A method and system for monitoring and warning of fire in ancient cultural relics buildings

Abstract

The present invention relates to the technical field of fire warning, and particularly relates to a method and system for monitoring and warning of fire in cultural relics and ancient buildings. The method includes: constructing a hierarchical classification model; training the hierarchical classification model using a training set; classifying the fire situation data of cultural relics and ancient buildings collected in real time using the trained hierarchical classification model to obtain a fire situation level, and giving a warning. That is, the solution of the present invention can accurately judge the fire situation level of cultural relics and ancient buildings.

Description

Technical Field

[0001] The present invention relates to the technical field of fire warning. More specifically, the present invention relates to a method and system for monitoring and warning of fire in cultural relics ancient buildings. Background Art

[0002] Cultural relics ancient buildings not only have important historical, artistic, architectural and archaeological values, but also carry rich cultural connotations. Among them, most cultural relics ancient buildings are mainly made of wood. The wood is dry after years and has a low fire resistance rating. After a fire breaks out, the burning speed is extremely fast. Therefore, the occurrence of a fire not only causes damage to cultural relics ancient buildings, but also leads to immeasurable cultural and historical losses. In particular, cultural relics ancient buildings are usually located in tourist areas with dense crowds, and often lack a perfect fire emergency management system. Therefore, once a fire breaks out, it is extremely difficult to extinguish the fire and evacuate people, and the destructiveness of the fire is more serious. Therefore, fire monitoring is crucial for their protection.

[0003] In recent years, with the development of technologies such as the Internet of Things (IoT), big data, and artificial intelligence, fire monitoring and warning technologies have gradually been widely applied.

[0004] In related technologies, for example, a Chinese patent document with the authorization announcement number CN205177047U discloses a fire situation timely alarm system, which mainly monitors through a plurality of temperature and smoke sensors and alarm devices arranged in the place for monitoring the fire situation. When a fire breaks out, the temperature and smoke sensors sense abnormal temperature and smoke concentration for warning.

[0005] The above solution only discloses the analysis of the fire situation by detecting the temperature and smoke concentration, and does not disclose the classification of the fire situation level. For the classification of the fire situation level, generally, the detected temperature and smoke concentration are compared with the set fire situation levels to determine the fire situation level when a fire breaks out.

[0006] However, the occurrence of a fire changes dynamically over time. Since the dynamic change speeds of the occurrence of a fire are different, it is difficult to capture accurate data for judging the fire situation level, resulting in possible inaccurate judgment of the fire situation level. Summary of the Invention

[0007] The purpose of the present invention is to provide a method and system for monitoring and warning of fire in cultural relics ancient buildings to solve the problem that there may be inaccuracies in judging the fire situation level in the prior art; for this purpose, the present invention provides solutions in the following two aspects.

[0008] In the first aspect, a method for monitoring and warning of fire in cultural relics ancient buildings provided by the present invention includes:

[0009] Construct a level classification model; train the level classification model using a training set;

[0010] Use the trained level classification model to classify the fire data of cultural relics and ancient buildings collected in real time, obtain the fire level, and issue a warning;

[0011] The training set is as follows: Obtain a plurality of historical fire data and the fire level labels of each historical fire data; the historical fire data includes temperature data or smoke concentration data;

[0012] Set a plurality of windows, use each window to divide all historical fire data, obtain multiple sequence segments of each historical fire data, and obtain a sequence segment set corresponding to each window;

[0013] Construct the objective function under each window; the objective function is positively correlated with the difference and the degree of chaos; the difference is the ratio of the sum of the first variance and the second variance of the first similarity between any two sequence segments in the sequence segment set, and the second variance is the variance of the second similarity between any two sequence segments under any fire level; the degree of chaos is the mean value of the relative entropy between the difference sequences of the corresponding time series segments in all two fire levels;

[0014] Construct a training set with all the sequence segments corresponding to the window with the maximum value of the objective function and the fire level labels corresponding to each sequence segment.

[0015] The solution of the present invention divides the historical fire data by using a plurality of windows, analyzes the sequence segments under each window, obtains the difference and the degree of chaos, judges whether the corresponding window is a suitable window, and further obtains the sequence segments that can characterize the fire change, so as to train the level classification model, thereby obtaining an accurate level classification model and realizing the fire level classification of the current fire data.

[0016] Optionally, the setting of a plurality of windows includes:

[0017] Set an initial window, and update the initial window according to a set step length to obtain an updated window, and so on, to obtain a plurality of windows, where the maximum value of the window size in the plurality of windows is less than the length of the historical fire data, and the minimum value of the window size is greater than 1.

[0018] The above solution provides support for obtaining a training set by setting a plurality of windows and deeply analyzing the fire data under each window.

[0019] Optionally, the objective function is: ;

[0020] Wherein, is the objective function corresponding to the m-th window, is the weight of the difference under the m-th window, is the difference under the m-th window, is the weight of the degree of chaos under the m-th window, is the degree of chaos under the m-th window.

[0021] The above scheme combines the differences and degrees of chaos under each window and can obtain the objective function under each window.

[0022] Optionally, the relative entropy is: ;

[0023] where, is the relative entropy between the k-th fire danger level and the t-th fire danger level, is the probability that the difference sequence in the j-th time series segment appears under the k-th fire danger level, is the probability that the difference sequence in the j-th time series segment appears under the t-th fire danger level, is the logarithmic function, and n is the number of sequence segments under the k fire danger levels or the t-th fire danger level.

[0024] The above scheme characterizes the chaos between different fire danger levels through relative entropy.

[0025] Optionally, the first variance is: ;

[0026] where, is the average similarity of all the first similarities under the m-th window, , is the total number of sequence segments in the sequence segment set under the m-th window, is the combination number of I sequence segments under the m-th window, is the first similarity between the i-th sequence segment and the (i + b)-th sequence segment in the sequence segment set under the m-th window.

[0027] Optionally, the first similarity is the Pearson correlation coefficient or the DTW distance calculated between any two sequence segments in the sequence segment set; the second similarity is the Pearson correlation coefficient or the DTW distance calculated between any two sequence segments under any fire danger level.

[0028] Optionally, it further includes the step of denoising the historical fire data and the fire data respectively.

[0029] The above scheme can obtain accurate historical fire data and fire data.

[0030] Optionally, the grade classification model is a BP network model or an LSTM model.

[0031] In the second aspect, a fire monitoring and early warning system for cultural relics ancient buildings includes:

[0032] a processor;

[0033] A memory stores computer instructions for fire monitoring and early warning of ancient cultural relics buildings. When the computer instructions are run by the processor, the system executes the above-mentioned method for fire monitoring of ancient cultural relics buildings.

[0034] The beneficial effects of the present invention are as follows:

[0035] The solution of the present invention can obtain the data that can represent the fire level in the historical fire data as the training set to train the level classification model, so as to accurately judge the level of the fire data collected in real time subsequently. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] By referring to the drawings and reading the following detailed description, the above and other objects, features, and advantages of the exemplary embodiments of the present invention will become easy to understand. In the drawings, several embodiments of the present invention are shown in an exemplary rather than restrictive manner, and the same or corresponding reference numerals are for the same or corresponding parts, where:

[0037] Figure 1 Schematically shows the flowchart of the steps of a method for fire monitoring and early warning of ancient cultural relics buildings in this embodiment;

[0038] Figure 2 Schematically shows the structural block diagram of a system for fire monitoring and early warning of ancient cultural relics buildings in this embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0039] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0040] Specifically, as Figure 1 shown, a method for fire monitoring and early warning of ancient cultural relics buildings in this embodiment includes the following steps:

[0041] Step S1, obtain a training set and use the training set to train a level classification model.

[0042] In this embodiment, the acquisition of the training set includes steps S11 - S14, specifically:

[0043] Step S11, obtain a historical fire data set and the labels corresponding to each historical fire data in the historical fire data set.

[0044] In one embodiment, sensors are used to collect data in simulated scenarios of different fire levels, obtaining corresponding historical fire data, and further obtaining a historical fire data set; the sensors can be temperature sensors, smoke sensors, or flame sensors.

[0045] Each historical fire data corresponds to a fire level. The fire levels corresponding to different historical fire data are different. The fire levels in this embodiment can be divided into level one, level two, level three, etc. The higher the level, the more serious the fire situation.

[0046] When the sensor is a temperature sensor, the historical fire data is the temperature data within a set time period, and at this time, the temperature data includes the temperatures at multiple sampling moments; when the sensor is a smoke sensor, the historical fire data is the smoke concentration data within a set time period. The set time period is a period of time, which can be specifically set according to the actual situation.

[0047] After obtaining each historical fire data, perform denoising processing on the collected historical fire data to obtain the denoised historical fire data.

[0048] In another embodiment, big data technology can also be used to obtain multiple historical fire data stored historically and label each historical fire data to obtain the corresponding fire level.

[0049] Step S12: Set multiple windows, and use each window to divide all historical fire data to obtain multiple sequence segments of each historical fire data.

[0050] Specifically, in this embodiment, using any window to divide all historical fire data respectively can obtain multiple sequence segments of each historical fire data, and further obtain a sequence segment set (all the sequence segments of all historical fire data under any window constitute a sequence segment set). Multiple windows correspond to multiple sequence segment sets.

[0051] It should be noted that one historical fire data corresponds to one fire level label; for multiple sequence segments of each historical fire data, each sequence segment also corresponds to one fire level label, and the fire level labels of multiple sequence segments of the same historical fire data are the same.

[0052] In this embodiment, the length of the window corresponds to the number of data in the sequence segment.

[0053] In one embodiment, the step of setting multiple windows is: set an initial window, and update the initial window according to a set step length to obtain an updated window, and so on, to obtain multiple windows, where the maximum value of the window size in multiple windows is less than the length of the historical fire data, and the minimum value of the window size is greater than 1.

[0054] In another embodiment, multiple windows can also be directly set according to experience.

[0055] Step S13: Construct the objective function for each window.

[0056] In order to obtain the data within the window that can accurately represent the fire danger level, in this embodiment, not only the static fire data is considered, but also the dynamic change situation of the fire data is considered. Specifically, in this embodiment, according to the sequence segment reflecting the static situation and the difference sequence reflecting the dynamic change situation, the objective function for each window is constructed.

[0057] First, calculate the difference between the sequence segments under all fire danger levels for each window.

[0058] Specifically, the difference is:

[0059] ;

[0060] where is the difference under the m-th window, is the first variance of the first similarity between any two sequence segments in the sequence segment set under the m-th window, is the second variance of all sequence segments under the k-th fire danger level in the m-th window, and N is the number of types of fire danger levels.

[0061] It should be noted that the smaller the variance under each fire danger level, and the larger the variance under all fire danger levels, the greater the difference between the sequence segments under all fire danger levels.

[0062] In this embodiment, the calculation process of the first variance is:

[0063] ;

[0064] where is the first variance of the first similarity between any two sequence segments in the sequence segment set under the m-th window, is the average similarity of all the first similarities under the m-th window, is the total number of sequence segments in the sequence segment set under the m-th window, is the combination number of I sequence segments under the m-th window, is the first similarity between the i-th sequence segment and the (i + b)-th sequence segment in the sequence segment set under the m-th window.

[0065] Among them, the average similarity is: .

[0066] Among them, the first similarity is the Pearson correlation coefficient or DTW distance calculated between any two sequence segments in the sequence segment set.

[0067] The second variance is the variance of the second similarities between any two sequence segments among all sequence segments at any fire danger level calculated. Wherein the second similarity is the Pearson correlation coefficient or the DTW distance between any two sequence segments among all sequence segments at any fire danger level calculated.

[0068] Since the calculation method of the second variance of all sequence segments at the same fire danger level is the same as that of the above-mentioned first variance, it will not be elaborated here too much.

[0069] Secondly, calculate the chaos degree of the difference sequences at different fire danger levels.

[0070] When the historical fire data is temperature data, the difference sequence is the difference between two adjacent temperatures in each sequence segment, that is, the difference between the latter temperature and the former temperature, and thus the difference sequences of each sequence segment can be obtained.

[0071] The process of obtaining the chaos degree in this embodiment is as follows:

[0072] Calculate the difference sequences of all sequence segments at each fire danger level, obtain the probability density of each difference sequence, and based on the probability densities of the difference sequences at different fire danger levels, obtain the chaos degree.

[0073] Specifically, the calculation of the chaos degree of the difference sequences at all fire danger levels is as follows:

[0074] ;

[0075] Wherein, is the chaos degree of the difference sequences at all fire danger levels, is the number of types of fire danger levels, is the relative entropy between the k-th fire danger level and the t-th fire danger level.

[0076] The relative entropy is: ;

[0077] Wherein, is the relative entropy between the k-th fire danger level and the t-th fire danger level, is the probability that the difference sequence of the j-th time series segment appears at the k-th fire danger level, is the probability that the difference sequence of the j-th time series segment appears at the t-th fire danger level, is the logarithmic function, and n is the number of sequence segments at the k-th fire danger level or the t-th fire danger level. Generally, the base of the logarithmic function is 2.

[0078] When is larger, the chaos degree of the difference sequences between two different fire danger levels is larger.

[0079] Then, construct an objective function according to the degree of difference and chaos.

[0080] Among them, the objective function is: ;

[0081] Among them, is the objective function corresponding to the m-th window, is the weight of the difference under the m-th window, is the difference under the m-th window, is the weight of the chaos degree under the m-th window, is the chaos degree under the m-th window.

[0082] In this embodiment, and both take the value of 0.5; of course, as other implementation manners, and can also take the values of 0.4 and 0.6 respectively.

[0083] Step S14, form a training set with all the sequence segments obtained after dividing the window corresponding to the maximum value of the objective function and the fire level labels corresponding to the sequence segments.

[0084] In this embodiment, different windows are set. According to the differences and chaos degrees under each window, the values of the objective function are obtained. The window corresponding to the maximum value of the objective function is the optimal window. That is, the historical fire data in the historical fire dataset can be divided according to the optimal window to obtain all the sequence segments after division.

[0085] It should be noted that determining the maximum value of the above objective function is to determine the optimal window for dividing each historical fire data. Since the optimal window is not only related to the differences between the sequence segments of different fire levels under this optimal window, but also related to the chaos degree of the differential sequences within the window of different fire levels. Therefore, the greater the chaos degree of the change trends of the sequence segments of different fire levels under the optimal window, the greater the difference in the change trends of the data within the sequence segments within this optimal window, and the higher the accuracy of judging the fire level according to the sequence segments corresponding to the optimal window.

[0086] At the same time, after obtaining all the sequence segments after division, the fire level labels of each sequence segment are also obtained, and all the sequence segments and their corresponding fire level labels form a training set.

[0087] After obtaining the training set, train the constructed level classification model.

[0088] In this embodiment, the constructed level classification model is a BP network model or an LSTM (Long Short-Term Memory) model.

[0089] Taking the LSTM model as an example, the specific training process is as follows:

[0090] Input the training set into the LSTM model to output the fire level labels corresponding to the sensor sequences. The loss of the model uses cross-entropy loss, and the gradient descent algorithm is used to update the model parameters. When the model reaches the maximum number of training times or the loss of the model is less than the set threshold, the model stops training to obtain the trained LSTM model.

[0091] Furthermore, the LSTM model can be trained multiple times, and the optimal LSTM model is selected according to the evaluation index precision rate of the model (where the evaluation index precision rate is the proportion of the number of correct predictions of the monitoring model in the total number).

[0092] Step S2: Use the trained level classification model to classify the fire data of the cultural relics ancient buildings collected in real time, obtain the fire level, and issue an early warning.

[0093] In this embodiment, the fire data detected in real time is input into the trained level classification model to output the fire level and issue an alarm. The alarm can be to send the fire level to the relevant department to remind the relevant department, and the relevant department makes corresponding arrangements according to the fire level to avoid losses caused by incorrect judgment of the fire level.

[0094] The solution of the present invention obtains an accurate training set that can characterize the fire level, trains the level classification model, and obtains the trained level classification model to judge the fire level of subsequent fire data.

[0095] The present invention also provides a fire monitoring and early warning system for cultural relics ancient buildings. As Figure 2 shown, the system includes a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, a fire monitoring and early warning method for cultural relics ancient buildings according to the above of the present invention is implemented.

[0096] The system also includes other components well known to those skilled in the art such as a communication bus and a communication interface. Their settings and functions are known in the art, so they will not be described in detail here.

[0097] In the present invention, the foregoing memory may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium may be any suitable magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory (RRAM), a dynamic random access memory (DRAM), a static random access memory (SRAM), an enhanced dynamic random access memory (EDRAM), a high-bandwidth memory (HBM), a hybrid memory cube (HMC), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of the device or accessible or connectable to the device. Any application or module described in the present invention may be implemented by computer-readable / executable instructions stored or otherwise maintained by such a computer-readable medium.

[0098] In the description of this specification, "a plurality of" means at least two, for example, two, three, or more, etc., unless otherwise specifically defined.

[0099] Although this specification has shown and described multiple embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will think of many changes, alterations, and alternative ways without departing from the spirit and scope of the present invention. It should be understood that various alternatives to the embodiments of the present invention described herein may be employed in practicing the present invention.

Claims

1. A method for monitoring and warning of fire in cultural relics ancient buildings, characterized in that, Including: Construct a hierarchical classification model; Use the training set to train the hierarchical classification model; Use the trained hierarchical classification model to classify the fire situation data of cultural relics ancient buildings collected in real time, obtain the fire situation level, and issue a warning; The training set is: obtain a plurality of historical fire situation data and the fire situation level labels of each historical fire situation data; the historical fire situation data includes temperature data or smoke concentration data; Set a plurality of windows, use each window to divide all historical fire situation data, obtain a plurality of sequence segments of each historical fire situation data, and obtain a sequence segment set corresponding to each window; Construct an objective function under each window; the objective function is positively correlated with the difference and the degree of chaos; the difference is the ratio of the sum of the first variance and the second variance of the first similarity between any two sequence segments in the sequence segment set, and the second variance is the variance of the second similarity between any two sequence segments under any fire situation level; the degree of chaos is the mean of the relative entropy between the difference sequences of the corresponding time series segments in all two fire situation levels; Construct a training set from all the sequence segments corresponding to the window with the maximum value of the objective function and the fire situation level labels corresponding to each sequence segment; The objective function is as follows: ; Among them, is the objective function corresponding to the m-th window, is the weight of the difference under the m-th window, is the difference under the m-th window, is the weight of the degree of chaos under the m-th window, is the degree of chaos under the m-th window.

2. The method for monitoring and warning of fire in ancient cultural relics buildings according to claim 1, wherein The setting of a plurality of windows includes: Set an initial window, and update the initial window according to a set step size to obtain an updated window, and so on, to obtain a plurality of windows, where the maximum value of the window size in the plurality of windows is less than the length of the historical fire situation data, and the minimum value of the window size is greater than 1.

3. The method for monitoring and warning of fire in ancient cultural relics buildings according to claim 1, characterized in that, The relative entropy is as follows: ; Among them, is the relative entropy between the k-th fire danger level and the t-th fire danger level, is the probability that the difference sequence in the j-th time series segment appears at the k-th fire danger level, is the probability that the difference sequence in the j-th time series segment appears at the t-th fire danger level, is the logarithmic function, and n is the number of sequence segments at the k fire danger levels or the t-th fire danger level.

4. A method for monitoring and warning of fire in cultural relics ancient buildings according to claim 1, characterized in that, The first variance is as follows: ; Among them, is the average similarity of all the first similarities under the m-th window, , is the total number of sequence segments in the sequence segment set under the m-th window, is the combination number of I sequence segments under the m-th window, is the first similarity between the i-th sequence segment and the (i + b)-th sequence segment in the sequence segment set under the m-th window.

5. A method for monitoring and warning of fire in ancient cultural relics buildings according to claim 4, characterized in that, The first similarity is the Pearson correlation coefficient or the DTW distance calculated between any two sequence segments in the sequence segment set; the second similarity is the Pearson correlation coefficient or the DTW distance calculated between any two sequence segments under any fire situation level.

6. The method for monitoring and warning of fire in cultural relics ancient buildings according to claim 1, wherein, It also includes the step of denoising the historical fire situation data and the fire situation data respectively.

7. A method for monitoring and warning of fire in cultural relics ancient buildings according to claim 1, characterized in that, The hierarchical classification model is a BP network model or an LSTM model.

8. A fire monitoring and early warning system for cultural relics and ancient buildings, characterized in that, Including: A processor; A memory that stores computer instructions for fire situation monitoring and warning of cultural relics ancient buildings. When the computer instructions are run by the processor, the system executes a method for fire situation monitoring and warning of cultural relics ancient buildings according to any one of claims 1-7.

Images