Smoke detection method and system based on image recognition technology

Abstract

The invention relates to the technical field of smoke detection, in particular to a smoke detection method and system based on an image recognition technology, and the system comprises an image collection module, an image preprocessing module, a smoke feature extraction module, a feature fusion and analysis module, a smoke region positioning module and an alarm and response module. Image data are collected in real time, an image preprocessing module preprocesses the image data, a smoke feature extraction module extracts multi-dimensional feature vectors of smoke from preprocessed images by using a deep convolutional neural network, and a feature fusion and analysis module carries out dimension reduction and classification analysis on extracted features. The smoke area positioning module positions the spatial position of the smoke based on the analysis result, and the alarm and response module automatically triggers the alarm device when detecting that the smoke exceeds a preset threshold value. According to the invention, high-precision smoke detection and positioning can be realized in a complex environment, and the early warning capability of fire risks is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of smoke detection, and in particular to a smoke detection method and system based on image recognition technology. Background Art

[0002] With the acceleration of urbanization and the increase in building density, fire prevention and control has become an important issue to ensure public safety. Among various fire prevention and control measures, early detection of smoke is crucial because smoke is usually one of the first signs of a fire. Traditional smoke detection methods mainly rely on traditional sensors such as ion smoke sensors and photoelectric smoke sensors. Although these devices can detect the presence of smoke to a certain extent, their detection range and sensitivity are often limited in large spaces, low-light environments or complex backgrounds, making it difficult to detect smoke quickly and accurately, resulting in delayed fire alarms and delaying the best time for emergency response.

[0003] However, the existing smoke detection technology based on image recognition still has some technical problems that need to be solved. First, traditional cameras often have difficulty obtaining clear image data in low light or complex backgrounds, resulting in reduced accuracy in smoke recognition. Second, existing image processing methods are insufficient in noise filtering and image enhancement, and are easily disturbed by ambient light and background clutter, affecting the extraction and recognition of smoke features. In addition, the current system generally lacks intelligent processing in smoke positioning and alarm response, and is unable to determine the spatial position and diffusion trend of smoke in a timely and accurate manner, delaying the timing of fire warning and emergency response. Summary of the invention

[0004] Based on the above objectives, the present invention provides a smoke detection method and system based on image recognition technology.

[0005] The smoke detection method based on image recognition technology includes the following steps:

[0006] S1, image acquisition: real-time video images are collected by multi-angle cameras set up in the monitoring area. The cameras are high-definition cameras with high sensitivity and low-light adaptability, ensuring that clear image data can still be captured in low-light environments and forming all-round coverage without blind spots in the monitoring area;

[0007] S2, image preprocessing: The collected video images are preprocessed. The preprocessing steps include image denoising, edge enhancement and color correction. The image noise is removed by combining adaptive median filtering and Gaussian filtering. The edge features of the image are enhanced by using the Canny edge detection algorithm to highlight the possible smoke area. The color deviation of the image is corrected by the color histogram equalization technology to ensure the accuracy of the subsequent processing steps.

[0008] S3, smoke feature extraction: In the preprocessed image, a deep convolutional neural network is used to extract features of the image. The features include color features, texture features, motion features and morphological features of the smoke. The features are extracted and quantified through the trained model to generate a set of smoke feature vectors;

[0009] S4, feature fusion and analysis: The extracted multi-dimensional smoke feature vectors are fused and analyzed, and the feature vectors are reduced in dimension using the principal component analysis method to reduce the computational complexity. At the same time, the Bayesian classifier is used to analyze the reduced-dimensional feature vectors to determine whether there is smoke in the image;

[0010] S5, smoke area positioning: Based on the analysis results, the regional growing algorithm is used to accurately locate the smoke, determine the position and range of the smoke in the monitoring area, and generate a three-dimensional model of the smoke distribution by combining it with the spatial position information of the camera to facilitate further fire risk assessment;

[0011] S6, Alarm and Response: When the smoke detection result exceeds the preset alarm threshold, the system automatically triggers an alarm signal and transmits the smoke location, concentration and diffusion information to the control center through the network.

[0012] Optionally, the step S1 specifically includes:

[0013] Camera location selection: Cameras are installed at key points in the surveillance area, including the four corners of the ceiling, both ends of the corridor, and above the center of the room to ensure full coverage;

[0014] Camera angle adjustment: According to the specific structure of the monitoring area, adjust the pitch angle and horizontal rotation angle of the camera to maximize the field of view and ensure the image quality in low light conditions;

[0015] Image acquisition strategy: The camera adopts an image acquisition strategy that combines fixed time intervals and event triggering. In low-light or no-light environments, infrared fill light is used to assist imaging, ensuring that high-quality image data can be captured even in extreme environments.

[0016] Optionally, the S2 specifically includes:

[0017] S21, adaptive median filtering: By dynamically adjusting the filter window size, it adaptively removes salt and pepper noise in the image and maintains the edge details of the image;

[0018] S22, Gaussian filtering: After adaptive median filtering, a Gaussian filter is used to further smooth the image to remove high-frequency noise while retaining the main features of the image;

[0019] S23, Canny edge detection: Perform Canny edge detection on the filtered image, use the double threshold method to detect and connect the significant edges of the image, and highlight the edge features of the smoke;

[0020] S24, color histogram equalization: performing color histogram equalization on the edge-detected image to correct image color deviation, enhance image contrast, and improve the recognition rate of smoke features.

[0021] Optionally, the S3 specifically includes:

[0022] S31, dataset construction: Use a large number of smoke image and non-smoke image datasets to build a training dataset, which contains images under different lighting, backgrounds and smoke forms;

[0023] S32, model architecture: A deep convolutional neural network model is used, which includes multiple convolutional layers, pooling layers and fully connected layers to automatically extract multi-level features in the image;

[0024] S33, training process: use the training data set to train the deep convolutional neural network model, adjust the model parameters through forward propagation and back propagation algorithms, and optimize the classification performance of the model;

[0025] S34, feature vector generation: The trained deep convolutional neural network model extracts features from the input image and generates a multi-dimensional feature vector including color features, texture features, motion features and morphological features.

[0026] Optionally, the S4 specifically includes:

[0027] S41, feature vector dimensionality reduction: principal component analysis is used to reduce the dimension of the high-dimensional smoke feature vector to reduce the dimension and computational complexity of the feature vector;

[0028] S42, Bayesian classifier analysis: The reduced feature vector is input into the Bayesian classifier, and the posterior probability of smoke in the image is calculated by combining the prior probability in the training phase and the observed feature data.

[0029] Optionally, the S5 specifically includes:

[0030] S51, region growing algorithm: based on the analysis results, an initial seed point is determined in the image, and the pixel area around the seed point is expanded using the region growing algorithm to finally determine the outline and range of the smoke;

[0031] S52, three-dimensional model generation: combining the spatial position information and positioning results of the camera, a three-dimensional model of smoke distribution is generated using triangulation method to accurately describe the spatial position and diffusion trend of smoke in the monitored area.

[0032] Optionally, the S6 specifically includes:

[0033] S61, alarm threshold setting: different alarm thresholds are set according to the specific conditions and environmental characteristics of the monitored area. The threshold parameters include smoke concentration, diffusion speed and smoke coverage area;

[0034] S62, alarm triggering mechanism: when the detection result exceeds the set threshold, the system automatically triggers the preset sound and light alarm device, and sends the alarm information to the control center through the network, and records the detection time, location, concentration and diffusion of the smoke;

[0035] S63, Response measures: The system automatically sends an alarm signal to a preset emergency response device, such as a fire protection system or a security monitoring system, to ensure that the fire risk can be controlled in a timely manner.

[0036] The smoke detection system based on image recognition technology is used to implement the above-mentioned smoke detection method based on image recognition technology, and includes the following modules:

[0037] Image acquisition module: It is equipped with multiple cameras, which are installed at different locations in the monitoring area to collect video images covering the monitoring area in real time;

[0038] Image preprocessing module: preprocesses the collected images. The preprocessing steps include image denoising, edge enhancement and color correction. Adaptive median filtering, Gaussian filtering and Canny edge detection algorithm are used to remove noise and enhance the edge features of the image. At the same time, the color histogram equalization technology is used to correct the color deviation of the image.

[0039] Smoke feature extraction module: Based on the deep convolutional neural network model, the smoke features are extracted from the preprocessed image, including color features, texture features, motion features and morphological features, and the corresponding feature vectors are generated;

[0040] Feature fusion and analysis module: It performs fusion analysis on the extracted multi-dimensional smoke feature vectors, uses principal component analysis to reduce the dimension, and uses Bayesian classifier to analyze the reduced feature vectors to determine whether there is smoke in the image;

[0041] Smoke area positioning module: Based on the analysis results, the regional growing algorithm is used to accurately locate the position and range of smoke in the monitoring area, and a three-dimensional model of smoke distribution is generated in combination with the spatial position information of the camera;

[0042] Alarm and response module: When the detection result exceeds the preset alarm threshold, the alarm device is automatically triggered, and the smoke location, concentration and diffusion information is transmitted to the control center through the network, and the emergency response device is triggered at the same time.

[0043] Beneficial effects of the present invention:

[0044] The system of the present invention uses advanced image preprocessing technologies such as adaptive median filtering, Gaussian filtering, Canny edge detection and color histogram equalization to effectively remove noise and enhance image edge features and color contrast. Combined with a deep convolutional neural network trained with a large amount of data, the system can accurately extract and quantify the multi-dimensional features of smoke, and accurately identify the presence of smoke even under complex background and lighting conditions. This efficient feature extraction and analysis process significantly improves the accuracy and robustness of smoke detection.

[0045] The present invention combines the regional growth algorithm with the spatial position information of the camera, so that the system can accurately locate the area where smoke occurs and generate a three-dimensional model to intuitively display the distribution and diffusion of smoke. When the detection result exceeds the preset threshold, the system can automatically trigger an alarm and quickly transmit the location information of the smoke to the control center, while linking the emergency response device. This intelligent positioning and response mechanism ensures a quick response when the fire risk first appears, effectively controls potential dangers, and ensures the safety of people and property. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0047] Figure 1 A schematic diagram of a method flow chart of an embodiment of the present invention;

[0048] Figure 2 Schematic diagram of a system flow of an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0050] It should be noted that the references to "one embodiment", "an embodiment", "an exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0051] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0052] like Figure 1 As shown, the smoke detection method based on image recognition technology includes the following steps:

[0053] S1, image acquisition: real-time video images are collected by multi-angle cameras set up in the monitoring area. The cameras are high-definition cameras with high sensitivity and low-light adaptability, ensuring that clear image data can still be captured in low-light environments and forming all-round coverage without blind spots in the monitoring area;

[0054] S2, image preprocessing: The collected video images are preprocessed. The preprocessing steps include image denoising, edge enhancement and color correction. The image noise is removed by combining adaptive median filtering and Gaussian filtering. The edge features of the image are enhanced by using the Canny edge detection algorithm to highlight the possible smoke area. The color deviation of the image is corrected by the color histogram equalization technology to ensure the accuracy of the subsequent processing steps.

[0055] S3, smoke feature extraction: In the preprocessed image, a deep convolutional neural network is used to extract features of the image. The features include color features, texture features, motion features and morphological features of the smoke. The features are extracted and quantified through a trained model to generate a set of smoke feature vectors. The deep convolutional neural network model is pre-trained on a large amount of smoke and non-smoke image data to ensure that smoke features can be accurately distinguished in complex backgrounds.

[0056] S4, feature fusion and analysis: The extracted multi-dimensional smoke feature vectors are fused and analyzed, and the feature vectors are reduced in dimension using the principal component analysis method to reduce the computational complexity. At the same time, the Bayesian classifier is used to analyze the reduced-dimensional feature vectors to determine whether there is smoke in the image;

[0057] S5, smoke area positioning: Based on the analysis results, the regional growing algorithm is used to accurately locate the smoke, determine the position and range of the smoke in the monitoring area, and generate a three-dimensional model of the smoke distribution by combining it with the spatial position information of the camera to facilitate further fire risk assessment;

[0058] S6, Alarm and Response: When the smoke detection result exceeds the preset alarm threshold, the system automatically triggers an alarm signal and transmits the smoke location, concentration and diffusion information to the control center through the network.

[0059] Optionally, the step S1 specifically includes:

[0060] Camera location selection: Cameras are installed at key points in the surveillance area, including the four corners of the ceiling, both ends of the corridor, and above the center of the room to ensure full coverage;

[0061] Camera angle adjustment: According to the specific structure of the monitoring area, adjust the pitch angle and horizontal rotation angle of the camera to maximize the field of view and ensure the image quality in low light conditions;

[0062] Image acquisition strategy: The camera adopts an image acquisition strategy that combines fixed time intervals and event triggering. In low-light or no-light environments, infrared fill light is used to assist imaging, ensuring that high-quality image data can be captured even in extreme environments.

[0063] By optimizing the position and angle of the camera and combining event triggering and infrared fill light technology, the system can achieve comprehensive and clear image acquisition under different environmental conditions, laying the foundation for subsequent smoke detection.

[0064] Optionally, the S2 specifically includes:

[0065] S21, adaptive median filtering: By dynamically adjusting the filter window size, it adaptively removes salt and pepper noise in the image and maintains the edge details of the image;

[0066] S22, Gaussian filtering: After adaptive median filtering, a Gaussian filter is used to further smooth the image to remove high-frequency noise while retaining the main features of the image;

[0067] S23, Canny edge detection: Perform Canny edge detection on the filtered image, use the double threshold method to detect and connect the significant edges of the image, and highlight the edge features of the smoke;

[0068] S24, color histogram equalization: performing color histogram equalization on the edge-detected image to correct the image color deviation, enhance the image contrast, and improve the recognition rate of smoke features;

[0069] Through the combination of multiple filtering and edge detection, the system can effectively remove noise, enhance image edge features and color contrast, and provide a high-quality image foundation for the subsequent accurate extraction of smoke features.

[0070] Optionally, the S3 specifically includes:

[0071] S31, dataset construction: Use a large number of smoke image and non-smoke image datasets to build a training dataset, which contains images under different lighting, backgrounds and smoke forms;

[0072] S32, model architecture: A deep convolutional neural network model is used, which includes multiple convolutional layers, pooling layers and fully connected layers to automatically extract multi-level features in the image;

[0073] S33, training process: use the training data set to train the deep convolutional neural network model, adjust the model parameters through forward propagation and back propagation algorithms, and optimize the classification performance of the model;

[0074] S34, feature vector generation: The trained deep convolutional neural network model extracts features from the input image and generates a multi-dimensional feature vector including color features, texture features, motion features, and morphological features;

[0075] Through the training and optimization of deep convolutional neural networks, the system can automatically extract and generate multi-dimensional vectors that best represent the characteristics of smoke, thereby improving the accuracy and robustness of smoke detection.

[0076] Optionally, the S4 specifically includes:

[0077] S41, feature vector dimensionality reduction: principal component analysis is used to reduce the dimension of the high-dimensional smoke feature vector to reduce the dimension and computational complexity of the feature vector;

[0078] S42, Bayesian classifier analysis: The reduced feature vector is input into the Bayesian classifier, and the posterior probability of smoke in the image is calculated by combining the prior probability in the training phase and the observed feature data, which is expressed as:

[0079]

[0080] Where P(smoke|X) is the probability of smoke in the image given the feature vector X, P(X|smoke) is the probability of the feature vector X given the presence of smoke, P(smoke) is the prior probability of smoke in the image, and P(X) is the marginal probability of the feature vector X;

[0081] Through the combination of principal component analysis and Bayesian classifier, the system can effectively reduce the computational complexity and accurately analyze the smoke characteristics in the image, thereby improving the real-time and reliability of smoke detection.

[0082] Optionally, the S5 specifically includes:

[0083] S51, region growing algorithm: based on the analysis results, an initial seed point is determined in the image, and the pixel area around the seed point is expanded using the region growing algorithm to finally determine the outline and range of the smoke;

[0084] S52, three-dimensional model generation: combining the spatial position information and positioning results of the camera, a three-dimensional model of smoke distribution is generated by using triangulation method to accurately describe the spatial position and diffusion trend of smoke in the monitored area;

[0085] Through region growing algorithm and three-dimensional model generation, the system can accurately locate the spatial position and range of smoke, providing intuitive and detailed information support for fire risk assessment.

[0086] Optionally, the S6 specifically includes:

[0087] S61, alarm threshold setting: different alarm thresholds are set according to the specific conditions and environmental characteristics of the monitored area. The threshold parameters include smoke concentration, diffusion speed and smoke coverage area;

[0088] S62, alarm triggering mechanism: when the detection result exceeds the set threshold, the system automatically triggers the preset sound and light alarm device, and sends the alarm information to the control center through the network, and records the detection time, location, concentration and diffusion of the smoke;

[0089] S63, Response measures: The system automatically sends an alarm signal to a preset emergency response device, such as a fire protection system or a security monitoring system, to ensure that the fire risk can be controlled in a timely manner;

[0090] By setting alarm thresholds and automatic response mechanisms, the system can quickly trigger an alarm when smoke detection reaches a dangerous level and transmit alarm information in a timely manner to ensure that fire risks can be quickly and effectively controlled.

[0091] like Figure 2 As shown, the smoke detection system based on image recognition technology is used to implement the above-mentioned smoke detection method based on image recognition technology, and includes the following modules:

[0092] Image acquisition module: It is equipped with multiple cameras, which are installed at different locations in the monitoring area to collect video images covering the monitoring area in real time;

[0093] Image preprocessing module: preprocesses the collected images. The preprocessing steps include image denoising, edge enhancement and color correction. Adaptive median filtering, Gaussian filtering and Canny edge detection algorithm are used to remove noise and enhance the edge features of the image. At the same time, the color histogram equalization technology is used to correct the color deviation of the image.

[0094] Smoke feature extraction module: Based on the deep convolutional neural network model, the smoke features are extracted from the preprocessed image, including color features, texture features, motion features and morphological features, and the corresponding feature vectors are generated;

[0095] Feature fusion and analysis module: It performs fusion analysis on the extracted multi-dimensional smoke feature vectors, uses principal component analysis to reduce the dimension, and uses Bayesian classifier to analyze the reduced feature vectors to determine whether there is smoke in the image;

[0096] Smoke area positioning module: Based on the analysis results, the regional growing algorithm is used to accurately locate the position and range of smoke in the monitoring area, and a three-dimensional model of smoke distribution is generated in combination with the spatial position information of the camera;

[0097] Alarm and response module: When the detection result exceeds the preset alarm threshold, the alarm device is automatically triggered, and the smoke location, concentration and diffusion information is transmitted to the control center through the network, and the emergency response device is triggered at the same time.

[0098] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0099] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A smoke detection method based on image recognition technology, characterized in that: The following steps are involved: S1, image acquisition: real-time video images are collected through multi-angle cameras set up in the monitoring area; S2, image preprocessing: preprocess the collected video images. The preprocessing steps include image denoising, edge enhancement and color correction. The image noise is removed by combining adaptive median filtering and Gaussian filtering. The edge features of the image are enhanced by using the Canny edge detection algorithm. The color deviation of the image is corrected by using the color histogram equalization technology. S3, smoke feature extraction: In the preprocessed image, a deep convolutional neural network is used to extract features of the image. The features include color features, texture features, motion features and morphological features of the smoke. The features are extracted and quantified through the trained model to generate a set of smoke feature vectors; S4, feature fusion and analysis: The extracted multi-dimensional smoke feature vectors are fused and analyzed, and the feature vectors are reduced in dimension using the principal component analysis method to reduce the computational complexity. At the same time, the Bayesian classifier is used to analyze the reduced-dimensional feature vectors to determine whether there is smoke in the image; S5, smoke area positioning: Based on the analysis results, the regional growing algorithm is used to accurately locate the smoke, determine the position and range of the smoke in the monitoring area, and generate a three-dimensional model of the smoke distribution by combining it with the spatial position information of the camera to facilitate further fire risk assessment; S6, Alarm and Response: When the smoke detection result exceeds the preset alarm threshold, the system automatically triggers an alarm signal and transmits the smoke location, concentration and diffusion information to the control center through the network.

2. The smoke detection method based on image recognition technology according to claim 1 is characterized in that: The step S1 specifically includes: Camera location selection: Cameras are installed at key points in the surveillance area, including the four corners of the ceiling, both ends of the corridor, and above the center of the room; Camera angle adjustment: According to the specific structure of the monitoring area, adjust the camera's pitch angle and horizontal rotation angle to maximize the field of view; Image acquisition strategy: The camera adopts an image acquisition strategy that combines fixed time intervals and event triggering. In low-light or no-light environments, infrared fill light is used to assist imaging.

3. The smoke detection method based on image recognition technology according to claim 2 is characterized in that: The S2 specifically includes: S21, adaptive median filtering: By dynamically adjusting the filter window size, it adaptively removes salt and pepper noise in the image and maintains the edge details of the image; S22, Gaussian filtering: After adaptive median filtering, a Gaussian filter is used to further smooth the image to remove high-frequency noise while retaining the main features of the image; S23, Canny edge detection: Perform Canny edge detection on the filtered image, use the double threshold method to detect and connect the significant edges of the image, and highlight the edge features of the smoke; S24, color histogram equalization: performing color histogram equalization on the edge-detected image to correct image color deviation, enhance image contrast, and improve the recognition rate of smoke features.

4. The smoke detection method based on image recognition technology according to claim 3 is characterized in that: The S3 specifically includes: S31, dataset construction: Use smoke image and non-smoke image datasets to construct a training dataset, which contains images under different lighting, backgrounds and smoke forms; S32, model architecture: A deep convolutional neural network model is used, which includes multiple convolutional layers, pooling layers and fully connected layers to automatically extract multi-level features in the image; S33, training process: use the training data set to train the deep convolutional neural network model, adjust the model parameters through forward propagation and back propagation algorithms, and optimize the classification performance of the model; S34, feature vector generation: The trained deep convolutional neural network model extracts features from the input image and generates a multi-dimensional feature vector including color features, texture features, motion features and morphological features.

5. The smoke detection method based on image recognition technology according to claim 4 is characterized in that: The S4 specifically includes: S41, feature vector dimensionality reduction: principal component analysis is used to reduce the dimensionality of high-dimensional smoke feature vectors to reduce the dimension and computational complexity of the feature vectors; S42, Bayesian classifier analysis: The reduced feature vector is input into the Bayesian classifier, and the posterior probability of smoke in the image is calculated by combining the prior probability in the training phase and the observed feature data.

6. The smoke detection method based on image recognition technology according to claim 5 is characterized in that: The S5 specifically includes: S51, region growing algorithm: based on the analysis results, an initial seed point is determined in the image, and the pixel area around the seed point is expanded using the region growing algorithm to finally determine the outline and range of the smoke; S52, three-dimensional model generation: combining the spatial position information and positioning results of the camera, a three-dimensional model of smoke distribution is generated using triangulation method to accurately describe the spatial position and diffusion trend of smoke in the monitored area.

7. The smoke detection method based on image recognition technology according to claim 6 is characterized in that: The S6 specifically includes: S61, alarm threshold setting: different alarm thresholds are set according to the specific conditions and environmental characteristics of the monitored area. The threshold parameters include smoke concentration, diffusion speed and smoke coverage area; S62, alarm triggering mechanism: when the detection result exceeds the set threshold, the system automatically triggers the preset sound and light alarm device and sends the alarm information to the control center through the network; S63, Response measures: The system automatically sends an alarm signal to the preset emergency response device to ensure that the fire risk can be controlled in time.

8. A smoke detection system based on image recognition technology, used to implement a smoke detection method based on image recognition technology as described in any one of claims 1 to 9, characterized in that: Includes the following modules: Image acquisition module: It is equipped with multiple cameras, which are installed at different locations in the monitoring area to collect video images covering the monitoring area in real time; Image preprocessing module: preprocesses the collected images. The preprocessing steps include image denoising, edge enhancement and color correction. Adaptive median filtering, Gaussian filtering and Canny edge detection algorithm are used to remove noise and enhance the edge features of the image. At the same time, the color histogram equalization technology is used to correct the color deviation of the image. Smoke feature extraction module: Based on the deep convolutional neural network model, the smoke features are extracted from the preprocessed image, including color features, texture features, motion features and morphological features, and the corresponding feature vectors are generated; Feature fusion and analysis module: It performs fusion analysis on the extracted multi-dimensional smoke feature vectors, uses principal component analysis to reduce the dimension, and uses Bayesian classifier to analyze the reduced feature vectors to determine whether there is smoke in the image; Smoke area positioning module: Based on the analysis results, the regional growing algorithm is used to accurately locate the position and range of smoke in the monitoring area, and a three-dimensional model of smoke distribution is generated in combination with the spatial position information of the camera; Alarm and response module: When the detection result exceeds the preset alarm threshold, the alarm device is automatically triggered, and the smoke location, concentration and diffusion information is transmitted to the control center through the network, and the emergency response device is triggered at the same time.

Images