present invention puts forward strict indicators: in the IGV lane scenario, the human body recognition time consumption should be controlled within 1 second, and the recognition rate should reach over 95%.

[0036] The foregoing description of specific exemplary embodiments of the present invention is for the purposes of illustration and exemplification. These descriptions are not intended to limit the present invention to the precise forms disclosed, and obviously, many changes and variations are possible in light of the above teachings. The purpose of selecting and describing the exemplary embodiments is to explain the specific principles of the present invention and its practical applications, so that those skilled in the art can implement and utilize various different exemplary embodiments of the present invention as well as various different selections and changes. The scope of the present invention is intended to be defined by the claims and their equivalents.

Claims

1. An alarm method based on secondary identification and screening, characterized in that: The following steps are involved: S1. Introducing AI algorithm training services on the GiSIM smart security platform that has been built at the port, conducting data training on a large number of IGV lane human recognition results generated previously, and optimizing the recognition model; S2. When a person appears in the IGV lane, the 3800 platform identifies the human body recognition event for the first time. The GiSIM smart security platform authorizes the 3800 platform to report the human body recognition event, and the recognition result is uploaded to the GiSIM smart security platform; S3. The AI ​​algorithm inference service is introduced into the GiSIM smart security platform built in the port to perform secondary inference and recognition on human recognition events reported by the 3800 platform; S4. GiSIM smart security platform determines whether to alarm based on the secondary recognition results and pushes the information to the client.

Images