Intelligent outdoor safety monitoring system

Abstract

The invention discloses an intelligent outdoor safety monitoring system which comprises a camera module, a sensor module, a video image processing module, an MCU data processing module and an AI machine vision recognition system. The video image processing module is respectively connected with the MCU data processing module and the plurality of camera modules; the sensor module is connected to the MCU data processing module; the AI machine vision recognition system is connected to the MCU data processing module, and the AI machine vision recognition system trains a model of the recognition system through data acquisition and machine learning so as to improve the recognition precision; the MCU data processing module is used for receiving and processing data transmitted by the video image processing module and the sensor module so as to realize intelligent identification and classification of people and objects in a target area; the beneficial effects of the invention are that the system can accurately recognize the personnel and objects in the target area.

Description

Technical Field

[0001] The present invention relates to the technical field of monitoring systems, and more specifically, to an intelligent outdoor security monitoring system. Background Art

[0002] With the progress of science and technology, security monitoring has been applied in all aspects of our lives. For example, monitoring devices for detecting the driving status of vehicles on the road, security monitoring devices for monitoring the entry and exit of personnel in communities, and fire safety detection devices set in the power industry and the oil industry. These security monitoring systems have provided strong guarantees for people's lives.

[0003] Currently, with the acceleration of the urbanization process and the complexity of the social security situation, the security problems in the outdoor environment have become increasingly prominent. Traditional security devices have problems such as monitoring blind spots, untimely information processing, and accurate identification of dangerous animals, and cannot meet the actual needs. Therefore, it is of great significance to develop an efficient and reliable outdoor security monitoring system and equipment. Summary of the Invention

[0004] In order to overcome the deficiencies of the prior art, the present invention provides an intelligent outdoor security monitoring system that can accurately identify people and objects in the target area.

[0005] The technical solution adopted by the present invention to solve its technical problems is: an intelligent outdoor security monitoring system, which is improved in that it includes a camera module, a sensor module, a video image processing module, an MCU data processing module, and an AI machine vision recognition system; The video image processing module is respectively connected to the MCU data processing module and a plurality of camera modules. The camera module is used to obtain video image information of the target area and transmit the video image information to the MCU data processing module through the video image processing module; The sensor module is connected to the MCU data processing module and is used to monitor the environment in real time and obtain parameters; The AI machine vision recognition system is connected to the MCU data processing module. The AI machine vision recognition system trains the model of the recognition system through data collection and machine learning to improve the recognition accuracy; The MCU data processing module receives and processes the data transmitted from the video image processing module and the sensor module, and realizes the intelligent recognition and classification of people and objects in the target area.

[0006] Further, after the camera module obtains the video image information, Labelme is used to mark the video image, including color marking, brightness marking, and image division marking.

[0007] Further, during the data acquisition process of the AI machine vision recognition system, multiple datasets need to be integrated for experimental testing. These datasets include MNIST, CIFAR-10, CIFAR-100, ImageNet, COCO, Kinetics-700, LSUN, and IMDB-Wiki.

[0008] Further, the intelligent outdoor security monitoring system further includes a power control module, which is electrically connected to the MCU data processing module and used to supply power to the MCU data processing module.

[0009] Further, a lithium battery is built in the power control module, and the power control module has a charging port, through which the lithium battery is charged by a power adapter.

[0010] Further, the intelligent outdoor security monitoring system further includes an internal storage module, which is connected to the MCU data processing module.

[0011] Further, the intelligent outdoor security monitoring system further includes a communication module, which is electrically connected to the MCU data processing module. The communication module is connected to an external network and used to upload monitoring data to a cloud server or a mobile client to achieve remote monitoring and management.

[0012] Further, the sensor module includes an infrared sensor, a sound sensor, and a smoke sensor to monitor the temperature, humidity, and light of the environment in real time.

[0013] Further, the system construction of the AI machine vision recognition system is mainly developed in the Python language, and uses LSTM long short-term neural network / CUN convolutional neural network for learning to optimize supervised learning.

[0014] The beneficial effects of the present invention are as follows: The advantages of the system include: First, it can achieve all-weather monitoring of the outdoor environment and improve the security prevention ability. Second, through intelligent recognition and classification technologies, the false alarm rate can be reduced and the monitoring effect can be improved. Third, it supports remote monitoring and management, facilitating users to perform real-time operations and data viewing. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a schematic block diagram of an intelligent outdoor security monitoring system of the present invention.

[0016] Figure 2 is a schematic block diagram of the AI machine vision recognition system in the present invention. EMBODIMENTS

[0017] The present invention will be further described below with reference to the drawings and embodiments.

[0018] The concept, specific structure and technical effects of the present invention will be clearly and completely described below in conjunction with the embodiments and the accompanying drawings, so as to fully understand the purpose, features and effects of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, other embodiments obtained by those skilled in the art without creative efforts shall fall within the protection scope of the present invention. In addition, all the connection / linkage relationships involved in the patent do not simply refer to the direct connection of components, but refer to the formation of a more optimal connection structure by adding or reducing connection accessories according to specific implementation situations. Each technical feature in the present invention can be combined interactively on the premise of not conflicting with each other.

[0019] It should be noted that all the directional indications (such as up, down, left, right, front, back...) in the embodiments of the present invention are only used to explain the relative position relationship and movement conditions between components in a specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indications will also change accordingly.

[0020] It should also be noted that when an element is referred to as "fixed to" or "disposed on" another element, it can be directly on the other element or there may be an intermediate element at the same time. When an element is referred to as "connected to" another element, it can be directly connected to the other element or there may be an intermediate element at the same time.

[0021] In addition, the descriptions involving "first", "second", etc. in the present invention are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of such features. In addition, the technical solutions between various embodiments can be combined with each other, but it must be based on the fact that those skilled in the art can implement them. When the combination of technical solutions conflicts with each other or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the protection scope required by the present invention.

[0022] Refer to Figure 1As shown in the figure, the present invention provides an intelligent outdoor security monitoring system, which is mainly applied to outdoor security monitoring and can realize information push. The system includes a camera module, a sensor module, a video image processing module, an MCU data processing module, and an AI machine vision recognition system. The video image processing module is respectively connected to the MCU data processing module and multiple camera modules. The camera module is used to acquire video image information of the target area and transmit the image information to the MCU data processing module through the video image processing module. In this embodiment, the camera module uses a high-definition image sensor, which can provide clear video image information to more accurately identify people and objects in the target area.

[0023] In this embodiment, the sensor module is connected to the MCU data processing module and is used to monitor the environment in real time and obtain parameters. The sensor module includes an infrared sensor, a sound sensor, and a smoke sensor to monitor the temperature, humidity, and light of the environment in real time. It can also monitor abnormal situations.

[0024] Furthermore, the AI machine vision recognition system is connected to the MCU data processing module. The AI machine vision recognition system trains the model of the recognition system through data collection and machine learning to improve the recognition accuracy. The MCU data processing module receives and processes the data transmitted from the video image processing module and the sensor module to realize the intelligent recognition and classification of people and objects in the target area. In this embodiment, the AI machine vision recognition system is embedded in the MCU data processing module and combined with Figure 2 As shown in the figure, the AI machine vision recognition system trains the model of the recognition system through data set collection and machine learning, and provides the recognition accuracy and distance of dangerous animals, mainly including snakes / mice / people. It is intelligently pushed to the mobile terminal device, and warning sounds and vibration information are issued. The architecture of the AI machine vision recognition system is mainly developed in the Python language, and uses LSTM long short-term neural network / CUN convolutional neural network learning to optimize supervised learning. The system adopts a configuration module, which can control the distance range for recognizing dangerous animals.

[0025] In the above embodiment, during the data collection process, the camera needs to collect the data set of the video in real time, and use Labelme to realize the marking of the image, including color marking, brightness marking, and image division marking. The system integrated and conducted experimental tests on multiple data sets, including COCO, Pascal VOC, etc.

[0026] In this embodiment, the supported data set types are: 1. MNIST: This is a handwritten digit dataset containing 60,000 training samples and 10,000 test samples. Each image is a grayscale image of 28x28 pixels. This dataset is commonly used for training basic digit recognition models.

[0027] 2. CIFAR-10 and CIFAR-100: These two datasets contain small color images with 10 and 100 classes respectively. Each class has 500 training samples and 100 test samples. These datasets are commonly used for training models for object and class recognition.

[0028] 3. ImageNet: This is a large image dataset containing 14 million training samples and 500,000 validation samples. This dataset is widely used for training deep learning models, especially convolutional neural networks.

[0029] 4. COCO: This is a large dataset for object detection, segmentation, and keypoint detection, containing over 200,000 images and over 5,000,000 annotated objects. This dataset is commonly used for training object detection, segmentation, and keypoint detection models.

[0030] 5. Kinetics-700: This is a video dataset containing 700 action classes, with hundreds of samples in each class. This dataset is commonly used for training action recognition models.

[0031] 6. LSUN: This is a large dataset for scene understanding, containing nearly 1 million labeled images corresponding to 10 scene classes and 20 object classes. This dataset is commonly used for training models for scene and object recognition.

[0032] 7. IMDB-Wiki: This is one of the largest publicly available face datasets containing gender, age, and names, and can be used for training face recognition models. In this embodiment, the AI machine vision recognition system, through the acquisition of the dataset, in the machine learning process, adopts the D2Go system, which has machine learning and generates corresponding json files during training, including file paths, labeled values, the height and width of the bounding boxes, the styles of the bounding boxes, the ids of the pictures, the ids of the classes, and the labels containing a single class. After training for a phased cycle, a model file is output, and the model file is called during visual recognition, graphic classification, and scene segmentation.

[0033] The module structure and characteristics of the system are as follows: 1. Data Loaders: Responsible for loading and preprocessing the dataset, and dividing the data into training set, validation set, and test set.

[0034] 2. Model Architecture: D2Go provides a variety of predefined model architectures, such as RetinaNet, Mask R-CNN, Cascade R-CNN, etc. Users can select the appropriate model architecture according to their needs.

[0035] 3. Loss Function: It is used to optimize the training process of the model. D2Go supports a variety of loss functions, such as cross-entropy loss, L1 loss, etc.

[0036] 4. Training Loop: It is responsible for the training process of the model, including forward propagation, calculating loss, backpropagation, and parameter update. The D2Go system performs periodic deep learning every day, and the recognition accuracy is higher than 93%.

[0037] 5. Evaluation Tools: They are used to evaluate the performance of the model. D2Go provides a variety of evaluation metrics, such as accuracy, precision, recall, etc.

[0038] 6. High-quality Pre-trained Models: To facilitate users to quickly build object detection models, D2Go provides a variety of high-quality pre-trained models, such as ResNet, MobileNet, etc. Users can select the appropriate pre-trained model according to the specific task and fine-tune on this basis.

[0039] The integrated model algorithms include Faster R-CNN, Mask R-CNN, RetinaNet, and DensePose, including: Cascade R-CNN, Panoptic FPN, and TensorMask. By moving the entire training pipeline to the GPU, it is more convenient to perform distributed training on GPU servers under various standard models, thus easily expanding the training dataset. The algorithm adds image tracking, continuously tracks the recognized objects, automatically gives an early warning when the target distance reaches a certain distance, and for each recognition result, the cross-entropy loss function is used for scoring to calibrate the recognition result.

[0040] 7. Scalability: D2Go has high scalability and supports a variety of different model architectures and training strategies. In addition, it provides rich plugins and extension libraries to facilitate users to customize and expand according to their needs. New software implementations for productizing the model deployment are added, including: implementing the standard internal data training workflow, model compression and quantization, and model conversion.

[0041] 8. Ease of Use: D2Go provides rich documents and sample codes, enabling users to get started easily. At the same time, it also supports development using Python and C++, and provides friendly interfaces and configuration files to facilitate users in model development and debugging.

[0042] 9. Efficiency: D2Go adopts efficient training strategies and optimization algorithms, enabling the model to achieve good performance within a relatively short training time. In addition, it also supports accelerated computing using GPUs to improve the training speed and efficiency.

[0043] 10. Flexibility: Users can freely adjust parameters such as the model architecture, loss function, and training strategy according to their needs to obtain the best performance. In addition, it also supports model training and evaluation using different datasets and experimental configurations to meet different requirements.

[0044] 11. Visual configuration management module, select to identify and give early warnings for certain types of objects. Configuration management for infrared distance identification and early warning. Complete visual recognition management configuration.

[0045] 12. Software-hardware combination Identify target objects, parameters such as distance and temperature. When the configured parameters are reached, give an automatic early warning, send out early warning information, call the API interface of the terminal device, emit early warning sounds and light information, control the device, and remind users of the danger level of the surrounding safety environment.

[0046] 13. Not limited to early warnings for the identification of dangerous objects, animals, and vehicles, and can be extended to other outdoor industries, visual management, and applications.

[0047] In addition, for image processing, this system uses a convolutional neural network (CNN) for feature extraction and R-CNN for object detection. By combining the Region Proposal Network (RPN) and the classifier, this system can efficiently detect objects in images and has high accuracy and robustness. The object detection methods of the system are usually based on manually extracted features and classifiers, which are difficult to handle complex image content and different object categories. With the continuous development of deep learning technology, object detection methods based on CNN have gradually become a research hotspot. However, existing methods still have certain limitations in dealing with large-scale data and complex scenes. Therefore, the present invention proposes a deep learning-based object detection method and system to solve the deficiencies of the prior art.

[0048] Image processing steps: Use CNN to extract features from the input image; use RPN to generate candidate regions; use the classifier to classify the candidate regions; generate object detection results according to the classification results. The system includes three main components: CNN, RPN, and classifier.

[0049] The object detection method and system of the present invention have the following advantages: using CNN for feature extraction can effectively process large-scale data and complex scenarios; combining RPN and a classifier can efficiently detect objects in images and has high accuracy and robustness; it can handle different object categories and has broad application prospects.

[0050] Based on the above embodiments, continue to combine Figure 1 As shown, the intelligent outdoor security monitoring system further includes a power control module, an internal storage module, and a communication module. The power control module is electrically connected to the MCU data processing module and is used to supply power to the MCU data processing module. In this embodiment, a lithium battery is built into the power control module, and the power control module has a charging port, and the lithium battery is charged through a power adapter. The internal storage module is connected to the MCU data processing module, and the monitoring data is stored through the internal storage module. In addition, the communication module is electrically connected to the MCU data processing module, and the communication module is connected to an external network and is used to upload the monitoring data to a cloud server or a mobile phone client to achieve remote monitoring and management.

[0051] An intelligent outdoor security monitoring system of the present invention, its camera module can realize personnel area recognition and personnel face recognition. Among them, for the function of personnel area recognition, it can recognize personnel within a range of 1 to 30 meters; use an outdoor detector to detect the distance range for dangerous object recognition. For personnel face recognition, face FACEID recognition within a range of 1 to 30 meters, record the face ID, configure the recognized person as dangerous or a friend, and recognize the information within a 30-meter safety range. If it is a dangerous person, trigger a warning message and sound; on the contrary, when recognized as a safe person, no prompt message and warning are pushed.

[0052] The AI machine vision recognition system in the present invention can run independently on edge devices without transmitting data to the cloud for processing, which greatly reduces latency. This further protects the data and privacy security of end users. When connected to the network, the AI machine vision recognition system will be automatically upgraded. Through the most advanced high-efficiency mobile device backbone network, it supports end-to-end model training, quantization, and deployment, and can easily export the TorchScript format through D2Go. It can create an FBNet model that has been optimized for mobile devices and efficiently execute tasks such as object detection, semantic segmentation, and key point estimation on mobile devices. When connected to the network, the video is stored in a network disk to provide data backup and serve as a dataset for supervised learning.

[0053] Based on this, an intelligent outdoor security monitoring system of the present invention uses a high-definition image sensor to provide clear image information for more accurate identification of people and objects in the target area. It integrates multiple sensor modules, can monitor environmental parameters in all directions, and issue warning signals in a timely manner. The data processing unit adopts advanced image recognition and sound analysis technologies, and can realize intelligent recognition and classification of people and objects in the target area. It is connected to an external network through a communication module to achieve remote monitoring and management, facilitating users to obtain monitoring data and perform operations at any time. The advantages of this system include: First, it can achieve all-weather monitoring of the outdoor environment and improve the security prevention ability. Second, through intelligent recognition and classification technologies, the false alarm rate can be reduced and the monitoring effect can be improved. Third, it supports remote monitoring and management, facilitating users to perform real-time operations and data viewing.

[0054] An intelligent outdoor security monitoring system of the present invention can be widely applied to public places, communities, factories and other places where security monitoring is required, has a good market prospect, and is particularly suitable for the security equipment monitoring of outdoor tourism, and is mobile and portable.

[0055] The above is a specific description of the preferred embodiment of the present invention, but the present invention is not limited to the described embodiment. Those skilled in the art can also make various equivalent deformations or substitutions without departing from the spirit of the present invention, and these equivalent deformations or substitutions are all included in the scope defined by the claims of this application.

Claims

1. An intelligent outdoor security monitoring system, characterized in that, it includes a camera module, a sensor module, a video image processing module, an MCU data processing module, and an AI machine vision recognition system; The video image processing module is respectively connected to the MCU data processing module and multiple camera modules. The camera module is used to obtain video image information of the target area, and transmit the video image information to the MCU data processing module through the video image processing module; The sensor module is connected to the MCU data processing module, and is used to monitor the environment in real time and obtain parameters; The AI machine vision recognition system is connected to the MCU data processing module. The AI machine vision recognition system trains the model of the recognition system through data collection and machine learning to improve the recognition accuracy; The MCU data processing module receives and processes the data transmitted from the video image processing module and the sensor module, and realizes the intelligent recognition and classification of personnel and objects in the target area.

2. The intelligent outdoor security monitoring system according to claim 1, characterized in that, after the camera module obtains the video image information, Labelme is used to mark the video image, including color marking, brightness marking, and image division marking.

3. The intelligent outdoor security monitoring system according to claim 1, characterized in that, during the data collection process of the AI machine vision recognition system, multiple data sets need to be integrated for experimental testing. The data sets include MNIST, CIFAR-10 and CIFAR-100, ImageNet, COCO, Kinetics-700, LSUN, IMDB-Wiki.

4. The intelligent outdoor security monitoring system according to claim 1, characterized in that, the intelligent outdoor security monitoring system further includes a power control module, which is electrically connected to the MCU data processing module and is used to supply power to the MCU data processing module.

5. The intelligent outdoor security monitoring system according to claim 4, characterized in that, a lithium battery is built in the power control module, and the power control module has a charging port, and the lithium battery is charged through a power adapter.

6. The intelligent outdoor security monitoring system according to claim 1, characterized in that, the intelligent outdoor security monitoring system further includes an internal storage module, which is connected to the MCU data processing module.

7. The intelligent outdoor security monitoring system according to claim 1, characterized in that, the intelligent outdoor security monitoring system further includes a communication module, which is electrically connected to the MCU data processing module. The communication module is connected to an external network and is used to upload monitoring data to a cloud server or a mobile phone client to realize remote monitoring and management.

8. The intelligent outdoor security monitoring system according to claim 1, characterized in that, the sensor module includes an infrared sensor, a sound sensor, and a smoke sensor to monitor the temperature, humidity, and light of the environment in real time.

9. An intelligent outdoor security monitoring system according to claim 1, characterized in that, the system construction of the AI machine vision recognition system is mainly developed in the Python language, and the LSTM long short-term neural network / CUN convolutional neural network is used for learning to optimize supervised learning.

Images