CN117612319A - Alarm information grading early warning method and system based on sensor and picture - Google Patents

Abstract

The invention relates to a sensor and picture-based alarm information grading early warning method and system, which belong to the field of intelligent fire control, and the method comprises the following steps: and acquiring a fire signal through a fire sensor, and processing the fire signal to obtain a fire signal value. And fusing signals of the same kind in the fire signal value set to obtain a similar signal fusion result. And inputting the same kind of signal fusion results into different kinds of signal fusion neural network models to obtain the fire evidence element fusion vector, and obtaining the fire occurrence probability through calculation. And alarming and grading early warning are carried out according to the occurrence probability of the fire. The on-site state is obtained through picture identification, and the alarm demand is obtained according to the on-site state. The fire safety monitoring system can help fire safety related personnel to timely acquire alarm information such as a fire automatic alarm system, an electric appliance fire monitoring system, a combustible gas alarm and the like, know fire conditions of an alarm point location site, reduce police resource waste caused by fire false alarm and improve alarm information disposal efficiency.

Description

A hierarchical early warning method and system for alarm information based on sensors and pictures

Technical field

The invention belongs to the field of smart fire protection, and specifically relates to a hierarchical early warning method and system for alarm information based on sensors and pictures.

Background technique

Fire accidents have occurred frequently in recent years. Due to insufficient timely acquisition of fire information, false fire alarms and delayed fire alarms usually occur, causing fire outbreaks to cause heavy property losses and casualties. Smart fire protection technology is a technology that fuses multiple sensor signals and then determines the probability of fire based on the fused signal, and then uses picture recognition to verify the fire situation. This technology is used in automatic fire alarm systems, electrical fire monitoring systems, and combustible gas alarms. and other fields. Help fire safety personnel detect fires promptly and accurately. Traditional fire identification technologies based on sensors and image recognition often fail to detect fires in time and cause false alarms.

Contents of the invention

In order to solve the above-mentioned problems existing in the prior art, the present invention provides a hierarchical early warning method and system for alarm information based on sensors and pictures. In the first aspect, the method of the present invention can be implemented through the following technical steps:

S1: Acquire fire signals through fire sensors. The fire sensors include: temperature sensors, residual current sensors, smoke concentration sensors, and CO concentration sensors. The fire signals are processed to obtain fire signal values;

S2: Obtain a fire signal value set D according to the fire signal value. The fire signal value set D includes a signal value set. , the signal value set/> , including n signal values x , the fire signal value set D is expressed as: ,/> , for the signal value set/> The signal values in are weighted and fused to obtain similar signal fusion results/> ;

S3: Fusion results of the similar signals Input different types of signals to fuse the neural network model to obtain fire evidence elements/> , according to the fire evidence element/> Get the fire evidence element fusion vector/> , based on the fire evidence element fusion vector/> Get the probability of fire/> ;

S4: Put the probability of fire occurrence into Sent to the alarm, which is based on the probability of fire occurrence , carry out fire alarm and graded early warning, and the graded early warning includes blue early warning, yellow early warning, orange early warning, and red early warning;

S5: When the alarm sounds, obtain on-site sample pictures based on on-site surveillance cameras, perform preprocessing on the on-site sample pictures to obtain target pictures, the preprocessing includes: color conversion, denoising, filtering and segmentation, and perform preprocessing on the target pictures. Perform static feature extraction to obtain the features of the target picture. The features include circularity features, similarity features, and texture features. The features are converted into feature vectors and input into the classification model according to the decision function. , obtain the on-site status, the sign () represents the sign function, the z represents the eigenvector, and the b represents the bias constant, and the alarm demand is obtained according to the on-site status.

Specifically, the method for obtaining the fire signal value in S1 is:

pass Calculate the temperature signal value, where,/> represents the temperature signal value, ξ represents the sensitivity of the temperature sensor, and a represents the linear relationship between voltage and temperature;

The current sensor obtains the residual current signal value according to Ohm's law;

The smoke concentration sensor passes an analog electrical signal, according to , calculate the smoke concentration signal, where C is the smoke concentration signal value, Rs is the resistance of the sensitive element of the smoke concentration sensor, m, n are the sensitivity constants of the smoke concentration sensor, log() is the logarithmic sign;

The CO concentration sensor uses SnO ₂ with low conductivity as a gas-sensitive material, and uses a high and low temperature cycle detection method to obtain the CO concentration signal value.

Specifically, the specific implementation steps of S2 are:

The signal value set includes a temperature signal value set , smoke signal value set/> , CO concentration signal value set/> , Leakage signal value set/> ,

According to the set of signal values Signal value in/> , n represents the total number of signal values, through , calculate the weighting factor of the signal value, where/> represents the weighting factor,/> Represents the measured variance of the signal value, by/> , the same kind of signal fusion result is calculated, where,/> Indicates the fusion result of the same type of signals,/> Represents the expectation for the set of signal values.

Specifically, the specific implementation steps of S3 are:

The fusion result of the same kind of signals Input the different categories of signal fusion neural network model to obtain the fire evidence element/> , according to/> , calculate the probability vector of the fire evidence element, where/> represents the probability vector;

To calculate the probability of the fire occurring, the calculation formula is: ,

Among them, P ( D ) is the probability of the fire, K is the conflict coefficient between the fire evidence elements, and ⊕ is the probability fusion symbol.

Specifically, the fire alarm and the hierarchical early warning in S4 are:

When the probability of fire occurrence , the alarm does not sound;

When the probability of fire occurrence When, the alarm emits a blue early warning, which indicates the occurrence of a non-fire alarm;

When the probability of fire occurrence When, the alarm issues a yellow warning, which indicates that a fire may have occurred;

When the probability of fire occurrence When, the alarm will issue an orange warning, which indicates that a fire is highly likely to occur;

When the probability of fire occurrence , the alarm will issue a red warning, which indicates that a fire is very likely to occur.

Specifically, the specific steps of preprocessing described in S5 are:

Convert the red, green, and blue color space models of the on-site sample pictures into hue, saturation, and intensity models to obtain the first on-site picture;

Use median filtering method to remove noise in the first scene picture to obtain a second scene picture;

Using the image binarization method, the intensity component image of the 0-255 gray level is represented by 0 and 1 pixels, and the second scene picture is converted into the second scene picture according to the saturation of the red component of the flame in the second scene picture as a threshold It is divided into a target picture and a background picture. The pixel corresponding to the target picture is 1, and the pixel corresponding to the background picture is 0.

Specifically, the specific steps of static feature extraction in S5 are:

according to Extract the roundness feature of the target image, where,/> is the roundness feature,/> is the area of the target image,/> is the perimeter of the target image;

according to , extract the similarity features of the target image, where,/> is the similarity feature,/> is the frame of the target picture,/> is the corresponding frame of the frame;

The texture features include: energy value , entropy/> , contrast/> , relevance/> , the formula for extracting the texture features is:

,

Among them, u represents the horizontal coordinate gray value, v the vertical coordinate gray value, Represents the gray value of the coordinate point,/> Represents the normalized gray level co-occurrence matrix element value.

Specifically, the steps in S5 to obtain the alarm requirement are:

When the decision function When , the on-site status is a fire, and when the decision function When the on-site status is that no fire has occurred, the alarm demand is determined according to the on-site status. When the on-site status is that a fire has occurred, the alarm demand is that an alarm needs to be sent out. When the on-site status is the When no fire occurs, the alarm requirement is that no alarm is required.

In a second aspect, the present invention provides an alarm information hierarchical early warning system based on sensors and pictures, which is run using the method described above, and is characterized by including the following modules:

Sensor module, local fusion module, global fusion module, alarm module, image recognition module;

The sensor module: obtains the fire signal through the fire sensor, and obtains the fire signal value according to the fire signal;

The local fusion module: obtains the fire signal value set according to the fire signal value, and fuses the signal values of the same type in the fire signal value set to obtain the same type of signal fusion result;

The global fusion module: input the same type of signal fusion results into the different types of signal fusion neural network models to obtain the fire evidence element fusion vector, and obtain the fire occurrence probability according to the fire evidence element fusion vector;

The alarm module: sends the fire occurrence probability to the alarm, and the alarm performs the fire alarm and the hierarchical early warning according to the fire occurrence probability;

The picture recognition module: obtains the on-site sample picture according to the on-site surveillance camera, performs the pre-processing on the on-site sample picture to obtain the target picture, and performs the static feature extraction on the target picture to obtain the Features are converted into feature vectors and input into the classification model to obtain the on-site status, and the alarm requirements are obtained based on the on-site status.

The beneficial effects of the present invention are:

(1) The probability of fire occurrence is obtained by fusing signals to help firefighters understand the status of the fire scene in a timely manner, which can not only detect fires in time but also improve the efficiency of alarm handling;

(2) Help fire safety personnel obtain fire information in a timely manner and understand the fire situation at the alarm point.

Description of drawings

In order to facilitate understanding by those skilled in the art, the present invention will be further described below in conjunction with the accompanying drawings.

Figure 1 is a schematic flow chart of the fire alarm and hierarchical early warning method based on sensors and picture recognition according to the present invention;

Figure 2 is a structural block diagram of the fire alarm and hierarchical early warning system based on sensors and picture recognition in the present invention.

Detailed ways

In order to further elaborate on the technical means and effects adopted by the present invention to achieve the intended inventive purpose, the specific implementation manner, structure, features and effects of the present invention are described in detail below with reference to the drawings and preferred embodiments.

Please refer to Figure 1. The process of a hierarchical early warning method for alarm information based on sensors and pictures includes the following steps:

S1: Acquire fire signals through fire sensors. The fire sensors include: temperature sensors, residual current sensors, smoke concentration sensors, and CO concentration sensors. The fire signals are processed to obtain fire signal values;

S2: Obtain a fire signal value set D according to the fire signal value. The fire signal value set D includes a signal value set. , the signal value set/> , including n signal values x , the fire signal value set D is expressed as: ,/> , for the signal value set/> The signal values in are weighted and fused to obtain similar signal fusion results/> ;

S3: Fusion results of the similar signals Input signals from different categories to the neural network model to obtain fire evidence elements/> , according to the fire evidence element/> Get the fire evidence element fusion vector/> , based on the fire evidence element fusion vector/> Get the probability of fire/> ;

S4: Put the probability of fire occurrence into Sent to the alarm, which is based on the probability of fire occurrence , carry out fire alarm and graded early warning, and the graded early warning includes blue early warning, yellow early warning, orange early warning, and red early warning;

S5: When the alarm sounds, obtain on-site sample pictures based on on-site surveillance cameras, perform preprocessing on the on-site sample pictures to obtain target pictures, the preprocessing includes: color conversion, denoising, filtering and segmentation, and perform preprocessing on the target pictures. Perform static feature extraction to obtain the features of the target picture. The features include circularity features, similarity features, and texture features. The features are converted into feature vectors and input into the classification model according to the decision function. , obtain the on-site status, the sign () represents the sign function, the z represents the eigenvector, and the b represents the bias constant, and the alarm demand is obtained according to the on-site status.

Specifically, the method for obtaining the fire signal value in S1 is:

pass Calculate the temperature signal value, where,/> represents the temperature signal value, ξ represents the sensitivity of the temperature sensor, and a represents the linear relationship between voltage and temperature;

The current sensor obtains the residual current signal value according to Ohm's law;

The smoke concentration sensor passes an analog electrical signal, according to , calculate the smoke concentration signal, where C is the smoke concentration signal value, Rs is the resistance of the sensitive element of the smoke concentration sensor, m, n are the sensitivity constants of the smoke concentration sensor, log() is the logarithmic sign;

The CO concentration sensor uses SnO ₂ with low conductivity as a gas-sensitive material, and uses a high and low temperature cycle detection method to obtain the CO concentration signal value.

Specifically, the specific implementation steps of S2 are:

The signal value set includes a temperature signal value set , smoke signal value set/> , CO concentration signal value set/> , Leakage signal value set/> ,

According to the set of signal values Signal value in/> , n represents the total number of signal values, through , calculate the weighting factor of the signal value, where/> represents the weighting factor,/> Represents the measured variance of the signal value, by/> , the same kind of signal fusion result is calculated, where,/> Indicates the fusion result of the same type of signals,/> Represents the expectation for the set of signal values.

Specifically, the specific implementation steps of S3 are:

The fusion result of the same kind of signals Input the different categories of signal fusion neural network model to obtain the fire evidence element/> , according to/> , calculate the probability vector of the fire evidence element, where/> represents the probability vector;

To calculate the probability of the fire occurring, the calculation formula is: ,

Among them, P ( D ) is the probability of the fire, K is the conflict coefficient between the fire evidence elements, and ⊕ is the probability fusion symbol.

Specifically, the fire alarm and hierarchical early warning in S4 are:

When the probability of fire occurrence , the alarm does not sound;

When the probability of fire occurrence When, the alarm emits a blue early warning, which indicates the occurrence of a non-fire alarm;

When the probability of fire occurrence When, the alarm issues a yellow warning, which indicates that a fire may have occurred;

When the probability of fire occurrence When, the alarm will issue an orange warning, which indicates that a fire is highly likely to occur;

When the probability of fire occurrence , the alarm will issue a red warning, which indicates that a fire is very likely to occur.

Specifically, the specific steps of preprocessing in S5 are:

Convert the red, green, and blue color space models of the on-site sample pictures into hue, saturation, and intensity models to obtain the first on-site picture;

Use median filtering method to remove noise in the first scene picture to obtain a second scene picture;

Using the image binarization method, the intensity component image of the 0-255 gray level is represented by 0 and 1 pixels, and the second scene picture is converted into the second scene picture according to the saturation of the red component of the flame in the second scene picture as a threshold It is divided into a target picture and a background picture. The pixel corresponding to the target picture is 1, and the pixel corresponding to the background picture is 0.

Specifically, the specific steps of static feature extraction in S5 are:

according to Extract the roundness feature of the target image, where,/> is the roundness feature,/> is the area of the target image,/> is the perimeter of the target image;

according to , extract the similarity features of the target image, where,/> is the similarity feature,/> is the frame of the target picture,/> is the corresponding frame of the frame;

The texture features include: energy value , entropy/> , contrast/> , relevance/> , the formula for extracting the texture features is:

,

Among them, u represents the horizontal coordinate gray value, v the vertical coordinate gray value, Represents the gray value of the coordinate point,/> Represents the normalized gray level co-occurrence matrix element value.

Specifically, the steps in S5 to obtain the alarm requirement are:

When the decision function When , the on-site status is a fire, and when the decision function When the on-site status is that no fire has occurred, the alarm demand is determined according to the on-site status. When the on-site status is that a fire has occurred, the alarm demand is that an alarm needs to be sent out. When the on-site status is the When no fire occurs, the alarm requirement is that no alarm is required.

As shown in Figure 2, the system according to the embodiment of the present invention includes the following modules:

Sensor module, local fusion module, global fusion module, alarm module, image recognition module;

The sensor module: obtains the fire signal through the fire sensor, and obtains the fire signal value according to the fire signal;

The local fusion module: obtains the fire signal value set according to the fire signal value, and fuses the signal values of the same type in the fire signal value set to obtain the same type of signal fusion result;

The global fusion module: input the same type signal fusion result into the different types of signal fusion neural network model to obtain the fire evidence element fusion vector, and obtain the fire occurrence probability according to the fire evidence element fusion vector;

The alarm module: sends the fire occurrence probability to the alarm, and the alarm performs the fire alarm and the hierarchical early warning according to the fire occurrence probability;

The picture recognition module: obtains the on-site sample picture according to the on-site surveillance camera, performs the pre-processing on the on-site sample picture to obtain the target picture, and performs the static feature extraction on the target picture to obtain the Features are converted into feature vectors and input into the classification model to obtain the on-site status, and the alarm requirements are obtained based on the on-site status.

The above are only preferred embodiments of the present invention, and do not limit the present invention in any form. Although the present invention has been disclosed above in preferred embodiments, it is not intended to limit the present invention. Anyone skilled in the art , without departing from the scope of the technical solution of the present invention, the technical contents disclosed above can be used to make some changes or modifications to equivalent embodiments with equivalent changes. However, without departing from the technical solution of the present invention, according to the technical solution of the present invention, In essence, any brief modifications, equivalent changes and modifications made to the above embodiments still fall within the scope of the technical solution of the present invention.

Claims (9)

1. A hierarchical early warning method for alarm information based on sensors and pictures, which is characterized by including the following steps: S1: Acquire fire signals through fire sensors. The fire sensors include: temperature sensors, residual current sensors, smoke concentration sensors, and CO concentration sensors. The fire signals are processed to obtain fire signal values; S2: Obtain a fire signal value set D according to the fire signal value. The fire signal value set D includes a signal value set. , the signal value set/> Including n signal values x , the fire signal value set D is expressed as: ,/> , for the signal value set/> The signal values in are weighted and fused to obtain similar signal fusion results/> ; S3: Fusion results of the similar signals Input signals from different categories to the neural network model to obtain fire evidence elements. , according to the fire evidence element/> Get the fire evidence element fusion vector/> , based on the fire evidence element fusion vector/> Get the probability of fire/> ; S4: Put the probability of fire occurrence into Sent to the alarm, which is based on the probability of fire occurrence/> , carry out fire alarm and graded early warning, and the graded early warning includes blue early warning, yellow early warning, orange early warning, and red early warning; S5: When the alarm sounds, obtain on-site sample pictures based on on-site surveillance cameras, perform preprocessing on the on-site sample pictures to obtain target pictures, the preprocessing includes: color conversion, denoising, filtering and segmentation, and perform preprocessing on the target pictures. Perform static feature extraction to obtain the features of the target picture. The features include circularity features, similarity features, and texture features. The features are converted into feature vectors and input into the classification model according to the decision function. , obtain the on-site status, the sign () represents the sign function, the z represents the eigenvector, and the b represents the bias constant, and the alarm demand is obtained according to the on-site status. 2. The method according to claim 1, characterized in that the method for obtaining the fire signal value in S1 is: pass Calculate the temperature signal value, where,/> represents the temperature signal value, ξ represents the sensitivity of the temperature sensor, and a represents the linear relationship between voltage and temperature; The current sensor obtains the residual current signal value according to Ohm's law; The smoke concentration sensor passes an analog electrical signal, according to , calculate the smoke concentration signal, where C is the smoke concentration signal value, Rs is the resistance of the sensitive element of the smoke concentration sensor, m, n are the sensitivity constants of the smoke concentration sensor, log() is the logarithmic sign; The CO concentration sensor uses SnO ₂ with low conductivity as a gas-sensitive material, and uses a high and low temperature cycle detection method to obtain the CO concentration signal value. 3. The method according to claim 1, characterized in that the specific implementation steps of S2 are: The signal value set includes a temperature signal value set , smoke signal value set/> , CO concentration signal value set/> , Leakage signal value set/> , According to the set of signal values Signal value in/> , n represents the total number of signal values, through , calculate the weighting factor of the signal value, where/> represents the weighting factor,/> Represents the measured variance of the signal value, by/> , the same kind of signal fusion result is calculated, where,/> Indicates the fusion result of the same type of signals,/> Represents the expectation for the set of signal values. 4. The method according to claim 1, characterized in that the specific implementation steps of S3 are: The fusion result of the same kind of signals Input the different categories of signal fusion neural network model to obtain the fire evidence element/> , according to/> , calculate the probability vector of the fire evidence element, where represents the probability vector; To calculate the probability of the fire occurring, the calculation formula is: , Among them, P ( D ) is the probability of the fire, K is the conflict coefficient between the fire evidence elements, and ⊕ is the probability fusion symbol. 5. The method according to claim 1, characterized in that the fire alarm and the hierarchical early warning in S4 are: When the probability of fire occurrence , the alarm does not sound; When the probability of fire occurrence When, the alarm emits a blue early warning, which indicates the occurrence of a non-fire alarm; When the probability of fire occurrence When, the alarm issues a yellow warning, which indicates that a fire may have occurred; When the probability of fire occurrence When, the alarm will issue an orange warning, which indicates that a fire is highly likely to occur; When the probability of fire occurrence , the alarm will issue a red warning, which indicates that a fire is very likely to occur. 6. The method according to claim 1, characterized in that the specific steps of preprocessing in S5 are: Convert the red, green, and blue color space models of the on-site sample pictures into hue, saturation, and intensity models to obtain the first on-site picture; Use median filtering method to remove noise in the first scene picture to obtain a second scene picture; Using the image binarization method, the intensity component image of the 0-255 gray level is represented by 0 and 1 pixels, and the second scene picture is converted into the second scene picture according to the saturation of the red component of the flame in the second scene picture as a threshold It is divided into a target picture and a background picture. The pixel corresponding to the target picture is 1, and the pixel corresponding to the background picture is 0. 7. The method according to claim 1, characterized in that the specific steps of static feature extraction in S5 are: according to , extract the roundness feature of the target image, where,/> is the roundness feature,/> is the area of the target image,/> is the perimeter of the target image; according to , extract the similarity features of the target image, where,/> is the similarity feature,/> is the frame of the target picture,/> is the corresponding frame of the frame; The texture features include: energy value , entropy/> , contrast/> , relevance/> , the formula for extracting the texture features is: , Among them, u represents the horizontal coordinate gray value, v the vertical coordinate gray value, Represents the gray value of the coordinate point,/> Represents the normalized gray level co-occurrence matrix element value. 8. The method according to claim 1, characterized in that the step of obtaining the alarm requirement in S5 is: When the decision function When , the on-site status is a fire, and when the decision function/> When the on-site status is that no fire has occurred, the alarm demand is determined according to the on-site status. When the on-site status is that a fire has occurred, the alarm demand is that an alarm needs to be sent out. When the on-site status is the When no fire occurs, the alarm requirement is that no alarm is required. 9. An alarm information hierarchical early warning system based on sensors and pictures, operating using the method described in any one of claims 1-8, characterized in that it includes the following modules: Sensor module, local fusion module, global fusion module, alarm module, image recognition module; The sensor module: obtains the fire signal through the fire sensor, and obtains the fire signal value according to the fire signal; The local fusion module: obtains the fire signal value set according to the fire signal value, and fuses the signal values of the same type in the fire signal value set to obtain the same type of signal fusion result; The global fusion module: input the same type signal fusion result into the different types of signal fusion neural network model to obtain the fire evidence element fusion vector, and obtain the fire occurrence probability according to the fire evidence element fusion vector; The alarm module: sends the fire occurrence probability to the alarm, and the alarm performs the fire alarm and the hierarchical early warning according to the fire occurrence probability; The picture recognition module: obtains the on-site sample picture according to the on-site surveillance camera, performs the pre-processing on the on-site sample picture to obtain the target picture, and performs the static feature extraction on the target picture to obtain the Features are converted into feature vectors and input into the classification model to obtain the on-site status, and the alarm requirements are obtained based on the on-site status.