Method for enhancing intelligent construction site fire recognition capability by using AI large model

Abstract

The invention discloses a method for enhancing intelligent construction site fire recognition capability by using an AI large model, relates to the field of fire recognition, and solves the technical problems of limited detection distance and insufficient blind fire detection in the existing fire recognition technology. The method comprises a blind fire identification model construction step and a blind fire identification analysis step, and the blind fire identification model construction step comprises the following steps: S11, collecting a blind fire training sample which comprises first characteristic values of a hidden heat source, an air expansion distortion phenomenon and a light refraction phenomenon; and S12, training the blind fire training sample through a deep convolutional neural network model. Compared with a traditional fire identification algorithm, the method reduces the dependence on a smoke sensing alarm and a temperature sensing fire detector, obviously increases the detection distance of the fire, can achieve the timely blind fire early warning, can prevent a part of the fire, and guarantees the project progress and construction safety.

Description

Technical Field

[0001] The present invention relates to the field of fire identification, and more specifically, to a method for enhancing the fire identification capability of a smart construction site by using an AI big model. Background Art

[0002] There are usually a large number of workers and managers on construction sites. Once a fire occurs, it may cause serious casualties and seriously affect the progress of the project and construction safety. Therefore, effective fire prevention measures can protect the lives of workers and are also an important measure to ensure construction safety. At present, the fire prevention measures on construction sites mainly include implementing fire management systems, rationally planning construction site layouts, equipping with sufficient firefighting equipment, and strengthening safety education. However, in terms of early detection and prevention of fires, it is still at a relatively weak stage.

[0003] At present, fire identification in smart construction sites mainly relies on high-definition cameras and supporting flame identification algorithms, and cooperates with smoke alarms and temperature-sensitive fire detectors for fire detection and identification. The current flame detection and identification technology is highly dependent on the quality of the camera and the efficiency and accuracy of the flame identification algorithm. In addition, the detection distance of smoke alarms and temperature-sensitive fire detectors is limited and cannot be detected at a long distance.

[0004] The current fire identification solutions for smart construction sites are also very inadequate in detecting hidden fires, or even unable to detect them at all. Due to their concealed nature, hidden fires have no obvious flames or smoke and may not be discovered for a long time. Moreover, when hidden fires develop to a certain extent, they are very likely to turn into open flames and eventually cause fires. Summary of the invention

[0005] The technical problem to be solved by the present invention is to address the shortcomings of the existing technology and provide a method for enhancing the fire identification capability of smart construction sites by using a large AI model, thereby solving the technical problems of the existing fire identification technology with limited detection distance and insufficient detection of hidden fires.

[0006] The present invention discloses a method for enhancing the fire identification capability of a smart construction site by using an AI large model, the method comprising a dark fire identification model construction step and a dark fire identification analysis step:

[0007] The dark fire identification model construction step includes:

[0008] S11: Collecting dark fire training samples, where the dark fire training samples include first characteristic values ​​of hidden heat sources, air expansion and distortion phenomena, and light refraction phenomena;

[0009] S12: training the Dark Fire training sample by a deep convolutional neural network model to extract a first eigenvalue from the Dark Fire training sample, and evaluating the first eigenvalue by a first standard feature set to obtain a first optimal eigenvalue;

[0010] S13: Generate a dark fire recognition model using the first optimal characteristic value as model data;

[0011] The dark fire identification and analysis step comprises:

[0012] S21: Collecting environmental video of the actual environment;

[0013] S22: the dark fire recognition model detects whether the video frame of the environment video contains air distortion and expansion phenomenon and light refraction phenomenon according to the first optimal characteristic value, and if so, proceeds to S23;

[0014] S23: analyzing the air distortion and expansion phenomenon by using the dark fire recognition model to obtain light parameters of the position where the air distortion and expansion phenomenon occurs in the video frame, and analyzing the heat source position and heat source temperature by using the light parameters;

[0015] S24: Obtain the size of the heat source according to the heat source position and the heat source temperature, and analyze the danger level of the heat source according to the heat source temperature and the heat source size, and when the danger level reaches a preset level threshold, trigger a dark fire alarm.

[0016] As a further improvement, in S12, the deep convolutional neural network model is trained to detect the air expansion and distortion phenomenon in the dark fire training samples by the following formula:

[0017]

[0018] Among them, D( χ ,n) represents the air expansion calculation function, and Indicates the nth frame image χ The coordinate vector of θ( χ )express and The angle value that produces the angle in the coordinate system, χ Represents a random coordinate point.

[0019] Further, in S12, evaluating the first eigenvalue by using the first standard feature set to obtain a first optimal eigenvalue specifically includes:

[0020] The first eigenvalue is compared with the first standard eigenvalues ​​in the first standard feature set one by one. If there is a first standard eigenvalue that is consistent with the first eigenvalue, the first eigenvalue is used as the first optimal eigenvalue.

[0021] Furthermore, it is detected whether the video frame contains air distortion and expansion phenomena and light refraction phenomena, specifically: the video frame is traversed according to the air expansion and distortion phenomena and light refraction phenomena in the first optimal eigenvalue, if the video frame has characteristic phenomena that are consistent with the air expansion and distortion phenomena and light refraction phenomena in the first optimal eigenvalue, it is determined that air distortion and expansion phenomena and light refraction phenomena exist.

[0022] Furthermore, in S22, light parameters are obtained by analyzing the phenomenon of air distortion and expansion, specifically:

[0023] Acquiring the temperature and pressure in the actual environment, and calculating the refractive index of the light according to the temperature and pressure;

[0024] The refraction angle of the light is calculated according to the refractive index.

[0025] Furthermore, in S22, the heat source position and the heat source temperature are analyzed by the light parameters, specifically: the light parameters are compared with the light parameter threshold of the light refraction phenomenon in the dark fire recognition model, if the light parameters are consistent with the light parameter threshold, or the light parameters exceed the light parameter threshold, then the position corresponding to the light parameters is judged to be the heat source position, and the heat source temperature of the heat source position is obtained at the same time.

[0026] Furthermore, the method also includes an open flame recognition model construction step and an open flame recognition analysis step;

[0027] The open fire recognition model construction step specifically includes:

[0028] S31: Collecting open fire training samples, where the open fire training samples include second characteristic values ​​of flame and smoke;

[0029] S32: training the open fire training sample by a deep convolutional neural network model to extract a second eigenvalue from the open fire training sample, and evaluating the second eigenvalue by a second standard feature set to obtain a second optimal eigenvalue;

[0030] S33: Generate an open flame recognition model using the second optimal characteristic value as model data;

[0031] The open fire identification and analysis step specifically includes:

[0032] The open fire recognition model detects whether there is flame and / or smoke in the video frame of the environmental video according to the second optimal eigenvalue; if so, obtains the smoke eigenvalue, motion eigenvalue and texture eigenvalue, and determines whether it is necessary to trigger an open fire alarm according to the smoke eigenvalue, motion eigenvalue and texture eigenvalue;

[0033] Furthermore, the detection expression of the smoke characteristic value is:

[0034]

[0035] Among them, U (χ,n) Refers to the chromaticity value of the pixel at position x in the nth frame image I, V (χ,n) Refers to the concentration value of the pixel at position x in the nth frame image I,

[0036] Set a brightness threshold T1, and the following formula is to obtain the brightness Y of the pixel at position x in the nth frame image I: (χ,n) The expression is:

[0037]

[0038] If the brightness Y of the pixel at position x in the nth frame image I (χ,n) When the brightness threshold T1 is less than the value of the smoke characteristic value P(χ,n) is -1; when the brightness Y (χ,n) When it is greater than T1, the smoke characteristic value P(χ,n) is

[0039] Furthermore, the detection expression of the motion feature is:

[0040]

[0041] Where I(χ,n) refers to the intensity value of the pixel at position χ in the nth frame image I, a is a smoothing constant between (0,1), B(χ,n+1) refers to the motion feature value of the n+1th frame image, and B(χ,n) refers to the motion feature value of the nth frame image;

[0042] When the motion feature value B(χ,n) of the nth frame image is greater than the preset motion feature threshold, B(χ,n+1)=B(χ,n); when the motion feature value B(χ,n) of the nth frame image is less than or equal to the motion feature threshold,

[0043] B(χ,n+1)=aB(χ,n)+(1-a)I(χ,n);

[0044] The texture feature detection formula:

[0045]

[0046] Among them, B fast (χ,n) and B slow (χ,n) refers to the way of updating at different rates, B fast (χ,n) is updated once per frame, B slow(χ,n) is updated once per second, T high and T low Refers to the different brightness values ​​set in the image attributes, A(χ,n) is the texture feature value; set a texture feature threshold range, when the rate difference B 1＝ |B fast (χ,n)-B slow When (χ,n)| is greater than the texture feature threshold range, the texture feature value A(χ,n) is equal to 1; when the rate difference B1 is less than the texture feature threshold range, the texture feature value A(χ,n) is equal to -1; when the rate difference B1 is within the texture feature threshold range, the rate difference B1 is substituted into Obtain the texture feature value A(χ,n).

[0047] Furthermore, it is determined whether an open fire alarm needs to be triggered based on the smoke feature values, motion feature values ​​and texture feature values. Specifically, the smoke feature values, motion feature values ​​and texture feature values ​​in the environmental video are compared one by one with the smoke standard feature values, motion standard feature values ​​and texture standard feature values ​​in the second standard feature set. If there are standard features that are inconsistent with the second standard feature set, an open fire alarm is triggered.

[0048] Beneficial Effects

[0049] The advantages of the present invention are:

[0050] The present invention sets an open flame recognition model construction step, an open flame recognition analysis step, a dark fire recognition model construction step and a dark fire recognition analysis step, and generates a neural network model by repeatedly training the collected training samples to obtain the location and range of the fire, thereby enhancing the flame recognition ability of the camera. Compared with the traditional fire recognition algorithm, the dependence on smoke alarms and temperature fire detectors is reduced, and the detection distance of the fire is significantly increased. Timely dark fire warning can prevent the occurrence of some fires, thereby ensuring the project progress and construction safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a flow chart of the dark fire identification steps of the present invention;

[0052] Figure 2 It is a flow chart of the open flame identification steps of the present invention;

[0053] Figure 3 It is a schematic diagram for deriving the light path calculation formula in the present invention. DETAILED DESCRIPTION

[0054] The present invention will be further described below in conjunction with the embodiments, but it does not constitute any limitation to the present invention. Any limited number of modifications made by anyone within the scope of the claims of the present invention are still within the scope of the claims of the present invention.

[0055] See also Figure 1-Figure 3 The present invention provides a method for enhancing the fire identification capability of a smart construction site by using an AI large model. The method includes a dark fire identification model construction step and a dark fire identification analysis step.

[0056] like Figure 1 As shown, the dark fire identification step includes:

[0057] The method comprises a dark fire identification model construction step and a dark fire identification analysis step.

[0058] The steps to build the dark fire recognition model include:

[0059] S11: Collect dark fire training samples, where the dark fire training samples include first eigenvalues ​​of hidden heat sources, air expansion and distortion phenomena, and light refraction phenomena.

[0060] S12: Training the Dark Fire training samples through a deep convolutional neural network model to extract a first eigenvalue from the Dark Fire training samples, and evaluating the first eigenvalue through a first standard feature set to obtain a first optimal eigenvalue.

[0061] In S12, the deep convolutional neural network model is trained to detect the air expansion and distortion phenomenon in the dark fire training samples using the following formula:

[0062]

[0063] Where χ represents a random coordinate point; and represents the coordinate vector of χ in the nth frame image, θ(χ) represents and The angle value generated in the coordinate system; when θ(χ) is small, the value of the air expansion calculation function D(χ, n) is close to -1.

[0064] In S12, the first eigenvalue is evaluated by the first standard feature set to obtain a first optimal eigenvalue, specifically including comparing the first eigenvalue with the first standard eigenvalues ​​in the first standard feature set one by one, and if there is a first standard eigenvalue consistent with the first eigenvalue, the first eigenvalue is used as the first optimal eigenvalue.

[0065] S13: Generate a dark fire recognition model using the first optimal eigenvalue as model data.

[0066] The dark fire identification and analysis steps include:

[0067] S21: Collect environmental video of the actual environment.

[0068] S22: The dark fire recognition model detects whether the video frame of the environmental video contains air distortion and expansion phenomenon and light refraction phenomenon according to the first optimal eigenvalue. If so, proceed to S23.

[0069] In S22, light parameters are obtained by analyzing the phenomenon of air distortion and expansion, specifically:

[0070] Obtain the temperature and pressure in the actual environment, and calculate the refractive index of light based on the temperature and pressure;

[0071] Calculate the refraction angle of light according to the refractive index;

[0072] The light path of light in a continuously changing medium is obtained by the following formula:

[0073]

[0074] like Figure 3 As shown, in a continuously changing medium, the light path can be described in integral form. The above formula is derived from Monte Carlo integration:

[0075] exist Figure 3 The middle ab interval (i.e. Figure 3 Take any point x in the middle gray area and get its corresponding f(x). Then, we consider the function graph to be a horizontal line, that is, its area becomes a rectangle with a base length of (ba) and a height of f(x). We use the area of ​​such a rectangle to approximate the area of ​​the irregular figure. Then, we repeat this process and get the areas of many rectangles. By adding them up and averaging them, we can get the approximate area of ​​the function graph. Each time we get a rectangle, it is a sampling process. That is to say, if we get the area of ​​the function graph, we can get the area of ​​the function graph. Figure 3 If this function value is sampled uniformly multiple times, it is actually equivalent to cutting the entire integral area into many rectangles and then adding up the areas of these small rectangles.

[0076] This leads to the definition of the definite integral function:

[0077]

[0078] Uniform random variable:

[0079]

[0080] This leads to the Monte Carlo integral equation:

[0081]

[0082] In order to get a smaller error value, importance sampling should also be considered. Of course, even if uniform distribution sampling is used, as long as there are enough sampling points, a relatively accurate result can still be obtained, but the performance overhead may increase, so uniform distribution is generally not used for sampling, but a sampling distribution function that fits the overall shape and the integrand is used to derive a more general formula:

[0083]

[0084] The ability to identify dark fires is enhanced. The principle is mainly based on the expansion and distortion of air. Although dark fires do not have flames or smoke, there must be a heat source. As the temperature of the heat source rises, the nearby air expands due to the heat and its density decreases, while the cold air around the heat source has a higher density. This density difference will cause light to refract during propagation, thereby creating a visual effect of air expansion and distortion.

[0085] Air expansion coefficient calculation formula:

[0086] a=(1 / v)*(dV / dT);

[0087] Wherein, a represents the coefficient of thermal expansion, V represents the volume of the gas, T represents the temperature, dV represents the change in volume, and dT represents the change in temperature.

[0088] Refractive index calculation formula:

[0089] n = c / v;

[0090] Among them, n is the refractive index, c is the speed of light, and v is the speed of light propagation in the medium.

[0091] The approximate calculation formula for atmospheric refraction under experimental conditions is:

[0092]

[0093] where n s is the refractive index of air at standard conditions, and n (T ,p ) It is the refractive index at a certain temperature and pressure, where the pressure p is in torr (1 torr = 1 / 760 atm) and the temperature T is in degrees Celsius.

[0094] Snell's law calculation formula:

[0095] n sin(θ1)=n′sin(θ2);

[0096] Where n is the refractive index of the incident medium, n' is the refractive index of the exit medium, θ1 is the incident angle, and θ2 is the refraction angle.

[0097] In S22, the heat source position and heat source temperature are analyzed by light parameters, specifically: the light parameters are compared with the light parameter threshold of the light refraction phenomenon in the dark fire recognition model. If the light parameters are consistent with the light parameter threshold, or the light parameters exceed the light parameter threshold, the position corresponding to the light parameters is judged to be the heat source position, and the heat source temperature of the heat source position is obtained at the same time.

[0098] A pre-trained convolutional neural network (CNN) model is used to train a large amount of video data with heat sources and air expansion and distortion. During the training process, the large model extracts information such as air expansion and distortion, light refraction, etc. in the video frames to form model data of dark fire heat source characteristics.

[0099] S23: Analyze the air distortion and expansion phenomenon through the dark fire recognition model to obtain light parameters of the position where the air distortion and expansion phenomenon occurs in the video frame, and analyze the heat source position and heat source temperature through the light parameters.

[0100] Detect whether the video frame contains air distortion and expansion phenomenon and light refraction phenomenon, specifically: traverse the video frame according to the air expansion and distortion phenomenon and light refraction phenomenon in the first optimal eigenvalue, if there are characteristic phenomena in the video frame that are consistent with the air expansion and distortion phenomenon and light refraction phenomenon in the first optimal eigenvalue, it is determined that the air distortion and expansion phenomenon and light refraction phenomenon exist.

[0101] S24: Obtain the size of the heat source according to the location and temperature of the heat source, and analyze the danger level of the heat source according to the temperature and size of the heat source, and when the danger level reaches a preset level threshold, trigger a dark fire alarm.

[0102] like Figure 2 As shown, the method further includes an open fire identification model building step and an open fire identification analysis step;

[0103] The steps to build the open flame recognition model include:

[0104] S31: Collecting open fire training samples, where the open fire training samples include second eigenvalues ​​of flame and smoke;

[0105] S32: Training the open flame training samples through a deep convolutional neural network model to extract a second eigenvalue from the open flame training samples, and evaluating the second eigenvalue through a second standard feature set to obtain a second optimal eigenvalue.

[0106] The principle of enhancing the ability to identify fireworks is mainly based on deep learning and video processing technology of large AI models. As a representative algorithm of deep learning, convolutional neural networks have representation learning capabilities and can extract high-order characteristics from input information. In the fireworks recognition model, convolutional neural networks extract representative high-frequency and low-frequency features in the image and extract more general and complete features from the signals in the hidden layer.

[0107] The convolution operation is defined as:

[0108] [(I*K)(x, y)=sum{i=-infty}^{infty}sum{j=-infty}^{infty}I(i, j)K(xi, yj)]:

[0109] Where (I) is the input image, (K) is the convolution kernel, and (x, y) represents the location of the output feature map.

[0110] The difference between a convolutional neural network and a common neural network is that a convolutional neural network contains multiple feature extractors consisting of convolutional layers and pooling layers. In the convolutional layer of a convolutional neural network, a neuron is only connected to some neurons in the adjacent layers. In a convolutional layer of a CNN, there are usually several feature planes (featureMaps), each of which is composed of some neurons arranged in a rectangular shape. The neurons in the same feature plane share weights, and the shared weights are the convolution kernels. The convolution kernels are generally initialized in the form of random decimal matrices, and the convolution kernels will learn to obtain reasonable weights during the training process of the network. The direct benefit of sharing weights (convolution kernels) is to reduce the connections between the layers of the network and reduce the risk of overfitting. Subsampling is also called pooling, which usually has two forms: mean subsampling and max subsampling. Subsampling can be regarded as a special convolution process. Convolution and subsampling greatly simplify the complexity of the model and reduce the parameters of the model.

[0111] Formula derivation of neural network:

[0112]

[0113] This unit can also be called a Logistic regression model. When multiple units are combined and have a hierarchical structure, a neural network model is formed. The training method of the convolutional neural network is similar to that of the Logistic model, but due to its multi-layer nature, the chain rule is also needed to derive the nodes of the hidden layer, that is, gradient descent + chain rule.

[0114] The expression of convolution is:

[0115] S(t)=∫x(ta)w(a)da;

[0116] The discrete form is:

[0117]

[0118] The matrix form is:

[0119] s(t)=(X*W)(t);

[0120] Convolutional Neural Network (CNN) defines two-dimensional convolution as:

[0121]

[0122] Where W is our convolution kernel and X is our input. If X is a two-dimensional input matrix, W is also a two-dimensional matrix. If X is a multi-dimensional tensor, W is also a multi-dimensional tensor.

[0123] Use pre-trained convolutional neural network (CNN) models (such as ResNet, VGG, etc.) as feature extractors to train a large amount of video data with fireworks. During the training process, the large model will extract the characteristic information of fireworks in the video frame, such as color features, motion features, texture features, and shape features, to form model data of fireworks features.

[0124] S33: Generate an open flame recognition model using the second optimal eigenvalue as model data.

[0125] Open flame identification analysis steps:

[0126] The open fire recognition model detects whether there is flame and / or smoke in the video frame of the environmental video based on the second optimal eigenvalue; if so, the smoke eigenvalue, motion eigenvalue and texture eigenvalue are obtained, and it is determined whether an open fire alarm needs to be triggered based on the smoke eigenvalue, motion eigenvalue and texture eigenvalue.

[0127] The detection expression of motion features is:

[0128]

[0129] Where I(χ,n) refers to the intensity value of the pixel at position χ in the nth frame image I, a is a smoothing constant between (0,1), B(χ,n+1) refers to the n+1th frame image, and B(χ,n) refers to the nth frame image.

[0130] When the motion feature value B(χ,n) of the nth frame image is greater than a preset motion feature threshold, B(χ,n+1)=B(χ,n); when the motion feature value B(χ,n) of the nth frame image is less than or equal to the motion feature threshold, B(χ,n+1)=aB(χ,n)+(1-a)I(χ,n).

[0131] The n+1th frame image B(χ,n+1) is estimated by a recursive formula involving frame I(χ,n) and image B(χ,n). The main algorithm used is the linear prediction algorithm. The following is the derivation principle:

[0132] Let B(χ,n) be the observed data of the nth frame, I(χ,n) be the intensity value of the pixel at position χ in the nth frame, and I(χ,n+1) be the predicted intensity value of the pixel at position χ in the n+1th frame. Assume that

[0133] I(χ,n+1)=aB(χ,n)+(1-a)I(χ,n);

[0134] It can be proved

[0135] I(χ,n+1)=aB(χ,n)+a(1-a)B(χ,n-1)+a(1-a) 2 B(χ,n-2)+…+a(1-a) n-1 B(χ, 1)+(1-a) n l0;

[0136] Here, l0=I(χ, 1) is regarded as the initial value.

[0137] Texture feature detection formula:

[0138]

[0139] Among them, B fast (χ,n) and B slow (χ,n) refers to the way of updating at different rates, B fast (χ,n) is updated once per frame, B slow (χ, n) is updated once per second, T high and T low Refers to the different brightness values ​​set in the image properties.

[0140] Among them, B fast (χ,n) and B slow (χ,n) refers to the way of updating at different rates, B fast (χ,n) is updated once per frame, B slow (χ,n) is updated once per second, T high and T low Refers to the different brightness values ​​set in the image attributes, A(χ,n) is the texture feature value; set a texture feature threshold range, when the rate difference B 1＝ |B fast (χ,n)-B slow When (χ,n)| is greater than the texture feature threshold range, the texture feature value A(χ,n) is equal to 1; when the rate difference B1 is less than the texture feature threshold range, the texture feature value A(χ,n) is equal to -1; when the rate difference B1 is within the texture feature threshold range, the rate difference B1 is substituted into The texture feature value A(χ,n) is obtained.

[0141] 0<T low <T high The threshold of is determined by experimental analysis. In the derivation process, we take T low =10,T high =30, then the texture feature threshold range is [10, 30].

[0142] By comparing two different images B fast (χ,n) and B slow (χ,n) is used to detect objects with slowly changing shapes within the camera’s surveillance range. If objects with slowly changing shapes are detected continuously within a certain period of time, an alarm will be generated, the area will be marked, and further detection will be performed.

[0143] Determine whether there is smoke according to the following formula:

[0144]

[0145] The main components of smoke are carbon dioxide, water vapor, carbon monoxide, trace minerals, and dust, so it appears gray and black. These smoke areas can be detected by setting the threshold in the YUV color space. (χ,n) Refers to the brightness value of the pixel at position x in the nth frame image I, U (χ,n) Refers to the chromaticity value of the pixel at position x in the nth frame image I, V (χ,n) Refers to the concentration value of the pixel at position x in the nth frame image I.

[0146] Among them, U (χ,n) Refers to the chromaticity value of the pixel at position x in the nth frame image I, V (χ,n) Refers to the concentration value of the pixel at position x in the nth frame image I. Set a brightness threshold T1. The following formula is used to obtain the brightness Y of the pixel at position x in the nth frame image I. (χ,n) The expression is:

[0147]

[0148] If the brightness Y of the pixel at position x in the nth frame image I (χ,n) When the brightness threshold T1 is less than the value of the smoke characteristic value P (χ,n) The value of is -1; when the brightness Y of the pixel at position x in the nth frame image I is (χ,n) When it is greater than T1, the chromaticity value U of the pixel at position x in the nth frame image I is (χ,n) and the concentration value V in the pixel at position x in the nth frame image I (χ,n) are both 128, then the smoke characteristic value P (χ,n)The value is 1. The brightness threshold T1 is set to exclude dark areas with low chromaticity values, because the colors in the area where fireworks exist may not have a clear contrast. For the case where the brightness value is higher than T1, the chromaticity value U of the pixel at position x in the nth frame image I is (χ,n) and the concentration value V of the pixel at position x in the nth frame image I (χ,n) When the mean is around 128, the chromaticity value U of the pixel at position x in the nth frame image I is (χ,n) and the concentration value V of the pixel at position x in the nth frame image I (χ,n) Substitution In, P (χ,n) The value is close to 1.

[0149] Determine whether it is necessary to trigger an open fire alarm based on the smoke feature values, motion feature values ​​and texture feature values. Specifically, compare the smoke feature values, motion feature values ​​and texture feature values ​​in the ambient video with the smoke standard feature values, motion standard feature values ​​and texture standard feature values ​​in the second standard feature set one by one. If there are standard features that are inconsistent with those in the second standard feature set, an open fire alarm is triggered.

[0150] Through the above steps, a training model for open flame recognition based on convolutional neural network can be built to achieve efficient and accurate open flame recognition, and the trained model can be deployed to actual application scenarios.

[0151] The above is only a preferred embodiment of the present invention. It should be pointed out that for those skilled in the art, several modifications and improvements can be made without departing from the structure of the present invention, which will not affect the effect of the implementation of the present invention and the practicality of the patent.

Claims

1. A method for enhancing the fire identification capability of smart construction sites using an AI big model, characterized in that: The method includes a dark fire identification model construction step and a dark fire identification analysis step: The dark fire identification model building step includes: S11: Collecting dark fire training samples, where the dark fire training samples include first eigenvalues ​​of hidden heat sources, air expansion and distortion phenomena, and light refraction phenomena; S12: training the Dark Fire training sample by a deep convolutional neural network model to extract a first eigenvalue from the Dark Fire training sample, and evaluating the first eigenvalue by a first standard feature set to obtain a first optimal eigenvalue; S13: Generate a dark fire recognition model using the first optimal characteristic value as model data; The dark fire identification and analysis step comprises: S21: Collecting environmental video of the actual environment; S22: the dark fire recognition model detects whether the video frame of the environment video contains air distortion and expansion phenomenon and light refraction phenomenon according to the first optimal characteristic value, and if so, proceeds to S23; S23: analyzing the air distortion and expansion phenomenon by using the dark fire recognition model to obtain light parameters of the position where the air distortion and expansion phenomenon occurs in the video frame, and analyzing the heat source position and heat source temperature by using the light parameters; S24: Obtain the size of the heat source according to the heat source position and the heat source temperature, and analyze the danger level of the heat source according to the heat source temperature and the heat source size, and when the danger level reaches a preset level threshold, trigger a dark fire alarm.

2. According to claim 1, a method for enhancing the fire identification capability of a smart construction site using an AI large model is characterized in that: In S12, the deep convolutional neural network model is trained to detect the air expansion and distortion phenomenon in the dark fire training sample by the following formula: Among them, D( χ ,n) represents the air expansion calculation function, and Indicates the nth frame image χ The coordinate vector of χ )express and The angle value that produces the angle in the coordinate system, χ Represents a random coordinate point.

3. The method of using an AI big model to enhance the fire identification capability of a smart construction site according to claim 1 is characterized in that: In S12, evaluating the first eigenvalue by using the first standard feature set to obtain a first optimal eigenvalue specifically includes: The first eigenvalue is compared with the first standard eigenvalues ​​in the first standard feature set one by one. If there is a first standard eigenvalue that is consistent with the first eigenvalue, the first eigenvalue is used as the first optimal eigenvalue.

4. The method of using an AI large model to enhance the fire identification capability of a smart construction site according to claim 1 is characterized in that: Detect whether the video frame contains air distortion and expansion phenomenon and light refraction phenomenon, specifically: traverse the video frame according to the air expansion and distortion phenomenon and light refraction phenomenon in the first optimal eigenvalue, if the video frame has characteristic phenomena consistent with the air expansion and distortion phenomenon and light refraction phenomenon in the first optimal eigenvalue, it is determined that the air distortion and expansion phenomenon and light refraction phenomenon exist.

5. The method of using an AI big model to enhance the fire identification capability of a smart construction site according to claim 4 is characterized in that: In S22, light parameters are obtained by analyzing the phenomenon of air distortion and expansion, specifically: the temperature and pressure in the actual environment are obtained, and the refractive index of the light is calculated according to the temperature and pressure; and the refractive angle of the light is calculated according to the refractive index.

6. The method of using an AI big model to enhance the fire identification capability of a smart construction site according to claim 1 is characterized in that: In S22, the heat source position and the heat source temperature are analyzed by the light parameters, specifically: the light parameters are compared with the light parameter threshold of the light refraction phenomenon in the dark fire recognition model, if the light parameters are consistent with the light parameter threshold, or the light parameters exceed the light parameter threshold, then the position corresponding to the light parameters is judged to be the heat source position, and the heat source temperature of the heat source position is obtained at the same time.

7. The method of using an AI big model to enhance the fire identification capability of a smart construction site according to claim 1 is characterized in that: The method also includes an open flame recognition model construction step and an open flame recognition analysis step; The open fire recognition model construction step specifically includes: S31: Collecting open fire training samples, where the open fire training samples include second characteristic values ​​of flame and smoke; S32: training the open fire training sample by a deep convolutional neural network model to extract a second eigenvalue from the open fire training sample, and evaluating the second eigenvalue by a second standard feature set to obtain a second optimal eigenvalue; S33: Generate an open flame recognition model using the second optimal characteristic value as model data; The open fire identification and analysis step specifically includes: The open fire recognition model detects whether there is flame and / or smoke in the video frame of the environmental video according to the second optimal eigenvalue; If so, obtain the smoke feature value, motion feature value and texture feature value, and determine whether it is necessary to trigger an open fire alarm based on the smoke feature value, motion feature value and texture feature value.

8. The method of using an AI big model to enhance the fire identification capability of a smart construction site according to claim 7 is characterized in that: The detection expression of the smoke characteristic value is: Among them, U (χ,n) Refers to the chromaticity value of the pixel at position x in the nth frame image I, V (χ,n) It refers to the concentration value of the pixel at position χ in the n-th frame image I, and P(χ,n) is the smoke characteristic value of the pixel at position χ in the n-th frame image I. Set a brightness threshold T1, and the following formula is to obtain the brightness Y of the pixel at position x in the nth frame image I: (χ,n) The expression is: When the brightness Y of the pixel at position x in the nth frame image I (χ,n) When the brightness threshold T1 is less than the value of the smoke characteristic value P(χ,n) is -1; when the brightness Y (χ,n) When it is greater than the brightness threshold T1, the smoke characteristic value P(χ,n) is 9. The method of using an AI big model to enhance the fire identification capability of a smart construction site according to claim 6 is characterized in that: The detection expression of the motion feature is: Where I(χ,n) refers to the intensity value of the pixel at position χ in the nth frame image I, a is a smoothing constant between (0,1), B(χ,n+1) refers to the motion feature value of the n+1th frame image, and B(χ,n) refers to the motion feature value of the nth frame image; When the motion feature value B(χ,n) of the nth frame image is greater than the preset motion feature threshold, B(χ,n+1)=B(χ,n); when the motion feature value B(χ,n) of the nth frame image is less than or equal to the motion feature threshold, B(χ,n+1)=aB(χ,n)+(1-a)I(χ,n); The texture feature detection formula: Among them, B fast (χ,n) and B slow (χ,n) refers to the way of updating at different rates, B fast (χ,n) is updated once per frame, B slow (χ,n) is updated once per second, T high and T low Refers to the different brightness values ​​set in the image attributes, A(χ,n) is the texture feature value; set a texture feature threshold range, when the rate difference B1=|B fast (χ,n)-B slow When (χ,n)| is greater than the texture feature threshold range, the texture feature value A(χ,n) is equal to 1; when the rate difference B1 is less than the texture feature threshold range, the texture feature value A(χ,n) is equal to -1; when the rate difference B1 is within the texture feature threshold range, the rate difference B1 is substituted into The texture feature value A(χ,n) is obtained.

10. The method of using an AI big model to enhance the fire identification capability of a smart construction site according to claim 6, characterized in that: Whether it is necessary to trigger an open fire alarm is determined based on the smoke feature value, motion feature value and texture feature value, specifically: the smoke feature value, motion feature value and texture feature value in the environmental video are compared one by one with the smoke standard feature value, motion standard feature value and texture standard feature value in the second standard feature set. If there are standard features that are inconsistent with the second standard feature set, an open fire alarm is triggered.

Images