Method and system for identifying fire fighting access occupation by using monitoring data
By amplifying and image repairing the monitoring video data, combining deep learning models and generative adversarial networks, accurate identification and automatic alarm of fire passage occupation are achieved, and the problem of inaccurate identification of fire passage occupation in the prior art is solved. Improves the efficiency and safety of fire rescue.
Patent Information
- Application Number
- CN202410993399.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-24
- Publication Date
- 2024-11-12
AI Technical Summary
The existing monitoring system cannot accurately identify whether the fire escape is occupied, resulting in the hindrance of rescue work when the fire occurs.
By collecting monitoring video data, the fire passage area is enlarged and image repaired, and the pre-constructed channel occupation recognition model is used to automatically identify the fire passage occupation and issue an alarm. Specific steps include using deep learning models to automatically identify fire channel boundaries, using super-resolution methods to amplify images, introducing attention mechanisms to extract key information, and using generative adversarial networks for image repair and quality evaluation.
It improves the accuracy and response speed of fire passage occupation identification, ensures that the fire passage is unobstructed, reduces the need for manual monitoring, and promptly notify relevant personnel to take measures.
Smart Images

Figure CN118942035A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of intelligent monitoring technology, and in particular to a method and system for identifying fire passage occupancy by using monitoring data. Background Art
[0002] At present, with the acceleration of urbanization, the problem of fire passage occupation has gradually become prominent. Fire passages are important escape routes in the event of a fire, and their unobstructed operation is crucial to protecting people's lives and property. However, due to various reasons, such as parking and stalls, fire passages are often occupied, which brings great difficulties to rescue work in the event of a fire. Although the existing monitoring system can monitor the fire passage, it often cannot accurately identify whether the passage is occupied. Therefore, developing a method that can accurately identify the occupation of fire passages is of great significance to maintaining public safety. Summary of the invention
[0003] The present invention provides a method and system for identifying fire passage occupation using monitoring data, which is used to solve the problem that the existing monitoring system cannot accurately identify fire passage occupation, including:
[0004] Collecting surveillance video data, and amplifying the fire passage area in the surveillance video data to obtain processed image data;
[0005] Performing image restoration on the processed image data to obtain channel restoration data;
[0006] According to the channel repair data, using a pre-built channel occupancy recognition model, obtain the fire channel area occupancy result in the channel repair data;
[0007] When the fire passage area occupancy result shows that the fire passage area is occupied, an alarm is automatically issued.
[0008] Optionally, the enlarging the fire passage area in the monitoring video data to obtain processed image data includes:
[0009] Automatically identify the fire passage boundary in the surveillance video data using a deep learning model to obtain the fire passage area in the surveillance video data;
[0010] The fire passage area is image-enlarged using a super-resolution method based on deep learning to obtain processed image data.
[0011] Optionally, the deep learning model includes the following training process:
[0012] Acquire pre-selected image monitoring data to form an image data set as a training data set; the image data set includes one or more of the following: illumination change image monitoring data, occlusion image monitoring data, and view angle change image monitoring data;
[0013] Using the image monitoring data as input of training data;
[0014] Outputting fire channel boundary information in the image monitoring data as training data;
[0015] Based on the input and output of the training data, the Faster R-CNN neural network is trained to obtain a deep learning model.
[0016] Optionally, the fire passage area is image-enlarged by using a super-resolution method based on deep learning to obtain processed image data, including:
[0017] Extracting features of the fire passage area to obtain passage feature data;
[0018] According to the channel feature data, an attention mechanism is introduced to obtain channel key information; the channel key information includes: fire signs, channel boundary boxes and evacuation routes;
[0019] According to the channel key information, a super-resolution convolutional neural network is used to perform image magnification processing to obtain processed image data.
[0020] Optionally, after performing image magnification processing using a super-resolution convolutional neural network according to the channel key information to obtain processed image data, the method further includes:
[0021] Acquire sensitive information in the processed image data, perform blurring on the sensitive information, and obtain privacy-protected data; the sensitive information includes: facial information, document information, and medical information;
[0022] The privacy protection data is encrypted to obtain encrypted image data.
[0023] Optionally, the channel occupancy recognition model includes the following training process:
[0024] Historical surveillance video data is used as input data;
[0025] The fire passage area occupancy results in the historical surveillance video data are used as output data;
[0026] Based on the input data and the output data, a convolutional neural network is trained to obtain a channel occupancy recognition model.
[0027] Optionally, performing image restoration on the processed image data to obtain channel restoration data includes:
[0028] Performing data preprocessing on the processed image data to obtain preprocessed repair data;
[0029] The pre-processed restoration data is restored by using a generative adversarial network to obtain channel restoration data.
[0030] Optionally, after performing image restoration on the preprocessed restoration data using a generative adversarial network to obtain channel restoration data, the method further includes:
[0031] Performing color correction, sharpening processing and denoising processing on the channel restoration data respectively to obtain post-processed image data;
[0032] A quality assessment is performed on the post-processed image data to obtain an image quality grade of the post-processed image data.
[0033] Based on the same inventive concept, the present invention also provides a system for identifying fire passage occupancy using monitoring data, comprising:
[0034] Video acquisition module: used to collect monitoring video data, amplify the fire passage area in the monitoring video data, and obtain processed image data;
[0035] Image restoration module: used to perform image restoration on the processed image data to obtain channel restoration data;
[0036] Channel occupancy module: used to obtain the fire channel area occupancy result in the channel repair data according to the channel repair data using a pre-built channel occupancy recognition model;
[0037] Alarm triggering module: used to automatically sound an alarm when the fire passage area occupancy result shows that the fire passage area is occupied.
[0038] Optionally, the video acquisition module is specifically used to:
[0039] Automatically identify the fire passage boundary in the surveillance video data using a deep learning model to obtain the fire passage area in the surveillance video data;
[0040] The fire passage area is image-enlarged using a super-resolution method based on deep learning to obtain processed image data.
[0041] Optionally, the deep learning model in the video acquisition module includes the following training process:
[0042] Acquire pre-selected image monitoring data to form an image data set as a training data set; the image data set includes one or more of the following: illumination change image monitoring data, occlusion image monitoring data, and view angle change image monitoring data;
[0043] Using the image monitoring data as input of training data;
[0044] Outputting fire channel boundary information in the image monitoring data as training data;
[0045] Based on the input and output of the training data, the Faster R-CNN neural network is trained to obtain a deep learning model.
[0046] Optionally, the video acquisition module uses a super-resolution method based on deep learning to enlarge the image of the fire passage area to obtain processed image data, including:
[0047] Extracting features of the fire passage area to obtain passage feature data;
[0048] According to the channel feature data, an attention mechanism is introduced to obtain channel key information; the channel key information includes: fire signs, channel boundary boxes and evacuation routes;
[0049] According to the channel key information, a super-resolution convolutional neural network is used to perform image magnification processing to obtain processed image data.
[0050] Optionally, the video acquisition module uses a super-resolution convolutional neural network to perform image magnification processing according to the channel key information to obtain processed image data, and further includes:
[0051] Acquire sensitive information in the processed image data, perform blurring on the sensitive information, and obtain privacy-protected data; the sensitive information includes: facial information, document information, and medical information;
[0052] The privacy protection data is encrypted to obtain encrypted image data.
[0053] Optionally, the channel occupancy recognition model in the channel occupancy module includes the following training process:
[0054] Historical surveillance video data is used as input data;
[0055] The fire passage area occupancy results in the historical surveillance video data are used as output data;
[0056] Based on the input data and the output data, a convolutional neural network is trained to obtain a channel occupancy recognition model.
[0057] Optionally, the image restoration module performs image restoration on the processed image data to obtain channel restoration data, including:
[0058] Performing data preprocessing on the processed image data to obtain preprocessed repair data;
[0059] The pre-processed restoration data is restored by using a generative adversarial network to obtain channel restoration data.
[0060] Optionally, after performing image restoration on the preprocessed restoration data using a generative adversarial network to obtain channel restoration data, the image restoration module further includes:
[0061] Performing color correction, sharpening processing and denoising processing on the channel restoration data respectively to obtain post-processed image data;
[0062] A quality assessment is performed on the post-processed image data to obtain an image quality grade of the post-processed image data.
[0063] Compared with the closest prior art, the present invention has the following beneficial effects:
[0064] The present invention provides a method and system for identifying fire passage occupation by using monitoring data, comprising: collecting monitoring video data, amplifying a fire passage area in the monitoring video data to obtain processed image data; performing image repair on the processed image data to obtain channel repair data; obtaining a fire passage area occupancy result in the channel repair data by using a pre-constructed channel occupancy recognition model according to the channel repair data; and automatically sounding an alarm when the fire passage area occupancy result shows that the fire passage area is occupied; the present invention can improve image resolution by amplifying the fire passage area in the monitoring video, so that subsequent analysis is more accurate, image repair technology can improve image quality, remove noise, blur or other visual interference, and ensure the accuracy of channel occupancy recognition, and by using a pre-trained channel occupancy recognition model, it can automatically monitor whether the fire passage is occupied in real time without continuous manual monitoring, and when it is detected that the fire passage is occupied, the system will automatically trigger an alarm, promptly notify relevant personnel to take measures, and improve the response speed.
[0065] Other features of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.
[0066] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS
[0067] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:
[0068] Figure 1 A flow chart of a method for identifying fire passage occupancy by using monitoring data in an embodiment of the present invention;
[0069] Figure 2 A schematic diagram of a model training structure in a method for identifying fire passage occupancy using monitoring data in a specific embodiment of the present invention;
[0070] Figure 3 The figure is a schematic diagram of a self-supervised training process in a method for identifying fire passage occupancy using monitoring data in a specific embodiment of the present invention. DETAILED DESCRIPTION
[0072] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.
[0073] 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. "Multiple" means two or more, unless otherwise clearly and specifically defined. 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.
[0074] 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.
[0075] Embodiment 1:
[0076] The present invention provides a method for identifying fire passage occupancy using monitoring data, such as Figure 1 As shown, including:
[0077] Step 1: Collect monitoring video data, and enlarge the fire passage area in the monitoring video data to obtain processed image data;
[0078] Step 2: performing image restoration on the processed image data to obtain channel restoration data;
[0079] Step 3: According to the channel repair data, using a pre-built channel occupancy recognition model, obtain the fire channel area occupancy result in the channel repair data;
[0080] Step 4: When the fire passage area occupancy result shows that the fire passage area is occupied, an alarm is automatically issued.
[0081] In one implementation, the fire passage area in the monitoring video data is enlarged in the above step 1 to obtain processed image data, including:
[0082] Automatically identify the fire passage boundary in the surveillance video data using a deep learning model to obtain the fire passage area in the surveillance video data;
[0083] The fire passage area is image-enlarged using a super-resolution method based on deep learning to obtain processed image data. A high-resolution camera array is used to cover the fire passage and its surrounding areas to ensure continuous collection of high-definition video data around the clock; in this implementation, the image data in the fire passage can be collected in real time through the video collection and processing system installed in the fire passage to be identified, and the data is transmitted back to the server.
[0084] In this implementation, the above-mentioned monitoring video data is collected by a video acquisition and processing system, which is installed near the fire passage area to be identified to ensure complete image coverage in the area to be monitored, and each installation location has a fixed location code to ensure that the corresponding location can be found according to the video acquisition system.
[0085] In this implementation, the above-mentioned deep learning model includes the following training process:
[0086] Acquire pre-selected image monitoring data to form an image data set as a training data set; the image data set includes one or more of the following: illumination change image monitoring data, occlusion image monitoring data, and view angle change image monitoring data;
[0087] Using the image monitoring data as input of training data;
[0088] Outputting fire channel boundary information in the image monitoring data as training data;
[0089] Based on the input and output of the training data, the Faster R-CNN neural network is trained to obtain a deep learning model; in the present invention, by using image monitoring data under complex conditions such as illumination changes, occlusion, and perspective changes as a training data set, the Faster R-CNN neural network is trained, which can enhance the robustness of the model, and the illumination change image monitoring data can help the model learn the ability to correctly identify objects under different lighting conditions. This is particularly important for scenes such as day and night monitoring, indoor and outdoor light changes. The occlusion image monitoring data allows the model to learn to accurately detect and identify even if some objects are occluded, which is very critical for object detection in complex scenes. For example, when detecting pedestrians, other pedestrians or obstacles may be occluded. The perspective change image monitoring data enables the model to identify the same object at different angles, which is very useful for multi-camera systems or mobile camera monitoring. Using the fire channel boundary information as a training output enables the model to accurately locate and identify the location of the fire channel, which is crucial for safety monitoring, especially in emergency situations such as fires, to ensure that the fire channel is unobstructed.
[0090] In one implementation, the fire passage area is enlarged using a super-resolution method based on deep learning to obtain processed image data, including:
[0091] Extracting features of the fire passage area to obtain passage feature data;
[0092] According to the channel feature data, an attention mechanism is introduced to obtain channel key information; the channel key information includes: fire signs, channel boundary boxes and evacuation routes;
[0093] According to the key information of the channel, a super-resolution convolutional neural network is used to perform image magnification processing to obtain processed image data. The above technical solution, i.e., extracting features of the fire channel area, introducing an attention mechanism to obtain key information of the channel, and using a super-resolution convolutional neural network to perform image magnification processing, can bring the following technical effects:
[0094] Improve detail recognition capabilities: The feature extraction process can capture key visual features within the fire passage area, such as the shape, color, and texture of the fire signs, as well as the geometric structure and edge information of the passage, which helps to more accurately identify and understand the passage environment.
[0095] Strengthening the focus on key information: After introducing the attention mechanism, the model can automatically focus on the most recognizable and informative parts of the image, such as fire signs, channel bounding boxes, and evacuation routes. This not only improves the processing efficiency of the model, but also enhances the recognition accuracy of important information and reduces the impact of background noise.
[0096] Improve image resolution: Using super-resolution convolutional neural networks to upscale images can improve image clarity and resolution without sacrificing details. This is especially important for long-distance monitoring or images captured by low-resolution cameras, and can help identify originally blurred details such as text, small objects, or distant human outlines.
[0097] Enhance the system's intelligent decision-making capabilities: Combined with the key channel information processed above, the system can more accurately determine whether the fire passage is unobstructed, whether the fire signs are intact, and whether the evacuation routes are clearly visible, providing more reliable data support for automatic alarm systems, emergency evacuation instructions, and maintenance inspections.
[0098] Improve safety and response speed: Clearer and more detailed image data can help quickly identify potential safety hazards, such as obstacles blocking fire escape routes or damaged firefighting equipment, thereby accelerating safety response and the implementation of preventive measures.
[0099] Optimizing resource allocation: The combination of attention mechanism and super-resolution processing can reduce the demand for computing resources because the model only performs high-precision analysis on key areas, while the background or other non-critical areas are processed with lower precision, thus achieving efficient resource utilization.
[0100] In this implementation, after the image magnification processing is performed using a super-resolution convolutional neural network according to the channel key information to obtain processed image data, the method further includes:
[0101] Acquire sensitive information in the processed image data, perform blurring on the sensitive information, and obtain privacy-protected data; the sensitive information includes: facial information, document information, and medical information;
[0102] The privacy protection data is encrypted to obtain encrypted image data; the above technical solution, that is, the sensitive information in the processed image data is blurred and encrypted, can bring the following important technical effects:
[0103] Privacy protection: By blurring sensitive information such as faces, documents, and medical information, personal privacy can be effectively protected to prevent such information from being illegally accessed or abused during transmission, storage, or analysis.
[0104] Enhanced data security: Data encryption is a further reinforcement of privacy-protected data. Even if the data is intercepted during transmission or storage, encryption technology can ensure that the data content cannot be easily cracked, thereby increasing the data security protection layer.
[0105] In one implementation, performing image restoration on the processed image data in the above step 2 to obtain channel restoration data includes:
[0106] Performing data preprocessing on the processed image data to obtain preprocessed repair data;
[0107] The pre-processed repair data is repaired using a generative adversarial network to obtain channel repair data; in this implementation, by using a generative adversarial network to repair the pre-processed repair data, a highly realistic image can be generated. When applied to channel repair, a repaired image with rich details, clear edges, and natural texture can be generated, which helps to improve the readability and usability of monitoring or recording images. In the image repair process, the generative adversarial network, especially the optimized variant, can restore occluded or missing details, which is particularly important for channel images because key visual information may be lost for various reasons.
[0108] In this implementation, after the above-mentioned using a generative adversarial network to perform image restoration on the pre-processed restoration data to obtain channel restoration data, it also includes:
[0109] Performing color correction, sharpening processing and denoising processing on the channel restoration data respectively to obtain post-processed image data;
[0110] Performing quality assessment on the post-processed image data to obtain an image quality grade of the post-processed image data;
[0111] In one implementation, the channel occupancy recognition model includes the following training process:
[0112] Historical surveillance video data is used as input data;
[0113] The fire passage area occupancy results in the historical surveillance video data are used as output data;
[0114] Based on the input data and output data, the convolutional neural network is trained to obtain a channel occupancy recognition model; in this implementation, the collected image data is used as input, and the channel occupancy recognition model constructed by the above method is used to detect whether the current fire channel is occupied, and the channel occupancy recognition model is continuously trained by using the data in the monitoring video and the self-supervised learning system based on image restoration to continuously improve the effect. Self-supervised learning allows the model to learn features from unlabeled data, which means that even without a large amount of labeled data, the model can improve its recognition ability by observing and understanding the structure and pattern in the image. Over time, the model can self-adjust and optimize to adapt to changes in the environment, such as light conditions, weather conditions or different types of obstacles; the present invention is based on deep learning technology, and the present invention proposes a fire channel occupancy recognition model, which can automatically identify whether the fire channel is occupied from the image. This solution adopts an image classification algorithm based on a convolutional neural network, and different models such as Resnet, densenet, efficientnet, etc. can be used as different embodiments.
[0115] In the above step 4, when the fire passage area occupancy result shows that the fire passage area is occupied, an alarm is automatically issued, and relevant information is pushed to the background to notify the management personnel.
[0116] In view of the problem that the existing monitoring system cannot accurately identify the occupation of fire passages, the present invention proposes a method for identifying the occupation of fire passages using monitoring data. The method performs image restoration on the fire passage data in the monitoring video, and then classifies the photos before restoration and the unrestored photos using a deep neural network, ultimately achieving accurate identification of the occupancy of the fire passage.
[0117] Embodiment 2:
[0118] A method for identifying fire passage occupancy by using monitoring data proposed by the present invention is described with a specific embodiment, which specifically includes:
[0119] (1) Data collection: Collect specific data on occupied and unoccupied fire escape passages, manually annotate the data, and assign category labels to each image, namely occupied and unoccupied.
[0120] (2) Image classification model training: Use the collected data to train neural network classification models such as ResNet, DenseNet, VGG, EfficientNet, etc. to complete the model training and enable it to have preliminary classification functions.
[0121] (3) Model deployment for image classification: The model is deployed to the fire escape occupancy recognition system, and each frame of the image captured by the surveillance camera is classified to determine whether the fire escape is occupied, and the classification results are labeled and saved.
[0122] (4) Generate image pairs using the images classified and labeled in step (3). Each pair of images contains an unoccupied image and an occupied image. Train the image restoration model. The model input is the occupied image, and the output target image is the unoccupied image. Use l2 loss to train this network. The model structure is similar to the UNet network structure, such as Figure 2 As shown in the figure, the final trained image has the function of restoring the occupied image to the unoccupied image.
[0123] (5) Input the occupied image and the unoccupied image after restoration into the image classification model in step (2) to continuously train the model and improve the classification accuracy of the model, thereby completing the self-supervised iteration of the model. The training process diagram is shown in the figure. Figure 3 As shown in the figure, this training method can improve the effect of the model.
[0124] Through the above specific embodiments, a large amount of data on the occupancy status of fire passages is collected and annotated, which provides a solid foundation for model training, enabling the system to make judgments based on real-world data and improve the accuracy and reliability of decision-making. Using deep learning models such as ResNet, DenseNet, VGG or EfficientNet, the automatic classification of fire passage images is realized to distinguish between the two states of "occupied" and "unoccupied", which greatly reduces the need for manual inspection and improves efficiency. By inputting the repaired image into the classification model again for training, a self-supervised learning mechanism is formed. This closed-loop training process can continuously improve the classification accuracy of the model, making it more accurate and robust. With the continuous iteration and optimization of the model, its classification ability in complex scenes will gradually increase, thereby reducing misjudgments, reducing the probability of false alarms and missed alarms, and improving the overall performance of the system. Ensuring that the fire passage is available at any time directly improves the life safety of people in the building, especially in emergency situations such as fires, unobstructed escape passages are crucial. Through the technical solution of the present invention, not only can the intelligent level of fire safety management be improved, but also data-driven decision-making can be promoted, and public safety and emergency management capabilities can be enhanced.
[0125] Embodiment 3:
[0126] The present invention provides a system for identifying fire passage occupancy by using monitoring data, the structural composition diagram of which is shown in FIG4 , including:
[0127] Video acquisition module: used to collect monitoring video data, amplify the fire passage area in the monitoring video data, and obtain processed image data;
[0128] Image restoration module: used to perform image restoration on the processed image data to obtain channel restoration data;
[0129] Channel occupancy module: used to obtain the fire channel area occupancy result in the channel repair data according to the channel repair data using a pre-built channel occupancy recognition model;
[0130] Alarm triggering module: used to automatically sound an alarm when the fire passage area occupancy result shows that the fire passage area is occupied.
[0131] The above-mentioned video acquisition module is specifically used for:
[0132] Automatically identify the fire passage boundary in the surveillance video data using a deep learning model to obtain the fire passage area in the surveillance video data;
[0133] The fire passage area is image-enlarged using a super-resolution method based on deep learning to obtain processed image data.
[0134] The deep learning model in the above video acquisition module includes the following training process:
[0135] Acquire pre-selected image monitoring data to form an image data set as a training data set; the image data set includes one or more of the following: illumination change image monitoring data, occlusion image monitoring data, and view angle change image monitoring data;
[0136] Using the image monitoring data as input of training data;
[0137] Outputting fire channel boundary information in the image monitoring data as training data;
[0138] Based on the input and output of the training data, the Faster R-CNN neural network is trained to obtain a deep learning model.
[0139] In the above-mentioned video acquisition module, the fire passage area is enlarged by a super-resolution method based on deep learning to obtain processed image data, including:
[0140] Extracting features of the fire passage area to obtain passage feature data;
[0141] According to the channel feature data, an attention mechanism is introduced to obtain channel key information; the channel key information includes: fire signs, channel boundary boxes and evacuation routes;
[0142] According to the channel key information, a super-resolution convolutional neural network is used to perform image magnification processing to obtain processed image data.
[0143] In the above-mentioned video acquisition module, after performing image magnification processing using a super-resolution convolutional neural network according to the channel key information and obtaining processed image data, it also includes:
[0144] Acquire sensitive information in the processed image data, perform blurring on the sensitive information, and obtain privacy-protected data; the sensitive information includes: facial information, document information, and medical information;
[0145] The privacy protection data is encrypted to obtain encrypted image data.
[0146] The channel occupancy recognition model in the channel occupancy module includes the following training process:
[0147] Historical surveillance video data is used as input data;
[0148] The fire passage area occupancy results in the historical surveillance video data are used as output data;
[0149] Based on the input data and the output data, a convolutional neural network is trained to obtain a channel occupancy recognition model.
[0150] The image restoration module performs image restoration on the processed image data to obtain channel restoration data, including:
[0151] Performing data preprocessing on the processed image data to obtain preprocessed repair data;
[0152] The pre-processed restoration data is restored by using a generative adversarial network to obtain channel restoration data.
[0153] After performing image restoration on the pre-processed restoration data using a generative adversarial network to obtain channel restoration data, the image restoration module further includes:
[0154] Performing color correction, sharpening processing and denoising processing on the channel restoration data respectively to obtain post-processed image data;
[0155] A quality assessment is performed on the post-processed image data to obtain an image quality grade of the post-processed image data.
[0156] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.
[0157] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0158] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0159] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0160] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.
Claims
1. A method for identifying fire passage occupancy using monitoring data, characterized in that: include: Collecting surveillance video data, and amplifying the fire passage area in the surveillance video data to obtain processed image data; Performing image restoration on the processed image data to obtain channel restoration data; According to the channel repair data, using a pre-built channel occupancy recognition model, obtain the fire channel area occupancy result in the channel repair data; When the fire passage area occupancy result shows that the fire passage area is occupied, an alarm is automatically issued.
2. The method according to claim 1, characterized in that The enlarging the fire passage area in the monitoring video data to obtain processed image data includes: Automatically identify the fire passage boundary in the surveillance video data using a deep learning model to obtain the fire passage area in the surveillance video data; The fire passage area is image-enlarged using a super-resolution method based on deep learning to obtain processed image data.
3. The method according to claim 2, characterized in that The deep learning model includes the following training process: Acquire pre-selected image monitoring data to form an image data set as a training data set; the image data set includes one or more of the following: illumination change image monitoring data, occlusion image monitoring data, and view angle change image monitoring data; Using the image monitoring data as input of training data; Outputting fire channel boundary information in the image monitoring data as training data; Based on the input and output of the training data, the Faster R-CNN neural network is trained to obtain a deep learning model.
4. The method according to claim 2, characterized in that The fire passage area is image-enlarged by using a super-resolution method based on deep learning to obtain processed image data, including: Extracting features of the fire passage area to obtain passage feature data; According to the channel feature data, an attention mechanism is introduced to obtain channel key information; the channel key information includes: fire signs, channel boundary boxes and evacuation routes; According to the channel key information, a super-resolution convolutional neural network is used to perform image magnification processing to obtain processed image data.
5. The method according to claim 4, characterized in that After the image magnification processing is performed by using a super-resolution convolutional neural network according to the channel key information to obtain processed image data, the method further includes: Acquire sensitive information in the processed image data, perform blurring on the sensitive information, and obtain privacy-protected data; the sensitive information includes: facial information, document information, and medical information; The privacy protection data is encrypted to obtain encrypted image data.
6. The method according to claim 1, characterized in that The channel occupancy recognition model includes the following training process: Historical surveillance video data is used as input data; The fire passage area occupancy results in the historical surveillance video data are used as output data; Based on the input data and the output data, a convolutional neural network is trained to obtain a channel occupancy recognition model.
7. The method according to claim 1, characterized in that The performing image restoration on the processed image data to obtain channel restoration data comprises: Performing data preprocessing on the processed image data to obtain preprocessed repair data; The pre-processed restoration data is restored by using a generative adversarial network to obtain channel restoration data.
8. The method according to claim 7, characterized in that After performing image restoration on the pre-processed restoration data using a generative adversarial network to obtain channel restoration data, the method further includes: Performing color correction, sharpening processing and denoising processing on the channel restoration data respectively to obtain post-processed image data; A quality assessment is performed on the post-processed image data to obtain an image quality grade of the post-processed image data.
9. A system for identifying fire escape occupancy using monitoring data, characterized in that: include: Video acquisition module: used to collect monitoring video data, amplify the fire passage area in the monitoring video data, and obtain processed image data; Image restoration module: used to perform image restoration on the processed image data to obtain channel restoration data; Channel occupancy module: used to obtain the fire channel area occupancy result in the channel repair data according to the channel repair data using a pre-built channel occupancy recognition model; Alarm triggering module: used to automatically sound an alarm when the fire passage area occupancy result shows that the fire passage area is occupied.
10. The system according to claim 9, characterized in that The video acquisition module is specifically used for: Automatically identify the fire passage boundary in the surveillance video data using a deep learning model to obtain the fire passage area in the surveillance video data; The fire passage area is image-enlarged using a super-resolution method based on deep learning to obtain processed image data.
