Fire safety assessment method and system based on machine learning

Abstract

The invention discloses a fire safety assessment method and system based on machine learning, and relates to a machine learning and fire safety assessment technology, and the method comprises the steps: collecting image data, equipment operation data and environment monitoring data of a fire-fighting place; inputting the first image feature vector into a pre-trained deep convolutional neural network model to obtain a risk level label, and inputting the first standardized data into a long-short term memory network model to obtain a first time sequence prediction result; performing feature fusion on the first risk level label and the first time sequence prediction result, constructing a fusion feature matrix, and inputting the fusion feature matrix into a random forest model for parameter learning; and inputting newly collected fire-fighting data into the learned random forest model to carry out safety level prediction so as to obtain a fire-fighting safety prediction result and generate risk early warning information. According to the invention, risk early warning information can be generated in real time, fire-fighting potential safety hazards can be quickly responded, and the fire-fighting management efficiency is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical fields of machine learning and fire safety assessment, and particularly to a fire safety assessment method and system based on machine learning. Background Art

[0002] Traditional fire safety assessment methods mainly rely on manual inspections and historical data analysis. Although these methods can ensure fire safety to a certain extent, there are many deficiencies. With the continuous development of sensing technology, computer vision, and big data technology, more and more intelligent technologies have been applied to the field of fire safety assessment. Through means such as image recognition, environmental data monitoring, and equipment status analysis, intelligent fire safety assessment methods have gradually replaced the traditional manual inspection mode. The application of machine learning methods such as deep learning, long short-term memory network (LSTM), and random forest has made fire safety assessment not only more accurate but also able to predict potential safety hazards in advance, further improving the fire safety prevention ability.

[0003] However, there are still some deficiencies in the application of existing intelligent fire safety assessment methods. In the prior art, risk assessment based on image data mainly relies on image feature extraction and classification, but often ignores the impact of equipment operation data and environmental changes on fire safety. In addition, existing methods usually use a single model for assessment, resulting in one-sided evaluation results and lack of comprehensiveness. For example, the limitations of traditional deep convolutional neural network (CNN) and LSTM models in processing time series data and spatial data may lead to low assessment accuracy and difficulty in effectively integrating various types of data for comprehensive safety assessment. Moreover, in practical applications, the data sources and types in fire protection sites are numerous, and how to accurately and effectively process and analyze these data remains an urgent problem in the field of intelligent fire safety assessment.

[0004] In view of the deficiencies of the prior art, the present invention proposes a fire safety assessment method and system based on machine learning. By collecting image data, equipment operation data, and environmental monitoring data of fire protection sites and combining multi-modal learning technologies of deep convolutional neural network, LSTM network, and random forest model, a comprehensive assessment of the safety risks of fire protection sites is achieved. Especially when dealing with complex spatio-temporal data and multi-dimensional information, it shows stronger advantages and can identify potential safety hazards in advance and generate effective risk warning information. Summary of the Invention

[0005] In view of the problems of insufficient data fusion and single model in existing fire safety assessment methods, the present invention is proposed.

[0006] Therefore, the problem to be solved by the present invention is how to improve the accuracy and predictability of fire safety assessment by integrating image data, equipment operation data, and environmental monitoring data, and using machine learning, so as to achieve a comprehensive assessment and real-time warning of the safety risks of fire protection sites, and reduce the occurrence of potential fire safety hazards.

[0007] To solve the above technical problems, the present invention provides the following technical solutions:

[0008] In a first aspect, an embodiment of the present invention provides a machine learning-based fire safety assessment method,

[0009] which includes collecting image data, equipment operation data, and environmental monitoring data of a fire protection site, extracting features from the image data to obtain a first image feature vector, and performing standardization processing on the equipment operation data and environmental monitoring data to obtain first standardized data; inputting the first image feature vector into a pre-trained deep convolutional neural network model to obtain a risk level label, and at the same time inputting the first standardized data into a long short-term memory network model to obtain a first time series prediction result; performing feature fusion on the first risk level label and the first time series prediction result to construct a fusion feature matrix, and inputting the fusion feature matrix into a random forest model for parameter learning; inputting newly collected fire data into the learned random forest model for safety level prediction to obtain a prediction result of fire safety and generating a risk warning message.

[0010] As a preferred solution of the machine learning-based fire safety assessment method of the present invention, the method of inputting newly collected fire data into the learned random forest model for safety level prediction to obtain a prediction result of fire safety and generating a risk warning message includes: preprocessing the newly collected fire data to construct a prediction feature matrix; inputting the prediction feature matrix into the learned random forest model, and calculating the probability distribution of the sample belonging to the safety level based on the voting results of each decision tree; determining the level corresponding to the first probability value in the probability distribution as the result of the safety level prediction, and at the same time recording the confidence value of the prediction result, where the confidence value is equal to the difference between the first probability value and the second probability; determining the safety level prediction result based on a risk threshold, and setting a warning level according to the determination result and the confidence value.

[0011] As a preferred solution of the machine learning-based fire safety assessment method of the present invention, wherein: the risk thresholds include a first threshold, a first threshold, a first threshold, and a first threshold; further, when the first probability value is greater than the first threshold and the confidence value is greater than the first preset confidence value, it is determined as a first-level risk in the risk level label; if it is a first-level risk, a red warning is immediately triggered, and fire-fighting equipment and an emergency response mechanism are linked; when the first probability value is between the first threshold and the second threshold and the confidence value is between the first preset confidence value and the second preset confidence value, it is determined as a second-level risk in the risk level label; if it is a second-level risk, a yellow warning is triggered, a risk warning is generated, a patrol task is assigned, and the on-site person in charge conducts a risk review; when the first probability value is between the second threshold and the third threshold and the confidence value is between the second preset confidence value and the third preset confidence value, it is determined as a third-level risk in the risk level label; if it is a third-level risk, a green warning is triggered, only monitoring records are retained, and a periodic risk assessment is carried out.

[0012] As a preferred solution of the machine learning-based fire safety assessment method of the present invention, wherein: feature fusion is performed on the first risk level label and the first time series prediction result to construct a fusion feature matrix, and the fusion feature matrix is input into a random forest model for parameter learning, including: performing one-hot encoding conversion on the risk level label to generate a 3D label vector; performing channel splicing on the label vector and the time series prediction result to generate an initial fusion feature; performing normalization processing on the initial fusion feature, and calculating the weight coefficient of each feature dimension by using an attention mechanism; performing weighted combination on the initial fusion feature according to the weight coefficient to construct a fusion feature matrix, where the rows of the fusion feature matrix represent the number of samples, and the columns represent the sum of the risk label dimension and the time series prediction result dimension; inputting the fusion feature matrix into a random forest model, and using the Gini coefficient as an evaluation index for node splitting, and constructing a number of decision trees through random feature selection and sample sampling, where the decision trees divide the fusion feature matrix by means of recursive binary splitting, and corresponding prediction results are obtained at the leaf nodes.

[0013] As a preferred embodiment of the machine learning-based fire safety assessment method of the present invention, the method includes: inputting the first image feature vector into a pre-trained deep convolutional neural network model to obtain a risk level label, and at the same time inputting the first standardized data into a long short-term memory network model to obtain a first time series prediction result, including: inputting the first image feature vector into the deep convolutional neural network model for feature extraction and non-linear transformation, performing deep learning processing on the first image feature vector, and outputting a risk level label, where the risk level label is divided into first-level risk, second-level risk, and third-level risk; reorganizing the first standardized data into a sample sequence according to the time series, constructing a long short-term memory network model, and training the long short-term memory network model using an Adam optimizer, where the long short-term memory network model includes two LSTM layers; inputting the sample sequence into the trained long short-term memory network model, and performing a probability distribution transformation on the prediction output of the long short-term memory network model to obtain a first time series prediction result.

[0014] As a preferred embodiment of the machine learning-based fire safety assessment method of the present invention, the method for obtaining the first standardized data is as follows: collecting image data in the fire protection site through cameras installed in various areas of the fire protection site; collecting equipment operation data through fire protection equipment sensors, and at the same time collecting environmental monitoring data through environmental sensors; preprocessing the image data, the equipment operation data, and the environmental monitoring data; based on the preprocessed image data, using a ResNet50 backbone network to extract deep features, and extracting the feature map output by the fifth convolutional layer; reducing the dimension of the feature map through a global average pooling layer to obtain a first image feature vector; according to the preprocessed equipment operation data, mapping the numerical range to the [0,1] interval, and at the same time making the data mean of the preprocessed environmental monitoring data 0 and the variance 1; using a feature splicing method to splice the preprocessed equipment operation data and environmental monitoring data according to the time stamp to form the first standardized data.

[0015] As a preferred embodiment of the machine learning-based fire safety assessment method of the present invention, the image data includes fire passage images, fire protection equipment layout point images, and safety exit images; the equipment operation data includes fire hydrant water pressure values, sprinkler system working states, and fire door switch states; the environmental monitoring data includes temperature values, smoke concentration values, and carbon monoxide concentration values.

[0016] In a second aspect, an embodiment of the present invention provides a fire safety assessment system based on machine learning, which includes: a collection module, configured to collect image data, equipment operation data, and environmental monitoring data of a fire prevention site, extract feature vectors of the image data to obtain first image feature vectors, and perform standardization processing on the equipment operation data and environmental monitoring data to obtain first standardized data; a feature extraction module, configured to input the first image feature vectors into a pre-trained deep convolutional neural network model to obtain risk level labels, and at the same time input the first standardized data into a long short-term memory network model to obtain a first time series prediction result; a data standardization module, configured to perform feature fusion on the first risk level labels and the first time series prediction results, construct a fusion feature matrix, and input the fusion feature matrix into a random forest model for parameter learning; a prediction module, configured to input newly collected fire prevention data into the learned random forest model for safety level prediction, obtain a prediction result of fire safety, and generate a risk warning message.

[0017] In a third aspect, an embodiment of the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and: when the computer program instructions are executed by the processor, the steps of the fire safety assessment method based on machine learning as described in the first aspect of the present invention are implemented.

[0018] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, and: when the computer program instructions are executed by the processor, the steps of the fire safety assessment method based on machine learning as described in the first aspect of the present invention are implemented.

[0019] The beneficial effects of the present invention are as follows: By collecting image data, equipment operation data, and environmental monitoring data of a fire prevention site, and combining a deep convolutional neural network, a long short-term memory network, and a random forest model, a comprehensive assessment and prediction of fire safety are realized; the standardization processing of image features and equipment and environmental data provides an accurate data basis for subsequent models; the use of a deep learning model to extract image features and the prediction of equipment and environmental change trends through a time series model improve the accuracy and real-time performance of the assessment; feature fusion and the random forest model enhance the robustness and reliability of the assessment results, can generate risk warning messages in real time, help quickly respond to fire safety hazards, and significantly improve fire management efficiency. Description of the Drawings

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. Among them:

[0021] Figure 1 This is a flow chart of the fire safety assessment method based on machine learning in Example 1. DETAILED DESCRIPTION

[0022] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0023] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0025] Example 1

[0026] Reference Figure 1 , which is the first embodiment of the present invention, and provides a fire safety assessment method based on machine learning, comprising:

[0027] S1: Collect image data, equipment operation data and environmental monitoring data of fire-fighting places, extract features from the image data to obtain a first image feature vector, and standardize the equipment operation data and environmental monitoring data to obtain first standardized data.

[0028] Specifically, image data is collected through cameras installed in fire-fighting places, where the image data includes images of fire passages, images of fire-fighting equipment layout points and images of emergency exits; equipment operation data is collected through fire-fighting equipment sensors, where the equipment operation data includes fire hydrant water pressure values, sprinkler system working status and fire door switch status; environmental monitoring data is collected through environmental sensors, where the environmental monitoring data includes temperature values, smoke concentration values ​​and carbon monoxide concentration values.

[0029] Furthermore, the image data was preprocessed, including size normalization to 256 × 256 pixels and RGB channel standardization.

[0030] Further, based on the preprocessed image data, the ResNet50 backbone network is used to extract deep features, and a 2048-dimensional feature map output by the fifth convolutional layer is extracted; the feature map is reduced in dimension through a global average pooling layer to obtain a 128-dimensional first image feature vector.

[0031] Specifically, the minimum-maximum normalization process is performed on the device operation data to map the numerical range to the [0, 1] interval; the Z-score normalization process is performed on the environmental monitoring data to make the data mean 0 and the variance 1; the feature splicing method is used to splice the preprocessed device operation data and environmental monitoring data according to the time stamp to form the first standardized data.

[0032] S2: Input the first image feature vector into a pre-trained deep convolutional neural network model to obtain a risk level label. At the same time, input the first standardized data into a long short-term memory network model to obtain a first time series prediction result.

[0033] Specifically, input the first image feature vector into the deep convolutional neural network model for feature extraction and non-linear transformation, perform deep learning processing on the first image feature vector, and output a risk level label, where the risk level label is divided into first-level risk, second-level risk, and third-level risk.

[0034] It should be noted that the deep convolutional neural network model includes 5 convolutional blocks, each convolutional block consists of two 3×3 convolutional layers and one max pooling layer; the last layer of the deep convolutional neural network model is a fully connected layer, and the output dimension is 3, corresponding to the risk level labels of first-level risk, second-level risk, and third-level risk respectively; the deep convolutional neural network model is fine-tuned and trained based on the cross-entropy loss function, the training batch size is set to 32, and the learning rate is set to 0.001.

[0035] Further, the first standardized data is reorganized into a sample sequence according to the time series, a long short-term memory network model is constructed, and the Adam optimizer is used to train the long short-term memory network model, where the long short-term memory network model includes two LSTM layers.

[0036] It should be noted that the time window length of the sample sequence is set to 24, and the sliding step size is set to 1; the long short-term memory network model contains two LSTM layers, the number of neurons in each layer is 64, and the dropout rate is set to 0.2; the input dimension of the long short-term memory network model is the same as the feature dimension of the first standardized data, and the output dimension is the prediction time step, and the prediction time step is set to 6.

[0037] Illustratively, the number of training rounds is set to 100 rounds. During the training process, the root mean square error is calculated as an evaluation index. When the evaluation index does not decrease significantly for 5 consecutive rounds, the early stopping mechanism is triggered.

[0038] Further, input the sample sequence into the trained long short-term memory network model, and perform probability distribution conversion on the prediction output of the long short-term memory network model to obtain the first time series prediction result.

[0039] S3: Perform feature fusion on the first risk level label and the first time series prediction result, construct a fusion feature matrix, and input the fusion feature matrix into a random forest model for parameter learning.

[0040] Specifically, the construction method of the fusion feature matrix is as follows: perform one-hot encoding conversion on the risk level label to generate a 3D label vector; perform channel concatenation on the label vector and the time series prediction result to generate an initial fusion feature; perform normalization processing on the initial fusion feature, and use the attention mechanism to calculate the weight coefficients of each feature dimension; perform weighted combination on the initial fusion feature based on the weight coefficients to construct a fusion feature matrix, where the rows of the fusion feature matrix represent the number of samples, and the columns represent the sum of the risk label dimension and the time series prediction result dimension.

[0041] It should be noted that the weight coefficients reflect the importance of different features for safety assessment; the 3D label vector includes that one-hot encoding uses [1, 0, 0] to represent the first-level risk, [0, 1, 0] to represent the second-level risk, and [0, 0, 1] to represent the third-level risk.

[0042] Further, input the fusion feature matrix into a random forest model, and use the Gini coefficient as the evaluation index for node splitting. Construct several decision trees through random feature selection and sample sampling. The decision trees divide the fusion feature matrix through recursive binary splitting and obtain corresponding prediction results at the leaf nodes.

[0043] It should be noted that the construction method of the random forest model is as follows: the random forest model includes 100 decision trees, and the maximum depth of each decision tree is set to 10; during the splitting process of each decision tree, randomly select a feature subset to search for the optimal splitting point, where the size of the feature subset is set to the square root of the fusion feature dimension; use the Gini coefficient as the evaluation index for node splitting, and when the number of node samples is less than 5 or all samples belong to the same category, stop splitting.

[0044] S4: Input the newly collected fire data into the learned random forest model for safety level prediction, obtain the prediction result of fire safety, and generate a risk warning message.

[0045] Specifically, preprocess the newly collected fire data to construct a prediction feature matrix.

[0046] Preferably, feature extraction is performed on the newly collected image data to obtain a second image feature vector, and the newly collected equipment operation data and environmental monitoring data are standardized to obtain second standardized data; the second image feature vector and the second standardized data are input into the deep convolutional neural network model and the long short-term memory network model in step S2 to obtain a second risk level label and a second time series prediction result, respectively; and a prediction feature matrix is ​​constructed according to the feature fusion scheme in step S3.

[0047] Furthermore, the prediction feature matrix is ​​input into the learned random forest model, and based on the voting results of each decision tree, the probability distribution of the sample belonging to the security level is calculated; the level corresponding to the first probability value in the probability distribution is determined as the result of the security level prediction, and the confidence value of the prediction result is recorded at the same time, wherein the confidence value is equal to the difference between the first probability value and the second probability; the security level prediction result is judged based on the risk threshold, and the warning level is set according to the judgment result and the confidence value.

[0048] Preferably, when the first probability value is greater than the first threshold value, and the confidence value is greater than the first preset confidence value, it is determined to be a level one risk in the risk level label; if it is a level one risk, a red warning is immediately triggered, and the fire-fighting equipment and emergency response mechanism are linked; when the first probability value is between the first threshold value and the second threshold value, and the confidence value is between the first preset confidence value and the second preset confidence value, it is determined to be a level two risk in the risk level label; if it is a level two risk, a yellow warning is triggered, a risk warning is generated, an inspection task is assigned, and the risk is reviewed by the on-site person in charge; when the first probability value is between the second threshold value and the third threshold value, and the confidence value is between the second preset confidence value and the third preset confidence value, it is determined to be a level three risk in the risk level label; if it is a level three risk, a green warning is triggered, only the monitoring records are retained, and periodic risk assessments are performed.

[0049] It should be noted that the risk threshold includes the first threshold, the second threshold, the third threshold and the fourth threshold; the first threshold is based on the statistical analysis of the first-level risk situation in the historical data, and through the analysis of multiple safety assessments and event data of fire-fighting places, the probability value that can best distinguish between the first-level and second-level risk events is determined; the second threshold is based on the frequency of occurrence of the second-level risk, and through the sampling and analysis of historical events, the interval of the moderate risk probability value under different environmental conditions is obtained; the third threshold is based on the observation results of the third-level risk scenario, taking into account the probability of occurrence of low-risk events under certain safety operations in fire-fighting places, and its reasonable range is determined; the fourth threshold is based on model verification in laboratory tests and simulation environments. Through repeated verification and experiments of actual test results, it is determined that the probability of misjudgment can be effectively reduced in the safety assessment model.

[0050] In this embodiment, the specific implementation method is that several risk thresholds and confidence thresholds need to be set in advance, which are used for the determination of primary risks, secondary risks, and tertiary risks respectively. For example, the first threshold is set to 0.8, the second threshold is set to 0.6, and the third threshold is set to 0.4, corresponding to the probability determination criteria for primary risks, secondary risks, and tertiary risks respectively. At the same time, the first preset confidence value is set to 0.9, the second preset confidence value is set to 0.7, and the third preset confidence value is set to 0.5 to ensure that the risk determination has sufficient reliability and accuracy.

[0051] In practical applications, when the first probability value is 0.85 and the confidence value is 0.92, since the first probability value is greater than the first threshold (0.8) and the confidence value is greater than the first preset confidence value (0.9), it is determined as a primary risk, that is, a high-risk state. At this time, the system immediately triggers a red alert, and at the same time, it links the fire-fighting equipment and the emergency response mechanism to ensure a rapid response in the primary risk state and minimize the harm caused by emergencies. When the first probability value is 0.75 and the confidence value is 0.8, since the first probability value is between the first threshold (0.8) and the second threshold (0.6), and the confidence value is between the first preset confidence value (0.9) and the second preset confidence value (0.7), it is determined as a secondary risk, that is, a medium-risk state. At this time, the system triggers a yellow alert, generates a risk warning, and assigns a patrol task, and the on-site person in charge conducts a risk review and further disposal to ensure sufficient verification and response when the potential risk is relatively high. When the first probability value is 0.55 and the confidence value is 0.6, since the first probability value is between the second threshold (0.6) and the third threshold (0.4), and the confidence value is between the second preset confidence value (0.7) and the third preset confidence value (0.5), it is determined as a tertiary risk, that is, a low-risk state. At this time, the system triggers a green alert, only retains the monitoring records, and conducts a periodic risk assessment. Through continuous monitoring and regular analysis, the effective management of the low-risk state can be ensured, and the potential harm caused by risk accumulation can be prevented from escalating.

[0052] Furthermore, by reasonably setting the risk thresholds and confidence thresholds and combining with actual data for dynamic determination, different risk levels can be effectively distinguished, ensuring timely response in the high-risk state, sufficient verification in the medium-risk state, and periodic assessment in the low-risk state, thereby realizing the intelligent and refined security management.

[0053] Specifically, perform a time-series analysis on the early warning results within a continuous time window to identify the changing trend of the risk level; when high-risk early warnings continuously appear in a certain area, increase the monitoring frequency of this area and shorten the data acquisition interval.

[0054] In summary, the present invention realizes the comprehensive evaluation and prediction of fire safety by collecting image data, equipment operation data, and environmental monitoring data of fire protection sites, and combining deep convolutional neural networks, long short-term memory networks, and random forest models; the standardized processing of image features, equipment, and environmental data provides an accurate data basis for subsequent models; the use of deep learning models to extract image features and the prediction of equipment and environmental change trends through time series models improve the accuracy and real-time performance of the evaluation; feature fusion and random forest models enhance the robustness and reliability of the evaluation results, can generate risk warning information in real time, help quickly respond to fire safety hazards, and significantly improve fire management efficiency.

[0055] Embodiment 2

[0056] This is the second embodiment of the present invention. This embodiment also provides a machine learning-based fire safety evaluation system, including:

[0057] A collection module, configured to collect image data, equipment operation data, and environmental monitoring data of a fire protection site, extract features from the image data to obtain a first image feature vector, and perform standardized processing on the equipment operation data and environmental monitoring data to obtain first standardized data;

[0058] A feature extraction module, configured to input the first image feature vector into a pre-trained deep convolutional neural network model to obtain a risk level label, and at the same time input the first standardized data into a long short-term memory network model to obtain a first time series prediction result;

[0059] A data standardization module, configured to perform feature fusion on the first risk level label and the first time series prediction result, construct a fusion feature matrix, and input the fusion feature matrix into a random forest model for parameter learning;

[0060] A prediction module, configured to input newly collected fire protection data into the learned random forest model for safety level prediction, obtain a prediction result of fire safety, and generate a risk warning information.

[0061] It should be noted that the technical solution of the machine learning-based fire safety evaluation system belongs to the same concept as the technical solution of the above-mentioned machine learning-based fire safety evaluation method. For the details not described in detail in the technical solution of the machine learning-based fire safety evaluation system in this embodiment, reference can be made to the description of the technical solution of the above-mentioned machine learning-based fire safety evaluation method.

[0062] The above-mentioned each unit module can be embedded in the processor of the computer device in a hardware form or be independent of it, or can be stored in the memory of the computer device in a software form, so that the processor can call and execute the operations corresponding to the above-mentioned each module.

[0063] This embodiment also provides an electronic device, which includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a multi-task edge computing resource scheduling method. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a button, a trackball, or a touchpad set on the housing of the computer device, or an external keyboard, a touchpad, or a mouse, etc.

[0064] This embodiment also provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by the processor, it implements the method proposed in the above embodiment.

[0065] The storage medium proposed in this embodiment and the method proposed in the above embodiment belong to the same inventive concept. For technical details not described in detail in this embodiment, reference can be made to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0066] From the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be implemented by means of software and necessary general-purpose hardware. Of course, it can also be implemented by hardware, but in many cases, the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as a floppy disk, a read-only memory (ROM), a random access memory (RAM), a flash memory (FLASH), a hard disk, or an optical disc of a computer, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the method of the embodiments of the present invention.

[0067] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not restrictive. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

[0068] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present application can be implemented in various computer languages.

[0069] The present application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows 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 the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the specified functions in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0070] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implements the specified functions in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0071] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the specified functions in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0072] Although the preferred embodiments of the present application have been described, additional changes and modifications can be made to these embodiments by those skilled in the art once they learn of the basic inventive concept. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications that fall within the scope of the present application.

[0073] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application is also intended to include these modifications and variations.

Claims

1. A fire safety assessment method based on machine learning, characterized in that: Including, Collecting image data, equipment operation data, and environmental monitoring data of a fire protection site, extracting features from the image data to obtain a first image feature vector, and performing normalization processing on the equipment operation data and environmental monitoring data to obtain first normalized data; Inputting the first image feature vector into a pre-trained deep convolutional neural network model to obtain a risk level label. At the same time, inputting the first normalized data into a long short-term memory network model to obtain a first time series prediction result; Performing feature fusion on the first risk level label and the first time series prediction result, constructing a fusion feature matrix, and inputting the fusion feature matrix into a random forest model for parameter learning; Inputting newly collected fire protection data into the learned random forest model for safety level prediction, obtaining a prediction result of fire safety, and generating a risk warning message.

2. The machine learning-based fire safety assessment method according to claim 1, wherein: Inputting newly collected fire protection data into the learned random forest model for safety level prediction, obtaining a prediction result of fire safety, and generating a risk warning message, including: Preprocessing the newly collected fire protection data to construct a prediction feature matrix; Inputting the prediction feature matrix into the learned random forest model, and calculating the probability distribution of the sample belonging to the safety level based on the voting results of each decision tree; Determining the level corresponding to the first probability value in the probability distribution as the result of the safety level prediction, and at the same time recording the confidence value of the prediction result, where the confidence value is equal to the difference between the first probability value and the second probability; Judging the safety level prediction result based on a risk threshold, and setting a warning level according to the judgment result and the confidence value.

3. The machine learning-based fire safety assessment method according to claim 2, characterized in that: The risk threshold includes a first threshold, a first threshold, a first threshold, and a first threshold; also including, When the first probability value is greater than the first threshold and the confidence value is greater than the first preset confidence value, it is determined as a first-level risk in the risk level label; if it is a first-level risk, immediately trigger a red warning and link the fire protection equipment and the emergency response mechanism; When the first probability value is between the first threshold and the second threshold and the confidence value is between the first preset confidence value and the second preset confidence value, it is determined as a second-level risk in the risk level label; if it is a second-level risk, trigger a yellow warning, generate a risk warning, assign a patrol task, and have the on-site person in charge conduct a risk review; When the first probability value is between the second threshold and the third threshold and the confidence value is between the second preset confidence value and the third preset confidence value, it is determined as a third-level risk in the risk level label; if it is a third-level risk, trigger a green warning, only retain the monitoring record, and conduct a periodic risk assessment.

4. The machine learning-based fire safety assessment method according to claim 3, wherein: Performing feature fusion on the first risk level label and the first time series prediction result, constructing a fusion feature matrix, and inputting the fusion feature matrix into a random forest model for parameter learning, including: Performing one-hot encoding conversion on the risk level label to generate a 3D label vector; Performing channel concatenation on the label vector and the time series prediction result to generate an initial fusion feature; Performing normalization processing on the initial fusion feature, and calculating the weight coefficient of each feature dimension using an attention mechanism; Perform weighted combination on the initial fusion features according to the weight coefficients to construct a fusion feature matrix, where the rows of the fusion feature matrix represent the number of samples, and the columns represent the sum of the risk label dimension and the time series prediction result dimension; Input the fusion feature matrix into a random forest model, and use the Gini coefficient as the evaluation index for node splitting. Construct a number of decision trees through random feature selection and sample sampling. The decision trees divide the fusion feature matrix through recursive binary splitting and obtain corresponding prediction results at the leaf nodes.

5. The machine learning-based fire safety assessment method according to claim 4, wherein: Input the first image feature vector into a pre-trained deep convolutional neural network model to obtain a risk level label. At the same time, input the first standardized data into a long short-term memory network model to obtain a first time series prediction result, including: Input the first image feature vector into a deep convolutional neural network model for feature extraction and non-linear transformation, perform deep learning processing on the first image feature vector, and output a risk level label, where the risk level label is divided into first-level risk, second-level risk, and third-level risk; Reorganize the first standardized data into a sample sequence according to the time series, construct a long short-term memory network model, and use the Adam optimizer to train the long short-term memory network model, where the long short-term memory network model includes two LSTM layers; Input the sample sequence into the trained long short-term memory network model, and perform probability distribution conversion on the prediction output of the long short-term memory network model to obtain a first time series prediction result.

6. The machine learning-based fire safety assessment method according to claim 5, characterized in that: The method for obtaining the first standardized data is as follows: Collect image data in the fire protection site through cameras installed in each area of the fire protection site; Collect equipment operation data through fire protection equipment sensors, and at the same time collect environmental monitoring data through environmental sensors; Preprocess the image data, the equipment operation data, and the environmental monitoring data; Based on the preprocessed image data, use the ResNet50 backbone network to extract deep features, and extract the feature map output by the fifth convolutional layer; Reduce the dimension of the feature map through a global average pooling layer to obtain a first image feature vector; According to the preprocessed equipment operation data, map the numerical range to the interval [0,1], and at the same time make the data mean of the preprocessed environmental monitoring data 0 and the variance 1; Use the feature splicing method to splice the preprocessed equipment operation data and environmental monitoring data according to the time stamp to form the first standardized data.

7. The machine learning-based fire safety assessment method according to claim 5, characterized in that: The image data includes fire passage images, fire protection equipment layout point images, and safety exit images; the equipment operation data includes fire hydrant water pressure values, sprinkler system working status, and fire door switch status; the environmental monitoring data includes temperature values, smoke concentration values, and carbon monoxide concentration values.

8. A fire safety assessment system based on machine learning, based on the fire safety assessment method based on machine learning according to any one of claims 1 to 7, characterized in that: Including: A collection module for collecting image data, equipment operation data, and environmental monitoring data of a fire protection site, extracting first image feature vectors from the image data, and performing standardization processing on the equipment operation data and environmental monitoring data to obtain first standardized data; A feature extraction module, configured to input the first image feature vector into a pre-trained deep convolutional neural network model to obtain a risk level label, and at the same time input the first normalized data into a long short-term memory network model to obtain a first time series prediction result; A data normalization module, configured to perform feature fusion on the first risk level label and the first time series prediction result, construct a fusion feature matrix, and input the fusion feature matrix into a random forest model for parameter learning; A prediction module, configured to input newly collected fire protection data into the learned random forest model for safety level prediction, obtain a prediction result of fire safety, and generate a risk warning message.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that: When the processor executes the computer program, the steps of the machine learning-based fire safety assessment method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, the steps of the machine learning-based fire safety assessment method according to any one of claims 1 to 7 are implemented.

Images